New Multielement Isotopic Compositions of Presolar SiC Grains: Implications for Their Stellar Origins

Presolar grains are microscopic solids that condensed around different types of stars at their advanced evolutionary stages before the formation of the solar system. As stellar condensates, presolar grains preserve a detailed record of nucleosynthesis and mixing processes in their parent stars and thus grant us the possibility of deriving a detailed picture of these stellar processes, if we can determine the types of stars they came from. Silicon carbide (SiC) is the best-studied presolar mineral phase. There are multiple pieces of evidence that the majority of presolar SiC grains condensed from carbon stars. For instance, presolar SiC grains show a ¹²C/¹³C ratio distribution similar to that of carbon stars (Zinner 2014), SiC grains are predicted to condense in C-rich stellar environments according to thermodynamic equilibrium calculations (Lodders & Fegley 1995), and SiC dust has also been observed to be ubiquitous in circumstellar envelopes around C-rich asymptotic giant branch (AGB) stars (e.g., Little-Marenin 1986; Chan & Kwok 1990; Yang et al. 2004). Presolar SiC grains were originally divided into five main groups (≥1% of the population), including mainstream (MS), X, Y, Z, and AB. MS grains were defined as SiC grains having 10 ⪅ ¹²C/¹³C < 100 and δ²⁹Si ⁷ and δ³⁰Si correlating with a slope of ∼1.35. In comparison, Y and Z grains have Si isotopic compositions deviating from the linear correlation observed for MS grains toward larger ³⁰Si enrichments relative to ²⁹Si, with the deviations larger for Z grains than for Y grains. In addition, while MS and Z grains overlap in their C and N isotopic compositions, Y grains were defined to have ¹²C/¹³C ≥100. AB grains have low ¹²C/¹³C ratios (⪅10) and a similar range of Si isotopic compositions as MS grains but exhibit a wider range of ¹⁴N/¹⁵N ratios. Finally, X grains are characterized by large ¹⁵N and ²⁸Si excesses and a wide range of ¹²C/¹³C ratios (see Zinner 2014 for a review). These classifications were based on C, N, and Si isotopic ratios mainly because, as the most abundant elements in the grains, they are the easiest to measure, and these elements thus dominate the existing isotope database for presolar SiC. In fact, interpreting the C, N, and Si isotopic compositions of presolar SiC in the context of stellar nucleosynthesis is quite complex. Thus, linking the different groups to different types of carbon stars has proven to be quite a challenge (e.g., Zinner et al. 2006, 2007). Boujibar et al. (2021; hereafter B21) recently assessed the existing classification scheme by applying the technique of cluster analysis, and their results highlight the importance of ¹⁴N/¹⁵N and the inferred initial ²⁶Al/²⁷Al ratios (hereafter ²⁶Al/²⁷Al, determined from ²⁶Mg excesses) to presolar SiC grain classification and the need for more multielement isotopic data for more accurate classification.

Multielement isotopic data for individual presolar SiC grains are limited. There are only ∼100 MS grains with both ¹⁴N/¹⁵N and ²⁶Al/²⁷Al (<50% errors) in the Presolar Grain Database (PGD; Hynes & Gyngard 2009; Stephan et al. 2021), although >10,000 MS SiC grains have been examined for their isotopic compositions. Moreover, asteroidal and terrestrial contamination complicates the data interpretation. Although the wide range of ¹⁴N/¹⁵N ratios observed in MS grains may provide important constraints on astrophysical processes, it is possible that some or all of this large range is simply caused by differing amounts of N contamination sampled during laboratory analyses. Furthermore, evidence that Al contamination strongly affected ²⁶Al/²⁷Al data in the literature was presented by Groopman et al. (2015). For the case of some trace-element measurements, e.g., Sr and Ba, special cleaning procedures have been developed (e.g., Liu et al. 2015), but similar methods have not been applied to studies of important elements like N, Mg, and Al.

