Research ArticleGEOCHEMISTRY

Survival of presolar p-nuclide carriers in the nebula revealed by stepwise leaching of Allende refractory inclusions

See allHide authors and affiliations

Science Advances  09 Jul 2021:
Vol. 7, no. 28, eabf6222
DOI: 10.1126/sciadv.abf6222


The 87Rb-87Sr radiochronometer provides key insights into the timing of volatile element depletion in planetary bodies, yet the unknown nucleosynthetic origin of Sr anomalies in Ca-Al–rich inclusions (CAIs, the oldest dated solar system solids) challenges the reliability of resulting chronological interpretations. To identify the nature of these Sr anomalies, we performed step-leaching experiments on nine unmelted CAIs from Allende. In six CAIs, the chemically resistant residues (0.06 to 9.7% total CAI Sr) show extreme positive μ84Sr (up to +80,655) and 87Sr variations that cannot be explained by decay of 87Rb. The extreme 84Sr but more subdued 87Sr anomalies are best explained by the presence of a presolar carrier enriched in the p-nuclide 84Sr. We argue that this unidentified carrier controls the isotopic anomalies in bulk CAIs and outer solar system materials, which reinstates the chronological significance of differences in initial 87Sr/86Sr between CAIs and volatile-depleted inner solar system materials.


Of the refractory elements enriched in calcium-aluminum–rich inclusions [CAIs; the oldest dated solids in the solar system: ~4.567 billion years (Ga)] (1, 2), strontium (Sr) holds a central place because its isotopic composition is controlled by three fundamentally different processes: (i) mixing of presolar carriers with distinct isotopic composition, (ii) radiogenic ingrowth of 87Sr from the decay of 87Rb, and (iii) mass-dependent isotope fractionation and/or resetting via physico-chemical processes. Hence, Sr can provide constraints on the origins of CAIs [e.g., (3, 4)], the genetic relationship between solar system materials [e.g., (410)], the chronology of volatile element depletion in planetary bodies (1113), and nebular, accretion, and alteration processes that shaped the solar system (3, 4, 1416). Because Rb is rather volatile while Sr is more refractory, the 87Rb-87Sr isotopic system is, in principle, one of the most powerful tools available to date the timing of volatile element depletion. In practice, however, many uncertainties remain due to the presence of nucleosynthetic anomalies in CAIs, meteorites and planetary materials, and the extent to which 87Sr/86Sr variations stem from 87Rb decay or incomplete homogenization of nucleosynthetic anomalies.

The existence of Sr nucleosynthetic anomalies in CAIs and their host meteorites, relative to Earth, is well documented (47, 9, 10, 13, 17, 18). These positive anomalies are on the order of a few tens to hundreds of parts per million and usually expressed in the μ84Sr notation: the deviation in parts per million from the measured NBS-987 value after internal normalization to a constant 86Sr/88Sr = 0.1194. Despite the apparent 84Sr excesses seen in internally normalized data, it remains unclear whether the anomalies observed in bulk CAIs result from variations in the p-nuclide 84Sr or reflect (through internal normalization) an r-process anomaly in 88Sr [e.g., (7, 19)]. Either interpretation is possible and is an enduring challenge in Sr isotopic studies of early solar system materials.

This uncertainty surrounding the nature of the 84Sr anomalies has implications beyond unraveling the presolar heritage of the solar system. In particular, it affects dating of volatile element depletion in the early solar system. Because Rb is more volatile than Sr (50% condensation temperatures under solar nebula conditions of 800 and 1464 K, respectively) (20), one can constrain the timing of the evaporation/condensation event(s) that lead to the depletions in moderately volatile elements observed in planetary bodies (5, 1113, 21), which depends on the assumptions (i) that the initial 87Sr/86Sr ratio of solar system materials is known and homogeneous and (ii) that 87Sr/86Sr variations are entirely due to 87Rb decay. These assumptions may not be warranted if isotopic anomalies of presolar origin are present in the nascent solar system and incompletely homogenized. Indeed, to correct for instrumental mass-dependent fractionation during analysis, calculations of 87Sr/86Sr use an internal normalization scheme by setting the 86Sr/88Sr ratio to a fixed value. If the nucleosynthetic anomalies in CAIs and bulk solar system materials reside in the p-nuclide 84Sr [as proposed in (4, 6, 7, 22, 23)], then the normalization is warranted and so are the chronological interpretations based on variable initial 87Sr/86Sr ratios. If, however, the anomalies lie on 86Sr (s-deficit) or 88Sr (r-excess), the normalization will shift the initial 87Sr/86Sr ratio (5), and variations in initial 87Sr/86Sr would instead reflect the compounded effect of radiogenic ingrowth and nebular isotopic heterogeneity (9, 13, 17).

These outstanding questions can be answered, provided that one finds materials that (i) formed early in the history of the solar system and (ii) were minimally affected by radiogenic ingrowth of 87Sr. As the oldest materials formed in the solar system, CAIs are the prime target for such investigations. Yet, finding CAIs with pristine, low Rb/Sr ratios is nontrivial because Rb (and to a lesser extent Sr) are fluid-mobile, and the Rb-Sr systematics in CAIs can be significantly disturbed during aqueous alteration [e.g., (4, 17, 24)]. Here, in an effort to isolate phases with low Rb/Sr ratios and constrain the nature of the carrier of 84Sr anomalies, we subjected to a step-leaching dissolution protocol a suite of nine fine-grained CAIs (fg-CAIs) from Allende (see Table 1 for listing and figs. S1 to S9 for imagery). The fg-CAIs were chosen for their relatively primitive characteristics. In general, they display condensate-like petrographic features (25), have never been melted, and their chemical compositions match those predicted by equilibrium thermodynamic calculation for partial condensation from a gas of solar composition gas [e.g., (2629)]. Our samples (figs. S1 to S9) share these features, and, in particular, their Rare Earth Elements (REE) patterns are classic type II volatility-controlled patterns (fig. S10 and table S1). Because in situ methods would provide insufficient precision, the isolation of pristine domains (i.e., with low primary Rb/Sr ratios) was performed chemically via step leaching.

Table 1 Sample names and masses of CAI samples.
View this table:

The Sr isotopic compositions of an unleached powder aliquot (i.e., bulk analysis) and each leached fraction were measured by thermal ionization mass spectrometry (TIMS) using internally normalized and double-spike methodologies. After visual inspection of slabs of the Allende CV chondrite, CAIs with fine-grained morphologies were extracted using clean steel dental tools (see Materials and Methods below). Given the limited mass of the samples (13 to 158 mg), and to ensure sufficient precision even on the smallest sample fractions (<10% of total Sr in the CAI), 1013-ohm feedback resistors (30) were used as these permitted high-precision Sr measurements on load sizes as low as 150 pg. Each leachate fraction was split into three aliquots to quantify (i) nucleosynthetic anomalies through internal normalization on an unspiked sample aliquot, (ii) stable 86Sr and 88Sr relationships using double-spike methodologies, and (iii) 87Rb abundances by high-sensitivity inductively coupled plasma mass spectrometry (ICP-MS) measurements to assess the contribution of 87Rb-decay on 87Sr/86Sr variations and estimate initial 87Sr/86Sr values using Rb-Sr isochrons. At each stage of the step-leaching process, the leached residues were weighed to estimate (i) the amounts of components dissolved and (ii) their chemical and Sr isotopic composition.


Step-leaching methodology and Sr release patterns

Each CAI sample was subjected to a five-step leaching protocol (see Materials and Methods) of increasingly aggressive acid type and molarity, and temperature. The leaching experiments and separation of the aliquots for internally normalized measurements were performed before any Sr double-spike materials were used in the laboratory (see the Supplementary Materials for details). Of the nine CAIs analyzed (Table 1), only one (Logan) had no material remaining after the fourth leaching step. In all cases, the first three leach steps contained >80% of the total Sr budget, and the amount of Sr released in each step typically decreased by two to three orders of magnitude across the total experiment (Fig. 1 and Supplementary Materials). To test the efficiency of the step-leaching protocol and accuracy of the gravimetric accounting for Rb and Sr yields, a separate fraction of powder from each CAI was dissolved in bulk, using a Parr bomb to ensure total dissolution, and analyzed similarly to the leach aliquots. Excellent agreement was observed between the Sr isotopic characteristics (Fig. 2) for the bulk samples versus the gravimetrically summed leach fractions.

