EVIDENCE FOR MULTIPLE SOURCES OF ¹⁰Be IN THE EARLY SOLAR SYSTEM

With a half-life of 1.4 Myr (Korschinek et al. 2010), the decay of ¹⁰Be to ¹⁰B is a key indicator of late spallogenic contributions toward the nucleosynthetic make-up of the solar system, given that ¹⁰Be is destroyed during stellar nucleosynthesis and formed near-exclusively by spallation reactions associated with cosmic rays (Clayton 2003). Evidence for the former presence of ¹⁰Be in meteorites has been inferred from the observation of a correlation between the excesses in ¹⁰B and the Be/B ratios of minerals in calcium–aluminum-rich inclusions (CAIs; McKeegan et al. 2000; Sugiura et al. 2001; MacPherson et al. 2003; Chaussidon et al. 2006), which are believed to be the earliest solids formed in our solar system (Amelin et al. 2011). The range of the solar system's initial ¹⁰Be/⁹Be ratios of ∼(3–8) × 10⁻⁴ inferred from these studies is higher than the expected contribution from the galactic background of ∼10⁻⁵ (McKeegan et al. 2000), requiring a late nucleosynthetic production of ¹⁰Be prior to, or shortly after the formation of our solar system. Several mechanisms have been suggested for this late enhancement of ¹⁰Be, including magnetic focusing and enhanced trapping of galactic cosmic rays (GCRs) in the protosolar molecular cloud (Desch et al. 2004), ¹⁰Be formation in an X-wind-type setting through spallation of solar nebula gas by solar cosmic rays (Clayton & Jin 1995; McKeegan et al. 2000), or in situ ¹⁰Be formation through irradiation of the refractory inclusions or their precursor materials (Lee et al. 1998; Gounelle et al. 2001; Liu et al. 2010). The proposed mechanisms can be tested by searching for non-chronometric variations in ¹⁰Be/⁹Be and cogenerated Li and B isotope anomalies among refractory inclusions that as a result of their varying petrogenesis may have preferentially sampled or recorded these reservoirs and processes.

Carbonaceous chondrite meteorites of the CV type (Vigarano type) contain refractory inclusions that formed with the canonical ²⁶Al/²⁷Al ratio of (5.252 ± 0.019) × 10⁻⁵ (Larsen et al. 2011; i.e., canonical CAIs) as well as the so-called FUN-CAIs (Fractionation and Unidentified Nuclear isotope anomalies; Wasserburg et al. 1977). Canonical CAIs appear to have formed in a gas of approximately solar composition as condensates from nearly completely vaporized nebular material and/or as remelted condensates (Grossman et al. 2000; Krot et al. 2004; Sugiura et al. 2009), possibly within 0.1 Myr of solar system formation (Bizzarro et al. 2004; Thrane et al. 2006; Jacobsen et al. 2008; Larsen et al. 2011). The CV CAI-forming reservoir was apparently homogeneous with respect to ²⁶Al (Larsen et al. 2011), oxygen (Krot et al. 2010), and other stable nuclides such as ⁵⁰Ti and ⁵⁴Cr (Trinquier et al. 2007, 2009). Therefore, these objects are samples of the homogenized refractory components of the gas, including its Be, and should not exhibit variable ¹⁰Be/⁹Be unless ¹⁰Be was locally synthesized in the gas or in the inclusions by exposure to particle irradiation. Whereas the majority of igneous canonical CAIs show small mass-dependent fractionation effects in Mg and Si isotopes consistent with melting and evaporation during transient heating events at relatively high ambient pressure (from ∼10⁻³ to ∼10⁻⁵), much larger mass-dependent fractionation effects are observed in FUN-CAIs, which appear to have experienced melt evaporation in near vacuum (Richter et al. 2007). The presence of large mass-independent isotope anomalies of nucleosynthetic origin in FUN-CAIs (Birck 2004) indicates that their precursors escaped complete evaporation–recondensation in the solar nebula. Moreover, FUN-CAIs are characterized by the absence or low initial abundance of ²⁶Al, suggesting that these objects may represent evaporative residues of molecular cloud material that formed prior to injection and homogenization of ²⁶Al in the protoplanetary disk and prior to the formation of canonical CAIs (Sahijpal & Goswami 1998). As such, FUN-CAIs may have retained the ¹⁰Be abundance of the primordial dust inherited from the protosolar molecular cloud.