In this study, we examined the problem of contamination by adopting two approaches. For C, N, and Si isotopes, we conducted NanoSIMS analyses on the same grains after (i) minimal presputtering (a few seconds to minutes of sputtering for the Si count rate to reach equilibrium) and (ii) extensive presputtering for comparison (see Appendix A for details), which enabled us to determine the source and degree of contamination for these elements. For Mg–Al, Ca–Ti, and Ti–V isotopic systematics, we used an RF plasma primary ion source, the Hyperion (Oregon Physics, LLC), which produces an O^– beam of reduced size compared to previous duoplasmatron ion sources (Malherbe et al. 2016). The improved spatial resolution enabled us to suppress contamination from the surrounding substrate for these elements.

All of the isotopic data were collected with the Cameca NanoSIMS 50L instrument at the Carnegie Institution in imaging mode; the analytical details are given in Appendix A. We obtained Si and Mg–Al isotopic data for all 85 grains investigated (Tables 1 and A1) and Ti isotopic data for 35 of the grains (Table A1). Prior to these isotopic analyses using the Hyperion O^– source, we used a Cs⁺ ion source to obtain C, N, and Si isotopic ratios (hereafter pre-Hyperion data) for all AB and X grains ⁸ and 25 of the MS and Y grains. For 47 of the MS and Y grains, we also collected their C and N isotopic data after the Hyperion analyses on the remaining grains (hereafter post-Hyperion data). The C and N isotopic ratios and ²⁶Al/²⁷Al are reported as absolute ratios, while the Mg, Si, and Ti isotopic ratios are reported in δ notation, defined as δⁱA = ((ⁱA/^jA)_grain/(ⁱA/^jA)_standard–1) × 1000, in which (ⁱA/^jA)_grain and (ⁱA/^jA)_standard denote the measured ratios of a grain and standard, respectively; the normalizing isotopes for Mg, Si, and Ti are ²⁴Mg, ²⁸Si, and ⁴⁸Ti, respectively. All data are reported with 1σ errors.

A comparison of the pre-Hyperion, post-Hyperion, and PGD data (PGD_SiC_2021-01-10; Stephan et al. 2021) for MS, Y, and Z grains is given in Figure 1. We chose these three types of grains for comparison because MS grains are the dominant type (Zinner 2014), and the three types are all inferred to have come from low-mass AGB stars given their comparable s-process Mo isotopic signatures (Liu et al. 2019). Figure 1(a) shows that (1) the PGD data have a distribution (in gray-scale color map) that peaks at ¹²C/¹³C ≈ 50–55 and ¹⁴N/¹⁵N ≈ 500, (2) our pre-Hyperion data for 25 MS and Y grains (Table A1) overlap with the PGD data (our average ¹²C/¹³C = 59.5, ¹⁴N/¹⁵N = 756), and (3) the post-Hyperion data for the same grains (Table 1) show significantly enhanced ¹⁴N/¹⁵N ratios, with the average being shifted upward to ∼2000. The significant increase in ¹⁴N/¹⁵N between the pre- and post-Hyperion data suggests that N contamination was present across the whole grain surface during the pre-Hyperion measurements, which could not be effectively suppressed even by selecting small regions of interest (ROIs) with low N within the grain ⁹ during the data reduction; instead, extensive sputtering was needed to expose clean grain surfaces for obtaining intrinsic N signals. Differences in ¹²C/¹³C between the pre- and post-Hyperion data are overall small; the averages are 59.5 and 56.7, respectively. The probability distribution of our 47 MS and Y grains (post-Hyperion data in Table 1) is shown in Figure 1(b) for comparison with the PGD data in Figure 1(a). Given its absence in our post-Hyperion data, the secondary peak at ¹²C/¹³C ≈ 90 and terrestrial/solar-like ¹⁴N/¹⁵N ratios for the PGD data are most likely the result of significant asteroidal/terrestrial contamination.

Zoom In Zoom Out Reset image size

Our MS and Y grains show ²⁶Mg/²⁴Mg values that are more than one order of magnitude higher than those of Huss et al. (1997). ¹⁰ Our higher ²⁶Mg/²⁴Mg values result from sampling less Mg contamination, as evidenced by our lower Mg/Si (Figure 2(b)) and enhanced Al/Mg ratios. The Huss et al. data and our data lie close to the same trend line in Figure 2(a). Hence, the differences in the Al/Mg ratios between the two studies cannot be explained by higher levels of Al contamination in our analyses because this would have moved our data horizontally to the right of the trend line in Figure 2(a). Thus, Figures 2(a) and (b) together point out that our study sampled much reduced Mg contamination, which was made possible by both the high spatial resolution achieved with the Hyperion O beam and the use of ROIs that were smaller than the full grains to exclude adjacent Mg contamination (see Figure A1 and Appendix A for details). This was not possible with the ion probe technology of the 1990s.