Fig. 1 μ84Sr values and associated Sr yields from step leaching.

Top: Internally normalized μ84Sr for each leach fraction (table S4). The gray shaded bar represents the measured isotopic composition of the bulk CAI (±2 SD). Bottom panels show the amount of Sr (ng) released in each leaching step. Note the use of log-scales throughout. See Table 1 for the CAI catalog numbers.

Fig. 2 Summary of μ84Sr data presented in this paper.

Top: Recombined estimates for μ84Sr for steps L1 to L3 (red circles) and L1 to L5 (blue circles) are compared to the bulk data for each CAI. The gray-shaded bar represents the external reproducibility of 22 ppm (2 SD) obtained in this study for the internally normalized μ84Sr for 1000-ng loads of the NBS-987 standard (table S1). Weighted averages were calculated using Isoplot (61). Bottom: Relative probability distribution function (PDF) (using Isoplot) (61) and histogram for all published Allende bulk CAI μ84Sr data for comparison [incorporating data from (4, 6, 9, 10, 13, 17, 18)].

Standard data

Two instruments were used in this work: the 1011-ohm equipped Triton TIMS at Victoria University of Wellington (VUW), which was used for large Sr loads (100 to 1000 ng), and the 1013-ohm equipped Triton XT at the Thermo Fisher factory in Bremen, which was used for Sr loads below 100 ng. Data accuracy was thoroughly evaluated on both instruments (see Materials and Methods). External reproducibility for internally normalized data on the VUW Triton was (2 SD, n = 9; 900 ng Sr loads): 84Sr/86Sr = 0.0564888 ± 12 [22 parts per million (ppm)] and 87Sr/86Sr = 0.7102604 ± 42 (5.9 ppm). External reproducibility on the Bremen instrument was evaluated for Sr loads of 1 ng, 500 pg, and 50 pg (fig. S11). The average (n = 10) 84Sr/86Sr ratio and 2 SD values were found to be, respectively, 0.0564997 ± 171 (302 ppm), 0.0564977 ± 180 (326 ppm), and 0.0564966 ± 1280 (2265 ppm). The average 87Sr/86Sr and 2 SD values were, respectively, 0.710260 ± 45 (64 ppm), 0.710248 ± 43 (60 ppm), and 0.710234 ± 150 (211 ppm) (table S2).

Similar tests performed on double-spike measurements [used to constrain mass fractionation effects, using methods from (8)] yielded external reproducibilities (2 SD) on the VUW instrument (900-ng loads, n = 8) of ±0.038 per mil (‰) and ± 0.030‰ for δ84Sr/86Sr and δ88Sr/86Sr, respectively. Using the same double-spike methodologies for 1-ng loads on the Bremen instrument yielded external reproducibilities of ±0.29 and ±0.19‰ for δ84Sr/86Sr and δ88Sr/86Sr, respectively (n = 10).

Strontium isotopic release patterns

Nucleosynthetic anomalies. The internally normalized 84Sr/86Sr data in each leach step (expressed as μ84Sr; see the “Notations” section in Materials and Methods) are shown in Fig. 1. Within each CAI, wide variations are observed. In all cases, μ84Sr values are positive, and the first three leaching steps (80 to 99% total Sr) have elevated μ84Sr values that are within the range of published values observed in bulk single CAIs (Fig. 2). In contrast, the final leach (L4) and residue digest (L5) show μ84Sr anomalies reaching per mil to percent levels, and the fractions most depleted in Sr also display the largest anomalies (Fig. 1). The largest anomaly was found in Hank, where the final digest (L5) revealed an extreme μ84Sr value of 80,655 ± 1807.

Rb/Sr systematics and initial 87Sr/86Sr ratios. The values for both 87Rb/86Sr and measured 87Sr/86Sr (Fig. 3) cover extreme ranges (e.g., 87Sr/86Sr measured for Logan in L1 of 0.936) and show three important features. (i) The 87Rb/86Sr relationships remain consistent across the first four leach steps, with ages and initial 87Sr/86Sr values overlapping within uncertainty (Fig. 3), despite the Rb abundances and Rb/Sr ratios in the first three leach fractions diminishing between each step. The mean square of weighted deviates (MSWD; equivalent to the reduced χ2) values for the best-fit lines for steps L1 to L4 are such, however, that some scatter beyond analytical error is present. (ii) Minor amounts of Rb persist into leaching step L4, with a similarly consistent pattern of correlation with 87Sr/86Sr measured ratios, and yield a precise initial ratio for L4 of 0.698784 ± 76 (2 SD). (iii) In marked contrast to the earlier leach steps, the 87Rb/86Sr versus 87Sr/86Sr relationships of the L5 residual refractory phases are incoherent. In all but one case (L5_Raven), the data scatter to more radiogenic 87Sr/86Sr values than the L4 isochron and show no meaningful age relationships.

Fig. 3 (A-D) 87Sr/86Sr versus 87Rb/86Sr values for each leach fraction.

Panels are scaled in each case to show a suitable range in values on the x and y axes. In (D), analyses from leach steps 4 and 5 are filled blue and gray symbols, respectively. Shaded box in each panel represents the area shown in each successive panel for comparison. Error bars are smaller than symbol size unless shown. Best-fit lines and 2 SD error envelopes are calculated using the model 3 fit of Isoplot (which assumes that the scatter is due to a combination of assigned errors plus a normally distributed variation in the y value) (61) and an 87Rb decay constant of 1.3971 ⨯ 10−11 year−1 (62). A ~13.8-Ga reference line has been added to illustrate that several of the L5 aliquots have extreme and unsupported 87Sr/86Sr ratios, emphasizing their exotic nature with respect to solar system materials. For comparison, isochron diagrams constructed based on all the leachates from each of the individual CAIs are shown in fig. S11.

A complementary view of the data can be obtained by plotting individual isochrons across all leach steps for each CAI separately (fig. S12). Plotted in this fashion, the leach steps define best-fit lines that are heavily leveraged by L1, but most yield ages within error of the solar system age. The scatter in the model ages and high MSWD values in Fig. 3 and fig. S12 imply that other effects are present. There may be a second-order influence from minor incongruent mineral dissolution induced by the step leaching process. Later-stage asteroidal alteration also cannot be ruled out with the available data. Other components within the same fragment of Allende would need to be analyzed in the same fashion to search for consistency that would indicate alteration in the meteorite parent body. Initial 87Sr/86Sr ratios range between 0.6967 ± 73 and 0.6994 ± 12 (fig. S12). These initial 87Sr/86Sr ratios are consistent within error and define a weighted mean of 0.69874 ± 11 (MSWD = 0.23), which agrees within error with the value obtained above by grouping the L4 data points across all nine CAIs.

Stable isotope fractionation patterns. The stable Sr isotopic compositions derived from the double-spike methodologies are plotted on δ88/86Sr versus δ84/86Sr three isotope plots in Fig. 4. For all but one sample (Hank), the leach fractions L1 to L3 define tight clusters in the three-isotope plots (Fig. 4B), with leach fractions L4 and L5 showing distinct compositions. The location of the clusters defined by the L1 to L3 fractions scatter between a CAI mass-dependent fractionation line (MDFL) defined from bulk CAI data (4), which is offset by 0.144‰ in the δ84/86Sr coordinate, and the equilibrium MDFL (Fig. 4A). The CAI MDFL is also identical, within error, to that previously defined by primitive meteorites (8). The CAI data presented here cover a total range of δ88/86Sr values from −1.94 to +1.14‰ and lie within the range of the values previously obtained for bulk CAIs (~5.3‰ in δ88/86Sr, from −3.67 to +1.67‰) (4). In stark contrast, the range of δ84/86Sr values from steps L4 and L5 shows clear positive departure from both the terrestrial and bulk CAI MDFLs, in some cases, to extreme values never before observed (up to δ84/86Sr = 81.41 ± 2.49‰).