To disentangle the possible nucleosynthetic contributions to the solar system's ¹⁰Be inventory, we conducted high-precision Li–Be–B isotope measurements of a suite of CV refractory inclusions. We studied four canonical CAIs (31E, E38, E48, and E104) from the reduced CV carbonaceous chondrite Efremovka and two FUN-CAIs (KT-1 and AXCAI 2771) from the oxidized CV chondrites NWA 779 and Axtell. The 31E, E38, and E48 CAIs are igneous Type B inclusions whereas E104 is a compound CAI consisting of a coarse-grained igneous Compact Type A (CTA) domain composed primarily of gehlenitic melilite, with minor perovskite, Al,Ti-pyroxene, and Fe,Ni-metal as well as a finer-grained Fluffy Type A (FTA) domain (Krot et al. 2002). 31E is a Type B1 CAI with a thick (2.5 mm) continuous melilite mantle and a core composed of pyroxene, melilite, and spinel. The inclusion is surrounded by a multilayered Wark–Lovering rim composed of spinel, anorthite, melilite, pyroxene, and forsterite. The petrology of the E48 and E38 has been described by Krot et al. (2002) and Sugiura et al. (2001). Likewise, the petrology of the FUN-CAIs KT-1 and AXCAI 2771 has been described by Thrane et al. (2008) and Srinivasan et al. (2000), respectively.

The Li–Be–B measurements were conducted by secondary ion-mass spectrometry (SIMS) using the University of Hawaii's (UH) Cameca ims-1280. The primary beam was O⁻ and the mass resolving power was set to 2400. Measuring sequence and counting times were ⁶Li (6 s), ⁷Li (2 s), and ⁹Be (2 s) in mono-collection mode, while ¹⁰B and ¹¹B were measured simultaneously (20 s), all on electron multipliers. Finally, ²⁴Mg was measured with a Faraday for 1.6 s. The analyses were mostly carried out in melilite, with a few analyses in pyroxene, and avoided any regions that experienced obvious alteration. The spots were pre-sputtered with a 30 μm rastered 10–20 nA beam for 15 minutes. Analyses were conducted with a 10 μm rastered 10 nA primary beam, while the B- and Li-rich standards were analyzed with 2–5 nA beams, in order to avoid high-intensity measurements with the electron multipliers. Ratios were calculated from counts summed over the entire analysis (Ogliore et al. 2011). Relative and absolute abundances of Li, Be, and B were calculated based on relative sensitivity factors derived from the analysis of the basaltic glass standards GSC-1G and GSD-1G, and measurements of the Mg content around the SIMS spots using the HU JEOL JXA-8500F electron microprobe. Instrumental mass fractionation for B and Li isotopes was corrected using measurement of the GSD and NIST SRM-612 standards, which have known B and Li isotope composition. After the initial sensitivity factor and instrumental mass fractionation determination, the GSD-1G standard was analyzed daily to quantify and correct for any drift. The sensitivity factor for Mg versus light elements varied more than the sensitivity factors between the light elements, and depended mostly on the analysis time/depth. We assign minimum 2σ errors of ±10% to the concentration measurements, ±5% to the ⁹Be/¹¹B and ⁹Be/⁶Li ratio measurements, and ±1‰ to the ¹⁰B/¹¹B and ⁶Li/⁷Li measurements. These errors were propagated with the counting statistics error into the final error estimates, which are 2σ confidence intervals.