Zoom In Zoom Out Reset image size

The high spatial resolution and exclusion of Al contamination (Figure A1) also resulted in higher ²⁶Al/²⁷Al ratios, calculated by assuming that all ²⁶Mg excesses were produced by the decay of ²⁶Al (see Appendix A for details), for our MS and Y grains (Figure 2(c)). We simulated the effect of having a larger primary beam on the levels of Al contamination by recalculating the ²⁶Al/²⁷Al ratios while including all Al signals within each grain (blue circles in Figure A2). This lowered the ²⁶Al/²⁷Al ratios by up to 60% (45% on average). The effects of Mg and Al contamination on reducing the Mg isotopic ratios and ²⁶Al/²⁷Al ratios, respectively, are further illustrated in Figure A3 by comparing the density distributions of our new data with those from Huss et al. (1997).

Our high-resolution ion images also revealed (1) common Ca and Cr contamination present on the substrate and (2) insignificant Si, Ti, and V contamination (Figure A1). The literature MS grains have an average δ⁵⁰Ti of 179‰ ± 75‰ (one standard deviation), and our MS grains have 153‰ ± 95‰. In agreement with the literature data, our AB grains exhibit less correlated Si and Ti isotopic compositions than the MS grains. For instance, δ⁴⁶Ti and ²⁹Si show a strong correlation among MS grains (with Pearson's correlation coefficient R = 0.79 and 0.78 for the PGD and our data, respectively) but a weaker correlation among AB grains (with R = 0.58 and 0.44, respectively).

In summary, we identified substantial N and Al contamination, which has certainly impacted the literature data by, for example, lowering the ¹⁴N/¹⁵N and initial ²⁶Al/²⁷Al ratios in the PGD and thus compromising the inferred ratio distributions. In the next section, our discussion focuses on the grain classification and AGB stellar nucleosynthesis based on our new ¹⁴N/¹⁵N and ²⁶Al/²⁷Al results for MS, Y, and AB grains.

3.1. Comparison with Observations for Carbon Stars

Stellar nucleosynthesis model calculations suffer from uncertainties in the initial stellar composition, nonstandard mixing processes, and nuclear reaction rates (see Busso et al. 1999; Palmerini et al. 2011 for reviews). Given the challenges in using model calculations to explain presolar grains' C and N isotopic ratios, we first compare the MS, Y, and AB grain data with observations for carbon stars (Table 2) in Figure 3(a). N-type carbon stars represent classical AGB stars and are the most numerous type of carbon star in the Galaxy. Most N-type carbon stars with near-solar initial metallicities are enhanced in s-process elements and F, which results from the recurrent mixing of He-burning products from the stellar interior into the envelope through third dredge-up (TDU) episodes (Abia et al. 2001, 2008; Abia 2008). J-type carbon stars represent about 10%–15% of all Galactic carbon stars (Abia & Isern 2000; Morgan et al. 2003) and are characterized by peculiar chemical compositions, including low ¹²C/¹³C ratios, Li enhancements (80% of the stars), and no enhancement in s-process elements (e.g., Sr, Zr, Ba, Nd) or F (Abia & Isern 2000; Abia et al. 2015). The origin of J-type carbon stars is unknown. SC-type carbon stars are rare in the Galaxy (Abia & Wallerstein 1998). They are characterized mainly by C/O ratios of around unity and are chemically similar to N-type carbon stars. In the classical picture, an SC-type carbon star represents a brief AGB phase when an oxygen-rich, class-M star transitions into a carbon star as TDU episodes bring increasing amounts of ¹²C to the stellar surface. N- and SC-type carbon stars in the solar neighborhood are inferred to be 1.5–3 M_⊙ AGB stars according to their derived luminosity functions (Abia et al. 2020).