Fig. 4 δ88/86Sr versus δ84/86Sr (obtained using the double-spike technique).

(A): Data for all CAIs organized by leach fraction. “Equilibrium MDFL” denotes the equilibrium mass-dependent fractionation line (solid line) that passes through the origin (the measured composition of NBS-987) with a slope of −0.978 (14). A best-fit line (dashed line) to the bulk CAI data from Charlier et al. (4) is plotted for comparison with the step leach data. The bulk CAI best-fit line is offset from the equilibrium MDFL by +0.144‰ in δ84/86Sr. 2 SD uncertainties are smaller than the symbol size unless shown. (B): Data for all leach fractions organized by CAI. For all CAIs except one (Hank), the data from leaching steps 1 to 3 are tightly clustered (shaded area), consistent with partial homogenization of the Sr isotopic composition of the CAIs during closed-system aqueous alteration (see the main text).


Bulk CAI versus individual leachates Sr data

The nucleosynthetic, radiogenic, and stable “bulk” CAI data obtained in this work (reconstructed from the individual masses and isotopic characteristics of materials at each leach step) are all in good agreement with literature data obtained on dissolutions of bulk CAI samples (4, 6, 9, 10, 13, 17, 18), indicating that the samples investigated here are representative of the type of CAIs previously studied at the bulk level (fine-grained group II). The step leaching protocol used here, however, provides insights into the history and isotopic heritage of CAIs that shed new light on bulk isotopic data.

There is a general consistency of Sr isotopic signatures (μ84Sr, δ84/86Sr, δ88/86Sr, and initial 87Sr/86Sr ratios) across the first three leaching steps (L1 to L3; e.g., Figs. 2 and 4), which together account for most of the Sr inventory of the bulk samples (80 to 99%). In three-isotope plots, the double-spike data (δ84/86Sr and δ88/86Sr) for fractions L1 to L3 cluster tightly (Fig. 4B), suggesting that either (i) a single phase is being digested in all three steps, or (ii) the different phases digested in these steps experienced the same extent of Sr mass-dependent fractionation. Major and trace element concentrations obtained on each leachate show that multiple phases are being digested in each step (fig. S13), starting with secondary phases like sodalite and nepheline in L1 and L2, followed by the digestion of hedenbergite, grossular, and some primary phases like melilite and diopside in L3, L4, and lastly the most refractory phases like spinel in L4 and L5, and other trace phases in L5. Yet, because fg-CAIs have condensate textures, one would not necessarily expect the different phases digested in steps L1 to L3 to be characterized by a similar extent of mass fractionation, at least as a primary feature. Because Allende is known to have been subjected to intense aqueous alteration, which is the most likely source of the high Rb content seen in L1 and L2, the similar δ84/86Sr and δ88/86Sr values in L1 to L3 could be the result of closed-system Sr redistribution and homogenization during aqueous alteration. This is consistent with the predominantly secondary nature of the phases digested in leachates L1 to L3 and the generally homogeneous μ84Sr values in these leachates, which are similar to the values in both fg- and coarse-grained bulk CAIs (μ84Sr = ~+125) (4). Furthermore, Allende as a whole has a μ84Sr value of +58 ± 17 (7), and the matrix alone should have an even lower μ84Sr value (to satisfy mass balance requirements). If extraneous Sr from the meteorite matrix had been introduced in the CAIs during parent body alteration, this would have lowered the μ84Sr value of the most fluid mobile component of the CAIs relative to the refractory components that are not sampled by leachates L1 to L3. We therefore conclude that, although aqueous alteration partially homogenized the Sr isotopic composition of the CAIs, it did not mobilize Sr beyond the boundaries of the CAIs and thus each CAI can be treated as an individual closed system object.

The final leachate (L4; 0.25 to 10.6% of total Sr) and residue digest (L5; 0.06 to 9.7% of total Sr), which together concentrate the chemically resistant phases, show Sr isotope systematics in stark contrast of those observed in the initial leachates. In particular, these leachates display positive μ84Sr anomalies of unprecedented magnitude, up to +80,655 ± 1807 in the Hank CAI (Fig. 1). In Fig. 2, we compare the bulk CAI μ84Sr values with those reconstructed by mass balance by taking into consideration the μ84Sr values and Sr amounts in each leachate fraction. We find agreement between the bulk value and the reconstituted value from L1 to L5. While the Sr amounts in L4 and L5 are small, they contribute significantly to bulk μ84Sr anomalies, and not considering them would lead to a discrepancy between measured and reconstructed bulk μ84Sr anomalies. This comparison between measured and reconstructed bulk μ84Sr values shows that the anomalies measured in L4 and L5 are real and not analytical artifacts due to the low amounts of Sr available for isotopic analysis in those leachates. These data demonstrate the existence and survival in CAIs of isotopically exotic materials, which have here been concentrated through our chemical step-leaching protocol.

A p-process origin for 84Sr excesses in CAIs

Whether the nucleosynthetic anomalies seen in CAIs and planetary bodies are due to variations in the p-nuclide (84Sr) or r-nuclide (88Sr) is a matter of contention [e.g., (410, 13, 17)]. If the positive μ84Sr values measured here were the result of r-process excesses in 88Sr propagating to 84Sr as an internal normalization (to 86Sr/88Sr = 0.1194) artifact, this would also result in apparent negative 87Sr shifts of a magnitude half those of 84Sr (i.e., the samples should appear less radiogenic) (5). Despite the large μ84Sr values in fractions L4 and L5, no negative 87Sr anomalies are seen in those fractions. The data from leach step L4 show coherent behavior (Fig. 3D), yielding an initial 87Sr/86Sr value of 0.698784 ± 0.000076 (2 SD, MSWD = 9.2) that lies within the error of the original ALL value of 0.69877 ± 0.00004 reported by Gray et al. (24) for Allende CAI USNM 3898 [see also (31)]. Similarly, a range of low apparent 87Sr/86Sr initial ratios can be calculated for the individual age-corrected analyses in L4 (87Sr/86Sr = 0.698587, L4_Ororo to 0.698883, L4_Jean), which, in turn, agree with those presented by Amelin et al. (32) for the leachates of the FUN (fractionation and unknown nuclear effects) CAI, CMS-1 (87Sr/86Sr = 0.69839 to 0.69900). Furthermore, despite the large μ84Sr values in leach step L5, no accompanying negative 87Sr anomalies are seen in those fractions, as all but one sample (Raven) plot significantly above the best-fit reference line for step L4 87Sr/86Sr values (Fig. 3). These results are inconsistent with the anomalies stemming from any substantive r-/s-process 88Sr variability and instead imply that the anomalies lie primarily in the p-nuclide 84Sr.

The double-spike data further support this conclusion. In the three-isotope plots (Fig. 4), the data for fractions L4 and L5 form a horizontal envelope (displacement toward higher 84Sr/86Sr ratios), which strongly argues against any notion that the μ84Sr variations are an artifact of normalization to an inaccurate value for the 88Sr/86Sr ratio. Any hypothetical variability in the 88Sr/86Sr ratio resulting from the proportions in r- versus s-process 88Sr may be discerned in individual L4-L5 data pairs but is wholly subordinate to the δ84/86Sr variations. This is demonstrated in Fig. 4B by the nearly horizontal slopes of hypothetical tie lines for those pairs showing extreme δ84/86Sr values, and the fact that no L5 analyses fall outside the range of δ88/86Sr values for the dataset (L1 to L4 and, by definition, reconstituted bulk samples) as a whole. In addition, the vertical trajectory of the L4 and L5 data in a plot of μ84Sr versus 87Sr/86Sr (Fig. 5) precludes the presence of any discernible 88Sr excess, which would drive the data to less radiogenic values for 87Sr/86Sr.