Isotope and elemental data are reported in Table 1, and a summary of Be/B and 1/B regressions in Table 2. The AXCAI 2771 and KT-1 FUN-CAIs have the inferred ¹⁰Be/⁹Be ratios of (2.77 ± 0.24) × 10⁻⁴ and (3.37 ± 0.20) × 10⁻⁴, respectively, as well as ¹⁰B/¹¹B initial ratios higher than that of primitive chondrites (i.e., ¹⁰B/¹¹B ∼ 0.2475 for CI chondrites; Zhai et al. 1996; Figure 1). The Li-isotope systematics of the two FUN inclusions are complex, with ⁶Li-rich compositions present within high Li-concentration domains, indicating a spallogenic contribution to the common Li, and an additional spallogenic component present in low-concentration domains, indicating post-igneous irradiation (e.g., Figure 2). The canonical Type B CAIs 31E, E38, and E48 show initial ¹⁰Be/⁹Be ratios of ∼(4.4–4.8) × 10⁻⁴ that are not distinguishable with current uncertainties, and are significantly higher than those recorded by the FUN-CAIs (Figure 1). 31E and E38 have super-chondritic ¹⁰B/¹¹B initial ratios, while E48 is chondritic within error. They also show complex Li-isotope patterns suggesting both inherited and post-igneous spallogenic overprints. The canonical compound Type A CAI E104 shows two isochrons with initial ¹⁰Be/⁹Be of (6.70 ± 0.86) × 10⁻⁴ and (5.50 ± 1.40) × 10⁻⁴ for the FTA and CTA domains, respectively (Figure 1). Both domains have chondritic initial ¹⁰B/¹¹B ratios. The CTA domain shows a linear correlation between ⁷Li/⁶Li and ⁹B/⁶Li, possibly hinting at the former presence of live ⁷Be in this inclusion. However, given the poor fit (MSWD = 2.5), low precision, and small sample size (n = 5), this result is likely coincidental. Finally, the FTA domain shows an overall chondritic Li-isotope composition (Figure 2).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Our results confirm the presence of ¹⁰B excesses correlated with their Be/B ratios, consistent with in situ decay of ¹⁰Be at the level of ¹⁰Be/⁹Be (3–7) × 10⁻⁴. All inclusions studied show isochronous correlations between ⁹Be/¹¹B and ¹⁰B/¹¹B that are in agreement with the estimates of analytical error (Table 2), while the ¹⁰B/¹¹B versus 1/B correlations, characteristic of spallogenic imprint, are less significant and show scatter in excess of analytical error (Table 2; Figure 2). Although the initial ¹⁰Be abundances inferred from our study broadly agree with earlier studies (McKeegan et al. 2000; Sugiura et al. 2001; MacPherson et al. 2003; Chaussidon et al. 2006), our analysis of E38 and E48 suggests ¹⁰Be/⁹Be ratios that are lower than reported by Sugiura et al. (2001). We infer that the cause of these discrepancies are differences in the estimated sensitivity factors for Be/B. Sugiura et al. (2001) determined a sensitivity factor of 2.67 ± 0.09 based on a suite of geological standards with different matrices. Although this value is similar to the value of ∼2.6 we determined for the NIST 612 silica-rich glass, such standards are compositionally very different from melilite and pyroxene and may not be appropriate for high-precision Be–B studies based on these minerals. Therefore, we used a suite of basaltic glasses with variable Be and B concentrations (GSC-G and GSD-G standards), which represents a closer compositional match to melilite and pyroxene, to determine a sensitivity factor of ∼1.78. In agreement with earlier work (Ottolini & Hawthorne 1999), we suggest that matrix composition can significantly affect the relative ionization efficiencies of Be and B. While this does not impact the validity of the relative difference in the initial ¹⁰Be/⁹B ratios we document between various inclusions, it emphasizes the need for well-described and matrix-matched standards when performing inter-laboratory comparison.

Our measurements demonstrate the presence of four distinct isochrons in apparent ¹⁰Be/⁹Be (Figure 1(h)), lower in the FUN-CAIs than in the canonical CAIs, and variable within these classes. We consider several possible explanations for this feature: (1) secondary alteration, (2) post-crystallization spallogenic overprint of FUN-CAIs, (3) differences in chronology, (4) heterogeneity caused by ongoing local spallogenesis, or (5) mixing of inherited and local sources.

Although parent body alteration processes could generate variations in apparent ¹⁰Be/⁹Be from initially identical isochrons, two lines of evidence argue against this mechanism. First, all inclusions show isochronous correlations between ⁹Be/¹¹B and ¹⁰B/¹¹B that are in agreement with the estimates of analytical error, even at the highest level of precision. Second, earlier studies (McKeegan et al. 2000; Chaussidon et al. 2006) indicate that canonical CAIs from oxidized CV chondrites record initial ¹⁰Be/⁹Be ratios that are similar or higher than that we report for inclusions from the pristine Efremovka chondrite, suggesting that the lower ¹⁰Be levels found in FUN-CAIs are not related to secondary alteration processes.