Fig. 5 Nucleosynthetic anomalies (μ84Sr) versus 87Sr/86Sr, both from internally normalized data.

Expected relationships for pure radiogenic ingrowth, 84Sr excess and correlation arising from internal normalization to 88Sr/86Sr in the presence of r-process anomalies (arrows) are shown. Component mixing follows curved trajectories in such a frame of reference (they would follow straight lines if the y axis were on a linear scale). Shaded area denotes the region plotted on the inset figure.

The unsupported 87Sr in the L5 fraction (i.e., 87Sr excesses that cannot be accounted for by in situ decay of 87Rb) further emphasizes the exotic nature of the carrier concentrated by the step leaching protocol. Although a broad correlation is observed between the μ84Sr versus 87Sr/86Sr (corrected or uncorrected for radiogenic ingrowth based on the Rb/Sr of the sample fraction; Fig. 5), the origin of the 87Sr excesses remains unclear. These could reflect (i) incongruent dissolution and fractionation of Rb/Sr, (ii) nucleosynthetic anomalies in 87Sr, (iii) radiogenic decay of 87Rb in the interstellar medium and incorporation into the presolar carriers, or (iv) contamination on the parent body with radiogenic Sr. None of these explanations can be ruled out at this point, because both the ages and initial ratios of these materials are unknown.

Nature, abundance, origin, and preservation of presolar carriers in CAIs

A long-held view is that presolar grains are only present in the fine-grained matrices of primitive meteorites that have experienced minimal high-temperature alteration or melting (33, 34), and CAIs, which formed at high temperature, should be thus devoid of presolar grains (35). In contrast with this general view, but in agreement with a recent stepwise pyrolysis study of noble gas isotopic signatures in fg-CAIs from Allende (36), our data provide evidence for the survival of distinct presolar carriers in CAIs. On the basis of the specific krypton, neon, and xenon s-process excesses observed, Pravdivtseva et al. (36) proposed that the carriers isolated in the 1200° to 1300°C pyrolysis step were presolar SiC (<0.2 μm in size), which were incorporated in fg-CAIs during their formation and survived any secondary processing. Although our data do not permit identification of the specific carrier(s) of the 84Sr anomalies, SiC grains are extremely unlikely candidates. 84Sr excesses represent p-nuclide anomalies and the few SiC grains that have been measured for Sr isotopes (n = 39, including both grains from Asymptotic Giant Branch (AGB) stars and supernovae) display either solar composition or negative anomalies (down to −90.6 ± 2.0%) (3739). No presolar grains, to date, have been found with excess 84Sr signatures.

The p-nuclides are thought to be formed primarily by photodisintegration of progenitor isotopes when the temperature in stars reaches ~3×109 K (40). Core collapse supernovae (41) and type Ia supernovae [e.g., (42)] are two likely sites of such process, yet in both cases, it is unclear whether grains would condense from those materials and, at least for type Ia supernovae, no grain condensation has been observed from spectroscopic observations. The only documented p-nuclide anomalies from presolar grains are in nano-diamonds, which upon heating release Xe with a Xe-HL (H and L stand for high and low mass isotopes, respectively) signature characterized by enrichment in both p- and r-process nuclides (43). The excesses in Sr p-nuclide without corresponding Sr r-process enrichments observed in our leach fractions L4 and L5 therefore point to the existence of presolar phases with only p-nuclide anomalies. These data, taken at face value, would indicate that at least two types of presolar grains (the p-nuclide–rich carrier and the SiC inferred from noble gas analyses) were incorporated and preserved in fg-CAIs.

The presence of presolar grains in CAIs, as evidenced by our Sr data and the noble gas data of Pravdivsteva et al. (36) is unexpected. Presolar grains would not be expected to survive either the high temperatures of the CAI forming region or the aqueous alteration and mild heating of the Allende parent body that destroyed some presolar grains (SiC) but not all (nanodiamonds) (33, 34). How and when presolar grains were incorporated into Allende fg-CAIs, and how they survived, is therefore an important question. One possibility is that these presolar grains are not a primary feature of the CAIs but were introduced in these inclusions during aqueous alteration on the parent body. The large positive 84Sr anomalies (maximum μ84Sr = +80655 ± 1807) observed in leach fractions L4 and L5 in our fg-CAIs argues, however, against such a scenario. These 84Sr enrichments are unique to fg-CAIs and in stark contrast with the large negative anomalies (down to μ84Sr = −4741 ± 70 in Murchison) in refractory residues of leaching experiments of most bulk carbonaceous chondrites (Ivuna, Murchison, and Tagish Lake) (6, 7, 10). These negative values are consistent with the high abundance of SiC grains in these meteorites (44), and their refractory nature would lead to concentration in the final steps of the leaching protocol. Furthermore, step-leaching of bulk Allende samples yielded essentially homogeneous, positive 84Sr signatures in all leaching steps, identical to the bulk Allende value (μ84Sr = ~+58) (7), consistent with the destruction of presolar grains during aqueous alteration, the homogenization of the Sr anomalies at the bulk Allende scale, and the low residual abundances of presolar grains in Allende (0.0062 ppm of SiC) compared to less altered carbonaceous chondrites (e.g., 13.5 ppm of SiC in Murchison) (44). The 84Sr-rich nature of the presolar carrier in fg-CAIs is thus inconsistent with introduction of these grains during fluid alteration on the Allende parent body.

The presolar component(s) in fg-CAIs, represented by the residues in leach steps L4 and L5, therefore appear to be distinct and primary phases, having been incorporated in fg-CAIs during their genesis rather than introduced by secondary processes on the Allende parent body. In theory, the geochemical signatures of the successive leachates could shed light on the nature of these carrier phases. Cross-plots of major and trace element atomic ratios in the leachates do not, however, allow us to pin down the nature of the carriers. The contribution of the carriers to the chemical composition of the leachate is overwhelmed by those of the major phases (fig. S13), and the L5 leachates define mixing arrays between a spinel pole and an Al-rich diopside pole. Without further knowledge of the nature of these carriers, their survival through the alteration events experienced by the Allende CAIs could be explained if they occurred as inclusions within refractory minerals (i.e., for which they could have acted as nucleation seeds), were isolated from the fluid flow path by aggregates of the refractory minerals in fg-CAIs, or were in themselves resistant to alteration by the metasomatic fluids. The first scenario would imply survival of the presolar carrier in high-T environments where the CAI minerals condensed. The second requires condensation of the minerals forming fg-CAIs in high-T regions, then transport to cooler regions where the presolar carriers would be present (or cooling of the CAI forming region and introduction of the presolar carriers) before, lastly, accretion of the CAIs from condensed grains. The third scenario would rule out presolar carriers that are prone to destruction by parent body metamorphism (e.g., silicate grains) as the carrier phase of the 84Sr excesses observed here.

Implications of p-process Sr anomalies in CAIs for the isotopic heritage of the solar system

The significant 84Sr excesses from fractions L4 and L5 suggest that most of the apparent 84Sr excesses in the bulk CAIs are due to one or more of the various p-processes (40). L4 and L5, which only contain 0.5% (Ororo) to 20.3% (Erik) of the total Sr inventory of the CAIs, contribute between 1.6% (Logan) and up to 69.7% (Jean) of the 84Sr excesses of the bulk CAIs (Fig. 2). In particular, in the Jean CAI, which displays one of the larger reported anomalies in CAIs (μ84Sr = +230), L4 and L5 only account for 1.3% of the Sr inventory, but their influence on the bulk μ84Sr value is dominant (μ84Sr = +191, 69.7% of the bulk anomaly).

Because the composition of many refractory elements in CAIs appears to be mainly controlled by variations in the relative contribution of the r- and s-process (Zr, Mo, Hf, W, and Nd), the same interpretation has often been assigned to Sr, Ba, and Sm anomalies [e.g., (18)]. However, the p-nuclides could represent up to 70 and 75% of the nucleosynthetic heritage of, respectively, Ba and Sm in bulk CAIs and all of the anomalies in Sr (45). A main control of p-processes on bulk CAI Sr anomalies, as shown by our data, means that the contribution of p-processes in setting the bulk CAI composition of Sm and Ba is possible and demands further investigation.