It is possible that the lower initial ¹⁰Be/⁹Be ratios observed in the KT-1 and AXCAI 2771 FUN-CAIs reflect a post-crystallization spallogenic overprint, given the evidence for spallogenic Li overprints in these inclusions. For example, assuming a power law (E⁻³) distribution in proton energy and using the cross-sections of Read & Viola (1984), the linear array in 1/Li space for KT-1 (Figure 2) can be explained by exposure to a proton fluence of ∼4.5 × 10¹⁵ cm⁻² (>50 MeV). Correcting for the co-produced spallogenic B with ¹⁰B/¹¹B of ∼0.4 (Yiou et al. 1968) only modifies the ¹⁰Be/⁹Be for KT-1 from (3.37 ± 0.2) × 10⁻⁴ to (3.26 ± 0.2) × 10⁻⁴, and marginally improves the fit of the isochron (MSWD = 0.44). Subtracting higher levels of this spallogenic B from KT-1 raises the apparent slope, until the relationship becomes errorchronous by exhibiting residual scatter significantly in excess of the estimates of analytical error (MSWD = 2.3, P < 0.05) at a fluence of 1 × 10¹⁶ cm⁻² and ¹⁰Be/⁹Be of 3.76 ± 0.46 (Figure 3). In contrast, subtracting spallogenic B from AXCAI lowers the slope, until it becomes errorchronous (MSWD = 2.3, P < 0.05) at a fluence of 1 × 10¹⁶ cm⁻² with a ¹⁰Be/⁹Be of 2.43 ± 0.45 (Figure 3). Although these estimates do not consider the effects of secondary particles, and may thus not accurately reflect the fluences, they do simulate the effect of addition of ¹⁰B/¹¹B of ∼0.4. Thus, post-crystallization spallogenic overprint cannot explain the ∼10% difference in ¹⁰Be/⁹Be ratios between the KT-1 and AXCAI 2771 FUN-CAIs, or the larger contrast between the FUN-CAIs and canonical CAIs.

Zoom In Zoom Out Reset image size

The observed variability in initial ¹⁰Be/⁹Be of canonical CAIs is discordant with high-precision ²⁶Al–²⁶Mg systematics, suggesting that canonical CAIs or their precursors formed within a short time frame compared to the ∼1.4 Myr half-life of ¹⁰Be (Bizzarro et al. 2004; Thrane et al. 2006; Jacobsen et al. 2008), possibly as short as 4000 yr (Larsen et al. 2011). Similarly, the lower initial ¹⁰Be/⁹Be ratios recorded by the FUN-CAIs is difficult to reconcile with a late formation of these objects, as the ¹⁰Be/⁹Be should have decreased by an order of magnitude during the >5 Myr of free decay required to reach the low levels of ²⁶Al in FUN-CAIs, if these formed after canonical CAIs. Moreover, late formation of FUN-CAIs conflicts with their large nucleosynthetic anomalies and ¹⁶O-rich compositions, suggesting formation contemporaneous with or prior to canonical CAIs (Sahijpal & Goswami 1998).

The range in apparent initial ¹⁰Be/⁹Be ratio among the canonical CAIs is a strong argument in favor of ongoing local gaseous or in situ spallogenesis, as it is not clear how these inclusions could otherwise exhibit homogeneity with respect to ²⁶Al but not ¹⁰Be. Although the systematics of the coarse-grained canonical CAI are in most cases clearly radiogenic instead of spallogenic, this does not rule out in situ spallogenesis. Indeed, if the precursor material of coarse-grained canonical CAIs were irradiated, subsequent melting would homogenize the 1/B systematics, leaving only ¹⁰Be to form purely radiogenic signatures. In contrast, petrological and isotopic constraints indicate that fine-grained inclusions such as FTA E104 formed as gas–solid condensates (Krot et al. 2002; Larsen et al. 2011). Since they escaped melting, these inclusions should retain 1/B patterns if they experienced in situ spallation. Therefore, the preference for Be/B over 1/B type regression in FTA E104 favors in situ decay over in situ spallation. We note that MacPherson et al. (2003) reported ¹⁰Be–¹⁰B data for the FTA CAI 477-5, suggesting a ¹⁰Be/⁹Be ratio within error of that we infer for the E104 FTA domain. However, the apparent level of spallation suggested by the 1/B systematics varies by an order of magnitude between the two inclusions, suggesting that ¹⁰B excesses in these objects cannot reflect admixing of spallogenic boron. This suggests that ¹⁰Be was present in the fine-grained CAI-forming gaseous reservoir and, by extension, that canonical CAIs inherited at least some portion of ¹⁰Be from the CAI-forming gas.