A p-process origin of Sr anomalies in CAIs also has ramifications for our understanding of the origin of the 84Sr anomalies in bulk meteorites. Initially identified based on clustering of planetary materials in ε50Ti, ε54Cr, and Δ17O isotope spaces (46), a clear isotopic dichotomy also exists in Sr isotopic characteristics between carbonaceous (CC) and non-carbonaceous (NC) objects (8). This dichotomy has also been documented in numerous other isotope systems and, in combination with elemental abundance data for both refractory and nonrefractory elements (Ti, Sr, Ca, Cr, Ni, Zr, Mo, Ru, Ba, Nd, Sm, Hf, W, and Os), reveals a general pattern whereby the CC reservoir is offset from the NC reservoir toward the composition of CAIs (10). This observation holds even for CC bodies where visible CAIs are extremely rare or absent (e.g., CI chondrites), implying the existence in CC materials of an as-yet unidentified (cryptic) refractory component, isotopically similar to CAIs (4, 10, 47). Recently, Burkhardt et al. (10) hypothesized that CAIs and this cryptic refractory dust originated both from an inclusion-like chondritic (IC) reservoir, the isotopic composition of which is similar to that of CAIs but the chemical composition of which is similar to bulk chondrites. Mixing of IC reservoir material with material from the NC reservoir would successfully explain the formation of the CC reservoir, and the two would become isolated from one another after the formation of Jupiter (10, 48).

In the framework presented above, the nucleosynthetic processes controlling the bulk composition of CAIs would also control the isotopic difference between CC and NC materials, up to and including planetary-sized bodies. The limited available Sr data obtained from step-leaching experiments of carbonaceous chondrites (6, 7, 10) support this hypothesis. In the studied samples (Ivuna CI chondrite, Murchison CM chondrite, and Tagish Lake C2-ungrouped chondrite), phases digested in dilute acetic acid and low molarity (<5 M) nitric acid at room temperature contain >90% of the total Sr of the meteorite and have positive μ84Sr values (+45 to +69), consistent with a p-nuclide origin of these anomalies. In contrast, the refractory components digested in later leaching steps extend to very negative μ84Sr values (Ivuna: −3240 ± 120; Murchison: −4741 ± 70; Tagish Lake: −3427 ± 17). The influence of these refractory components is consistent with the presence of SiC grains and variable s-process contributions to these bulk CC meteorites, as observed for other elements (e.g., Mo) (49). Yet, these unambiguous s-process signatures only result in moderate decreases in the bulk isotopic anomalies of the meteorites measured (μ84Sr = −17 in Ivuna, −17 to 27 in Murchison, and −33 in Tagish Lake), which supports the idea that the positive 84Sr anomalies in bulk CC materials are mainly controlled by materials with a p-nuclide excess (through addition of CAI and CAI-like components).

There is a clear isotopic difference between the refractory components isolated from fg-CAIs (whose composition is thought to represent the average early infalling molecular cloud material: the IC reservoir) and bulk CC meteorites (thought to represent a mixture of IC and NC materials). This observation supports proposals of a heterogeneous distribution of presolar grains in the presolar molecular cloud, as a result of changes in the composition of the material infalling on the protosolar nebula (45, 48, 50). The extreme 84Sr excesses reported here in fg-CAIs also deepen the gap between the “normal” and FUN CAIs, which display large negative 84Sr anomalies. The bulk FUN CAIs EK-1-4-1 and C1 (5) are characterized by negative μ84Sr values of about −3200 and −700, respectively, and the leached FUN CAI CMS-1 (32) extends to about −420 and −560. These results suggest that the main carriers of 84Sr anomalies in fg-CAIs are fundamentally different from those in FUN Allende CAIs.

Implications for Rb-Sr chronologies of the early solar system

Our data have important implications for the chronology of volatile element depletion in planetary materials (12, 17). Taken at face value, the differences between the 87Sr/86Sr initial ratios of CAIs and differentiated NC meteorites, such as eucrites and angrites, yield model ages for these meteoritic parent bodies that are equivalent to (17), or longer than (24), their core formation and accretion time scales as derived from Hf-W and Mn-Cr systematics (e.g., (51)). This feature was considered by Halliday and Porcelli (12) to represent formation of the eucrite/angrite parent bodies with a solar or higher than solar Rb/Sr ratio, followed by a period of rapid ingrowth of 87Sr, before large-scale volatile depletion to explain the very alkali-poor nature of these meteorites (present Rb/Sr ~1 to 2 orders of magnitude lower than the solar value).

As an alternative, Moynier et al. (13) and Hans et al. (17) argued that the indistinguishable initial 87Sr/86Sr for eucrites and angrites [that is, BABI (basaltic achondrite best initial)] versus the lower value for CAIs reflected an anomaly in the normalizing 86Sr/88Sr ratio and that when “corrected for” removed the 84Sr anomaly and made the Gray et al. (24) USNM 3898 value for initial 87Sr/86Sr match that of eucrites and angrites. Settling this debate therefore boils down to identifying which nucleosynthetic process is responsible for the apparent 84Sr anomalies of CAIs relative to the NC planetary materials (angrites and eucrites). The magnitudes and direction of the 84Sr offsets we measured in leach fractions L4 and L5 show unequivocally that the anomalies must lie in 84Sr. This conclusion renders the proposal of Hans et al. (17) untenable and agrees with previous work (22, 23, 32, 52) that pointed out that the Allende FUN inclusions similarly show diverse 84Sr variations yet coherent initial 87Sr/86Sr initial values, closely comparable to the canonical USNM 3898 CAI (24). In our results, the initial 87Sr/86Sr value of 0.698784 ± 0.000076 yielded by the isochron from the step L4 aliquots (i.e., the closest our step-leaching protocol came to isolating the primary condensate mineralogy in the CAIs) is indistinguishable, within error, from those derived for ALL [0.69877 ± 0.00004 (24) and 0.698808 ± 0.000020 (21)]. Thus, if there are any variations in the r- versus s-process isotopic balance of 88Sr (within the limits of error of the measurements we make; Fig. 5), these are concealed within the variability of the bulk data. Our results thus reinstate the distinction in initial 87Sr/86Sr values between the inner solar system–differentiated materials (NC reservoir) represented by eucrites and angrites (11, 53) on the one hand, and CAIs and CC meteorites on the other hand, and suggest that this difference holds a chronological significance. Translating this distinction into a time scale is, however, hindered by our lack of knowledge of the initial Rb/Sr ratios of the angrite and eucrite parent bodies. Although beyond the scope of this paper, careful re-evaluation of this issue will be needed to place robust constraints on the time scales of volatile depletion in the early solar system based on Rb-Sr systematics.



Nine fg-CAIs from Allende were investigated in this study. Table 1 gives the sample numbers as used here, plus the catalog numbers in the American Museum of Natural History database, and sample masses taken for both the step-leaching experiments and bulk analysis. The samples were initially identified on the basis of their texture (fine-grained), color (from white/gray to purple, with most being pink due to FeO in spinel), and irregular shapes. They were excised from cut slabs of Allende and crushed in a class 1000 laboratory at the Massachusetts Institute of Technology (McGee lab), using the methods described in Tissot et al. (54). A small chip of each CAI was extracted using cleaned stainless steel dental tools (Hu-Friedy: see below for the metal composition), under a stereoscopic zoom microscope, and mounted in epoxy for microbeam investigation to confirm the classification of the samples as fg-CAIs, based on their typical textures, compositions, and mineralogy [e.g., (25, 54)] For each CAI, secondary electron and back-scattered electron images, along with mineral identifications, are shown in figs. S1 to S9.

In most samples, primary mineral phases were found to be spinel (predominantly), perovskite (minor occurrences), and sometimes melilite. Two samples (Erik and Logan) showed, respectively, hibonite and pyroxene, as another major phase. These features are consistent with the typical mineralogy of fg-CAIs (55, 56). As is common for fg-CAIs from Allende, all samples have experienced significant alteration as evidenced by the abundance of secondary phases such as nepheline, sodalite, grossular, anorthite, and hedenbergite, as well as the presence of clinopyroxene rims around spinel (57). Forsterite was found in most samples, mostly concentrated in the rim of the CAIs, similar to the outermost zone C in inclusions described by Hashimoto and Grossman (57), and as individual occurrences inside some of the samples. Trace occurrences of Fe-Ni metal as well as iron and/or titanium oxides (i.e., ilmenite and magnetite) were also observed.

Sample preparation and step leaching

The powdered CAIs were weighed out into preweighed, clean conical-bottomed perfluoroalkoxy alkane (PFA) Savillex beakers, and step leaching was carried out using acids of different concentration while varying the temperature and duration of treatment. The entire leaching process and subsequent aliquoting, column chemistry, and dilution procedures were performed at VUW in an ultraclean environment using high-purity distilled acids and water produced by two-time subboiling distillation in Teflon stills (TD reagents; see information on procedural blanks below). Step leaching (L) dissolution used the following sequence:

1) L1: 0.5 M CH3CO2H (acetic acid), 1 day, 20°C, shaken up by hand every few hours;

2) L2: 0.5 M HNO3, 3 days, 20°C, shaken up by hand two to three times a day;

3) L3: 6 M HCl, 1 day, 120°C, sonicated for 15 min at the beginning;

4) L4: 14 M HF, 9 M HNO3, 1 day, 120°C, sonicated for 15 min at the beginning;

5) L5 (final digest): 14 M HF, 9 M HNO3, 3 days, 220°C, high pressure Parr bomb digest.

This protocol broadly follows previous studies [e.g., (58)]: L1 should only mobilize the soluble alteration phases; L2 would digest more of the alteration phases; L3 would presumably start digesting oxide, silicate, and metal; L4 would digest most minerals except for the most refractory phases; and L5 is a full digestion of the residue in a pressurized vessel. The aliquot fractions were labeled by leach step (L1 to L5) and sample name (_name). After each leaching step, the beakers were secured within custom-made holders and centrifuged at 5000 rpm (5 min). The supernatant liquid was pipetted off to a separate clean beaker, taking care not to disturb the sediment pellet in the conical tip of the beaker. TD water (1 to 2 ml) was then added to the beaker, and the sediment pellet was resuspended via shaking and several minutes of sonication. This process was repeated three times to ensure full leachate solution recovery. The beaker containing the pellet was then returned to a hotplate and taken to dryness at 120°C. After re-equilibration to ambient temperature, the beakers containing the pellets were weighed again, and the difference between initial weight at the start of the leach step and the postleaching weight was taken to represent the amount of sample digested at each leach step (allowing calculation of trace elements concentration in the mass fraction of the sample digested in each step).

Leachate solutions produced in steps L1 to L4 were taken to dryness, treated with aqua regia (3:1 HCl/HNO3), dried once more, and then redissolved in 2 ml of 9 M HNO3. The Logan CAI was completely dissolved by the end of the L4 stage, and no residues were visible under a binocular microscope. For all other samples, the remaining grains in the final residue (L5) were removed from the conical vial, weighed using an ultra-microbalance, transferred into Parrish-type capsules, and digested in HNO3-HF (1:5) over 3 days using Parr bomb high pressure digestion vessels in an oven at 220°C. After the final digestion, L5 samples were rinsed out of their capsules into conventional Savillex vials, and complete dissolution following the addition of 6 M HCl was verified using a binocular microscope. The samples were taken to dryness and lastly taken up in 2 ml of 9 M HNO3.

For all sample solutions (L1 to L5) in 9 M HNO3, a 5 to 10% aliquot was taken and diluted appropriately for trace element analysis using a Thermo Fisher Element 2 sector field ICP-MS instrument, although for this paper only, the Rb, Sr, and REE contents are used (the full dataset will be presented elsewhere). Calibration curves (i.e., concentration versus count rates) for each element were generated using a range of USGS rock standards (BHVO-2, BCR-2, and AGV-2) as well as synthetic standards prepared gravimetrically from commercially available, certified, single-element standards. The range in standards was chosen to bracket the calculated ranges in concentrations of the diluted leach and residue fractions. Count times were extended for Rb and Sr to reduce the uncertainty on the parent-daughter ratio determinations, and instrumental drift was monitored and corrected for using repeated analysis of a standard measured for every two unknowns. Blank subtraction for the trace element data was carried out using the measured concentrations in a total procedure blank that had been produced under the same conditions as for sample processing (see below).

For the Sr isotope work, the remaining sample was further split at the 9 M HNO3 stage into two fractions. One half was spiked with the VUW 84Sr-87Sr double spike (DS aliquot) (4) for stable isotope determination and the other half left unspiked (natural aliquot) for conventional, internally normalized analysis. Purification of Sr from the sample matrix was performed by extraction chromatography using Sr-spec resin by sorption onto the resin and matrix elution using 9 M HNO3 and then back extraction of the Sr in 0.05 M HNO3. Natural and DS aliquots were processed separately using labware specifically designated for spiked and natural samples. The techniques for this part of the study are identical to those detailed in the supplementary materials for Charlier et al. (8).

Aliquots of the bulk powder from each CAI were also investigated. These were first weighed using an ultra-microbalance and then transferred into Savillex vials for initial HF-HNO3 digestion under low pressure at 130°C. Any remaining residues were then transferred to Parrish-type capsules and digested in HNO3-HF (1:5) over 3 days at 220°C in a high-pressure Parr bomb. After complete digestion, total bulk CAI samples were processed using the same column protocols as those for the leach fractions.

Because of the susceptibility of the exceedingly small load sizes encountered in L5, great care was taken to eliminate all possibility of contamination with materials that could influence the 84Sr abundances. Spike contamination can be eliminated as a possible influence on the data for the following reasons.

1) The step leaching process was carried out on the samples before the isotopic spike material was used in the laboratory. The samples, in solution in their factory new beakers, had a fraction aliquoted for ICP-MS trace element analysis, and a second aliquot was taken and processed through column chemistry for natural ratio analysis (which is what yielded the values used by internal normalization for the μ84Sr anomalies obtained for this study). Only then did the remaining material have the spike added to it. The first two steps, in all cases, were undertaken before the spike was used in the clean handling facilities (which had not previously seen the use of any enriched Sr isotopic spike materials).

2) All materials used for spiked analyses were kept entirely separate: independent collections of Teflon beakers for natural and double-spiked samples, separate filaments, and working spaces with only one operator active at any stage for this work. Cleaning protocols associated with laboratory items and consumables for spike versus natural materials involved entirely separate vessels and reagents.

3) In addition to the standards run during data collection for this study, we have also subsequently measured ~1-ng samples of natural standard BCR-2 processed using the same protocols as our CAI samples in the VUW laboratories but after the Sr spike material had been opened and used extensively in the laboratory. The NBS-987 data on picogram-level loads (table S2 and fig. S11) serve to demonstrate that any possible contamination could not have occurred at the loading (VUW) or analysis (Bremen) stages, as the 1- to 2-ng-level BCR-2 data (i.e., comparable sample loads to the L4, L5 fractions were analyzed in this study) show no detectable contamination (table S3 and fig. S11).

Mass spectrometry

Strontium isotopic compositions were measured using two TIMS instruments. A Thermo-Finnigan Triton at VUW, equipped with nine Faraday collectors connected to 1011-ohm feedback resistor amplifiers, and a Thermo Scientific Triton XT at the Thermo Fisher Scientific factory in Bremen, Germany, equipped with a switchable selection of both 1011- and 1013-ohm amplifiers. Gain calibration was performed before each day of analysis, and the interchannel gain stability of the 1011- and 1013-ohm amplifier equipped instruments was better than 25 ppm/24 hours and 10 ppm/24 hours (2 SD), respectively.