The presence of ¹⁰Be in gas, however, fails to explain the abundance of ¹⁰Be in the two FUN-CAIs. Primary thermal processing of FUN-CAIs took place at low ambient pressures and, hence, with minor interaction with any solar gas. As such, the ¹⁰Be/⁹Be ratios recorded by the FUN-CAIs must reflect the ¹⁰Be abundance of their precursor material. It is possible that the FUN precursors experienced in situ spallogenesis within the solar system. A prediction of this model is the co-production of significant ⁶Li, which should shift the Li-isotope composition of a refractory precursor material by >10% (Chaussidon & Gounelle 2006). Yet, such extreme Li-isotope compositions are not observed in the FUN-CAIs. Although it is possible that the original Li-isotope compositions were erased during secondary events, the magnitude of the predicted ⁷Li/⁶Li anomalies requires that the near totality of the Li budget was exchanged with a reservoir of chondritic Li-isotope composition. However, the preservation of an ¹⁶O-rich composition in coarse-grained pyroxenes (Thrane et al. 2008) defining the post-crystallization irradiation array for KT-1 (Figure 2(d)) is at odds with pervasive isotopic exchange. We note that the ¹⁰Be/⁹Be level of ∼3 × 10⁻⁴ is in agreement with the ¹⁰Be abundance predicted to result from magnetic focusing and enhanced trapping of GCRs in the presolar molecular cloud (Desch et al. 2004). Therefore, we suggest that the ¹⁰Be/⁹Be level recorded by the FUN-CAIs represents a baseline level present in presolar material inherited from the molecular cloud, generated via enhanced trapping of GCRs. The higher and variable ¹⁰Be/⁹Be ratios present in the canonical CAIs reflects an additional ongoing ¹⁰Be generation in the gaseous reservoir from which these inclusions formed, thereby suggesting the presence of at least two nucleosynthetic sources of ¹⁰Be in the early solar system. The higher initial ¹⁰Be/⁹Be in KT-1 as compared to AXCAI 2771 is, in principle, inconsistent with a strict cloud capture scenario, as variations present in the parent cloud are expected to be efficiently homogenized during collapse. This high initial ¹⁰Be/⁹Be may reflect a higher level of pre-igneous in situ irradiation in KT-1, consistent with the anomalous Li-isotope composition of this inclusion, or alternatively a ∼150 kyr formation interval between KT-1 and AXCAI 2771.

The most promising location for spallogenesis of the additional ¹⁰Be present in canonical CAIs is close to the proto-Sun during its early mass-accreting stages, as these stages are thought to coincide with periods of intense particle irradiation similar to those inferred from astronomical observations of young stellar objects (Kastner et al. 2004; Skinner et al. 2006; Skinner et al. 2009). The variability in initial ¹⁰Be/⁹Be ratios recorded by canonical CAIs suggests that ongoing ¹⁰Be synthesis occurred within the brief formation interval of 4000 yr inferred for these objects or their precursors (Larsen et al. 2011). These timescales are consistent with the temporal evolution of FU Orionis stars (Bell & Lin 1994; Bell et al. 1995), where quiescent low mass-accretion periods (∼10⁻⁸ M_☉ yr⁻¹) lasting ∼1000 yr are punctuated by short-lived ∼100 yr intervals of high mass-accretion rates (∼10⁻⁵ M_☉ yr⁻¹) and, by extension, intense irradiation. Alternatively, it is possible that the variability in initial ¹⁰Be/⁹Be ratios reflect inherent heterogeneity in the CAI-forming gaseous reservoir associated with a single short-lived irradiation event. This would imply, however, that the formation interval of canonical CAIs was considerably shorter than the timescales required for effective isotopic homogenization of the gaseous CAI-forming reservoir.

The isochronous ²⁶Al–²⁶Mg systematics of canonical CAIs (Larsen et al. 2011) allows us to place constraints on the maximum amount of ²⁶Al co-produced during ¹⁰Be spallogenesis. We focus our discussion on the FTA E104 inclusion, as it records the highest initial ¹⁰Be/⁹Be ratio amongst canonical CAIs (Figure 3). Considering the uncertainty of the ²⁶Al–²⁶Mg regression, we calculate that FTA 104 could have formed with an ²⁶Al/²⁷Al that was at most 2.5% lower while retaining an isochronous relationship with the remaining inclusions. We note that the production of ¹⁰Be/⁹Be ∼ 10⁻³ by proton irradiation in impulsive ³He-rich events and/or steep energy spectra would result in co-production of >10% of the canonical ²⁶Al/²⁷Al (Gounelle et al. 2001, Leya et al. 2003). In contrast, first-order calculations indicate that low ³He and flatter energy spectra characteristic of gradual flares result in ²⁶Al co-production that is orders of magnitude lower (Clayton et al. 1977). Thus, the apparent ²⁶Al homogeneity is consistent with the observed ¹⁰Be heterogeneity if local spallogenesis of the additional ¹⁰Be inventory can be explained by gradual flare dominated models.

The Centre for Star and Planet Formation is financed by the Danish National Science Foundation. We thank Steve Desch for his thorough and constructive review.