The VUW instrument was used for the analysis of the fractions L1 to L3, where the Sr load was >100 ng in all cases. Analyses consisted of 540 cycles (16-s integrations) measured statically with amplifier rotation, a 1-min baseline measurement, and a prebaseline wait time of 20 s after each rotation (4, 8). The Bremen instrument, with the improved signal-to-noise ratio afforded by the 1013-ohm amplifiers (30), was used for analysis of the smaller load sizes produced in leach fractions L4 and final digest L5. This instrument was configured with the 1013-ohm amplifiers connected to cups L1, C, H1, and H2 for measurement of masses 84Sr, 85Rb, 86Sr, and 87Sr, respectively, using cup H3 to collect 88Sr using a 1011-ohm amplifier. No amplifier rotation was performed, and a single 20-min baseline was measured with the analyzer valve closed during sample warm-up. After tuning and instrument optimization, measurements were made in static mode, using 4.194-s integration time, and collecting up to 400 cycles, or until the sample was completely exhausted. After each block of 40 cycles, the signal was checked and the evaporation filament current adjusted by the software so that the signal remained within 80 and 120% of the intensity measured at the start of the run. The bulk CAI samples were run later on the VUW instrument after upgrading with four 1013-ohm amplifiers, one of which was used to measure mass 84 during those runs (other masses measured on 1011-ohm amplifiers).

For all measurements, any 87Rb interferences at mass 87 were corrected for using the beam intensity measured at 85Rb and assuming 87Rb/85Rb = 0.385707 (59). In all runs, the raw 87Rb contribution to the 87Sr signal never exceeded a few microvolts and was observed to decay rapidly before the start of analysis. Natural (unspiked) runs were corrected for instrumental fractionation using the exponential law to 86Sr/88Sr = 0.1194. Double spike measurements and subsequent data reduction were carried out using the methods detailed in Charlier et al. (4, 8). All samples were loaded on outgassed zone-refined single Re filaments, and a Ta2F5 activator (60) was used to increase ionization efficiency. Natural and spiked measurements were made on separate sample wheels with a clean ion-source extraction plate for each wheel change. Loading of filaments was carried out at VUW using identical techniques throughout and the same batch of Re wire. Care was taken to retain consistency in both the nature and quality of sample loading with the intention of keeping the total procedural blank as uniform as possible for all analyses. Filaments loaded at VUW were sealed within a clean environment and carefully transported to Bremen for analysis.


Following the convention for the reporting of 84Sr anomalies used in other studies, we use the μ84Sr notation (i.e., parts per 106 deviation of the 84Sr/86Sr, internally normalized to 86Sr/88Sr = 0.1194, from the average value of the standard)μ84Sr=[(S84r/S86r)samp(S84r/S86r)NBS987 ave1]×106

Stable Sr data derived from the double-spike natural measurements are presented in the delta notation (in per mil) relative to our NBS-987 standardδ84/86Sr ()=[(S84r/S86r)samp(S84r/S86r)NBS9871]×103δ88/86Sr ()=[(S88r/S86r)samp(S88r/S86r)NBS9871]×103

All of the fractionation-corrected 87Sr/86Sr data have been renormalized to NBS-987 = 0.710250.

Data precision and accuracy

The NBS-987 strontium standard was run at various load sizes to assess the reproducibility for the load sizes measured on each instrument. The details of these measurements are given in table S2, and the standard reproducibility was summarized in Results. For the large loads encountered in L1 to L3, internally normalized measurements of nine ~1000-ng loads of the standard were carried out over the period of analysis on the VUW instrument. Given that the double-spike data involve two independent mass spectrometric measurements, eight NBS-987 loads optimally spiked with the VUW 84Sr-87Sr double spike were measured during the same period, and the data were reduced against each of the natural NBS-987 runs to yield 72 possible permutations to calculate the external reproducibility. For the Bremen instrument, 10 individual measurements of 1-ng, 500-pg, and 50-pg NBS-987 loads were carried out. One sample of the 50-pg loading was discarded because of an ion yield of <0.1%. Average Sr ion yields (data were only measured for acquisition where signals were higher than 10 mV on 88Sr) were 1.5, 1.8, and 1.5%, and the average 84Sr signals were 5.5, 3.2, and 0.3 μV, for, respectively, the 1-ng, 500-pg, and 50-pg Sr loadings. The external precisions for individual analyses (table S2) are within a factor of 1.5 of the average internal precisions as defined by the counting statistics. For the stable strontium measurements, an overall external reproducibility for δ84/86Sr and δ88/86Sr for small load sizes was determined for the 1-ng loads where the 10 analyses of unspiked, natural NBS-987 were deconvolved against nine optimally spiked 1-ng aliquots.

As an additional demonstration of data consistency between the VUW and Bremen instruments, both the natural and spiked aliquots of two samples (L3_Jean and L3_Logan) were measured first at VUW then later at Bremen (fig. S14) and yielded identical results (within uncertainty). These samples were chosen on the basis that enough Sr remained on the filament for two measurements since the measured isotopic composition for the natural aliquots had not yet fractionated through the 86Sr/88Sr = 0.1194 value assumed for fractionation correction.

To check for the integrity and consistency of isotopic data from the gravimetric recombined leach fractions, we carried out analysis of total digests of the remaining fractions of bulk powder from the individual CAIs. In addition, these analyses served to constrain the magnitude of any possible spike contribution to any 84Sr values in the leachate samples from the natural runs. The μ84Sr values from the bulk samples are presented in table S4 and plotted in Fig. 2 and show good agreement with the reconstituted L1 to L5 compositions. In particular, Jean, which has the largest μ84Sr value, shows a match within error for reconstituted L1 to L5 and the bulk sample, and both of these are distinct from the L1 to L3 data (Fig. 2).

Possible sources of inaccuracies

Blank contributions. An identical conical beaker to that used in the step-leaching procedure, but containing no sample, was manipulated in the same way as the sample beakers and the subsequent sample-free supernatants passed through identical column chemistry and aliquoting for trace element analysis. The blank amount for each leach step was calculated via isotope dilution (ID) using an 84Sr single spike, which was diluted from an “as-supplied” IRMM-635 1 M HNO3 stock solution and calibrated for concentration against a gravimetric solution of NBS-987. ID mass spectrometric measurements were made at different times to those of the unknowns and used an entirely separate suite of labware and filament components to circumvent any possibility of cross-contamination with 84Sr.

To blank-correct the isotopic data, both the blank amount and its isotopic composition must be known. From past work in the VUW laboratory, the largest contributor to the total procedural blank (TPB) has been found to derive chiefly from the Sr-spec resin used for separation. The wash fractions from the resin cleaning procedure were thus collected (60), and their isotope composition was precisely measured, yielding a value of 87Sr/86Sr = 0.709866 ± 0.000003 (2 SD; normalized to NBS-987 87Sr/86Sr = 0.710250) as a best estimate for the TPB composition for correction of radiogenic data. The internally normalized 84Sr/86Sr (μ84Sr) data were blank-corrected using the measured 84Sr/86Sr value for NBS-987 obtained from the respective instrument. Double spike data were corrected for blank assuming “average BSE” values of 0.3 and −0.3‰ for δ88/86Sr and δ84/86Sr, respectively (8). The TPBs were remarkably consistent at 55 ± 13 pg (2 SD) total strontium for all leach steps. In the case of L1 to L3, the TPB never exceeded 0.14% of the Sr present. For steps L4 and L5, TPB contributions ranged from 0.23 to 5.30% and 0.082 to 24.82% of the Sr present in each sample aliquot, respectively (table S4). The two TPBs measured for the bulk CAI digests gave an average of 66.2 pg, representing a maximum blank contribution of 0.11% in these analyses. To ensure consistency, blank corrections were made to the isotopic data for the bulk analyses and all leach fractions, regardless of total strontium content.

Tests for coherent fraction behavior of processed samples. To test for the possibility that separation of Sr from a matrix dominated by Ca and Al could lead to anomalous fractionation behavior during ionization from the filament, some “synthetic CAI” solutions were prepared. These solutions were made from ICP-MS single-element standards containing an identical mixture of the major elements (Mg, Al, Ca, and Fe) to mimic the composition of the final leachates determined for the three most extreme μ84Sr-bearing samples (L5_Hank, L5_Ororo and L5_Peter; μ84Sr = +80655, +11599, and + 5225, respectively). Ten nanograms of NBS-987 Sr was added to each of these solutions, which were then subjected to the same Sr-spec column separation as that used for the unknowns throughout this study. Natural logs of the measured ratios of the separated Sr were used to produce regression lines in three-isotope 87Sr denominator space, from which a slope could be determined using a best-fit York-type regression to the block average data. The slopes of the lines for each synthetic solution were then compared and found to agree, within uncertainty, with those expected for theoretical exponential law fractionation (table S5). This demonstrates that the highly anomalous data for residues measured in L4 and L5 are not an artifact of nonexponential law ionization behavior during TIMS analysis and that the use of internal normalization using the exponential law is justified.

Tests on the extraction tools used for sampling CAIs. The CAIs were extracted from their host meteorite using cleaned stainless steel and nickel-cobalt-molybdenum alloy dental tools [after (54)]. To test for possible contamination from these tools, their Sr contents and isotopic compositions were determined. Tool fragments were removed using a diamond saw blade, lubricated, and cooled with water. Each tool was cut half way through then covered with a laboratory tissue and snapped by hand against a hard surface. The fragments of the four tools (DT-1 to DT-4) were cleaned repeatedly using Milli-Q H2O and ultrapure ethanol in an ultrasonic bath and then dried in an oven. They were then weighed and dissolved using increasingly aggressive acids, culminating in full digestion in aqua regia for the stainless-steel tools (DT-1, DT-3, and DT-4). DT-2 (NiCoMo alloy), however, was not fully dissolved even after protracted acid digestion at elevated temperatures. This fragment was reweighed, and the difference between the initial and incompletely digested fragment mass was taken as the sample weight in solution. All solutions were equilibrated in 9 M HNO3 before aliquoting and column chemistry. A ca. 10% aliquot was mixed gravimetrically with the same 84Sr spike as that used for TPBs to determine the Sr concentration via ID, and the remainder left unspiked for isotopic analysis. Both aliquots were processed separately through columns using the same methods as for the natural samples. Sr concentrations and isotopic compositions in dental tool materials are given in table S6.

Given that the tools were not weighed before and after the CAI sampling, their exact contribution to the Sr budget of the sample is uncertain, yet, as shown below, significant contamination can be ruled out. For all samples, full digestion after the final bomb digest was ascertained by careful inspection of the liquid under a binocular microscope after removal from the bomb. No discernable particulate matter was observed in any samples after the final digestion (L5). Given the resistance to acid digestion described above (particularly for DT-2), any metal swarf generated from the tools from abrasion against the sample would have been unlikely to be dissolved in the first three leach steps and may possibly have survived through to the final digest where its effect on the isotopic composition of the sample would have had the largest influence.

To test the possible influence of the dental tools on the isotopic composition of the samples, we examined an extreme worst-case scenario of 1 mg of dental tool being present in the final digestion. This figure is based on observations that the tools were visually unaffected by the abrasion process, and a loss of 1 mg or more of material would have been obvious from visual inspection. In terms of the 84Sr influence, we consider CAI sample L5_Hank, which contained the smallest Sr amount (0.372 ng) but the most extreme μ84Sr (+80,655). Taking DT-2 as the tool likely to exert the largest effect on the Sr budget of the sample (due to its high Sr content; 215.6 parts per billion), 1 mg of this tool represents 0.2156 ng of Sr. For L5_Hank, this would correspond to a “blank” contribution of ~58% and yield a μ84Sr value blank-corrected upward to +192,184 (on account of the μ84Sr of the dental tools having essentially normal μ84Sr). In contrast, correction of the measured 87Sr/86Sr would be revised downward from 0.704163 to a very unradiogenic value of 0.697339. In this scenario, the presence of dental tool material in the analysis would serve to reduce the measured μ84Sr, not increase it, but raises the possibility that the 87Sr/86Sr might be shifted to apparently more radiogenic measured values if this dental tool material were present.

In terms of 87Sr, to test whether the more radiogenic 87Sr/86Sr values encountered for some samples (in particular, in the L5 step) were affected by contamination from dental tool material, we consider the possible effects of tool material addition to L5_Scott. This sample contained 58.2-ng Sr in the final digest and has the most radiogenic measured 87Sr/86Sr of all the final leach fractions (87Sr/86Sr = 0.706701). Again, assuming the worst-case scenario of 1 mg of tool contamination from DT-2, the blank contribution in this situation would be only 0.37%. This would result in negligible shifts in the corrected ratios (μ84Sr, +854 revised upward to +857; 87Sr/86Sr, 0.706701 revised downward to 0.706692). On this basis, the values measured for CAI 9 and others with extreme values in the L5 step must therefore represent the primary isotopic compositions of these samples.

The same mass balance considerations using the other three sampling tools yield the same conclusions and demonstrate that the observed highly anomalous μ84Sr values encountered in steps L4 and L5 cannot be attributed to any admixture of material from the sampling process, since none of the dental tools have an elevated μ84Sr. Similarly, the measured, highly radiogenic 87Sr/86Sr values in steps L4 and L5 cannot be attributed to material derived from any of the dental tools used and must be of primary origin, particularly those measured in the L5 step. Last, the influence on the measured Sr isotopic compositions of any dental tool material contribution at an earlier stage in the leaching process (L1 to L3) would have been trivially small on account of the much higher Sr contents encountered in these preceding leach fractions.

Geochemical insights into the nature of the 84Sr carrier

In theory, the geochemical signatures of the successive leachates could shed light on the nature of these carrier phases. As major and trace elements were measured on the leachate solutions, we attempted to place constraints on the nature of the carrier using these data. The most informative cross-plots of major and trace element atomic ratios in the leachates are shown in fig. S13. Although the data clearly indicate which of the major phases is being digested in which step (i.e., mixing arrays in between well-defined poles), they do not, however, allow us to pin down the nature of the carriers. This is evident when looking at the third and fourth panels (in which the data for L5 are shown). The chemical composition of the leachate is overwhelmed by those of the major phases, and the L5 leachates define mixing arrays between a spinel pole and an Al-rich diopside pole. This is consistent with a very low abundance of nucleosynthetic anomaly carriers with extremely high 84Sr excess, as observed in L5 in the Hank CAI. Establishing the nature of the p-process carriers will therefore require additional work and the physical isolation, imaging, and isotopic analysis of the carrier. These steps are beyond the scope of the present study.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: CAI samples were generously provided by D. Ebel and S. Alpert (American Museum of Natural History). B.L.A.C. thanks the late C. Douthitt for support for the analytical work at Bremen, and F.L.H.T. thanks D. McGee for providing access to his clean laboratory for sample extraction. Funding: This research was supported by a VUW Research Establishment Grant 225107 to B.L.A.C.; an NSF-EAR grant (1824002), a Crosby Postdoctoral Fellowship (from MIT), and start-up funds (provided by Caltech) to F.L.H.T.; and NASA grants 80NSSC17K0744 (Habitable Worlds), 359NNX17AE86G (LARS), NNX17AE87G, and 80NSSC20K0821 (Emerging Worlds), and NSF grant EAR-2001098 (CSEDI) to N.D. Author contributions: B.L.A.C., F.L.H.T., and N.D. designed the research. F.L.H.T. selected and extracted the samples. F.L.H.T. and R.T.M. characterized the samples. B.L.A.C. processed the samples and developed techniques with H.V. B.L.A.C. and H.V. analyzed the samples. B.L.A.C., F.L.H.T., N.D., R.T.M., and C.J.N.W. interpreted the data. B.L.A.C., F.L.H.T., and C.J.N.W. wrote the first draft of the manuscript, which was subsequently edited by all authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional date related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article