Zoom In Zoom Out Reset image size

Table 2. Abundances and Chemical Characteristics of Carbon Stars

Carbon Star	Abundance	12C/13C	14N/15N	s-process Enrichment
N-type	Large majority (e.g., Lloyd Evans 2010)	20–90	1000–5000	Yes
J-type	10%–15% in the Galaxy and the Large Magellanic Cloud	≤13	1000–3000	No
SC-type	Rare (Abia & Isern 2000; Morgan et al. 2003)	5–50	100–1000	Yes

Download table as:  ASCIITypeset image

3.1.1. MS/Y Grains and N- and SC-type Carbon Stars

The grain-observation comparison in Figure 3(a) links MS and Y grains to N-type carbon stars, which are classical low-mass AGB stars. MS and Y grains make up more than 90% of presolar SiC grains, which is consistent with the observation that N-type carbon stars are the large majority of carbon stars in the Galaxy and the Large Magellanic Cloud (e.g., Lloyd Evans 2010). Our MS and Y grain C and N isotopic data are very similar to those of N-type carbon stars in Figure 3(a); i.e., the stars mostly lie within the region where most of the grain data are plotted (Figure A4). The consistency implies that N contamination was effectively suppressed in our analyses, such that the measured N (and C) isotopic ratios are representative of the envelope compositions of the grains' parent stars. Thus, our results solve a long-standing discrepancy highlighted by many previous studies (e.g., Hedrosa et al. 2013), in which ¹⁴N-poor MS grains could not be reproduced by theoretical AGB stellar models and were linked to astrophysical sites other than N-type carbon stars.

Despite the general consistency, a slight difference is seen in the range of ¹⁴N/¹⁵N; while N-type stars generally have ¹⁴N/¹⁵N above 1000, some of our MS grains show ¹⁴N/¹⁵N down to the terrestrial value. This raises the question of whether these grains represent SiC from a different type of star or their lowered ratios were caused by sampling N contamination. It was pointed out by B21 that MS grains with ¹⁴N/¹⁵N ≲ 1000 (hereafter ¹⁵N-rich MS grains; brown ellipse in Figure 3(b)) tend to show higher ²⁶Al/²⁷Al than those with higher ¹⁴N/¹⁵N (hereafter ¹⁴N-rich MS grains; blue ellipse), on the basis of which they divided MS grains into two groups. In agreement with the B21 results, ²⁶Al/²⁷Al ≳ 3 × 10⁻³ was found in four of our seven ¹⁵N-rich MS grains (57.1%) but none of our 33 ¹⁴N-rich MS/Y grains. However, Figure 3(b) shows that the definition for these two clusters given by B21 based on the PGD is not applicable to our new grain data due to the PGD's N and Al contamination problem.

Figure 3(a) suggests that the ¹⁵N-rich MS grains may have come from SC-type carbon stars. The rareness of SC-type carbon stars (3% of all AGB stars) naturally explains the scarcity of ¹⁵N-rich MS grains. Abia et al. (2017) observed that, compared to N-type carbon stars, SC-type carbon stars tend to exhibit higher ¹⁷O/¹⁶O ratios, leading to the suggestion that SC-type carbon stars could have higher initial stellar masses (≳3–4 M_⊙) than typical N-type carbon stars (≲3 M_⊙). However, the stellar models of Karakas (2014) for 3–4 M_⊙ carbon stars (compared to ≲3 M_⊙ carbon stars) predict enhanced ¹⁴N production and no ²⁶Al/²⁷Al increase at the stellar surface (Karakas & Lugaro 2016) and thus cannot explain the ¹⁵N- and ²⁶Al-rich nature of ¹⁵N-rich MS grains. Instead, these minor isotope enrichments, together with the observed high ¹⁷O/¹⁶O ratios for SC-type carbon stars, point to the occurrence of high-temperature (≳1 × 10⁸ K) H-burning process(es), i.e., nova-like nucleosynthesis (José et al. 2004), in SC-type carbon stars, which are not predicted by any of the state-of-the-art stellar models for 1.5–4 M_⊙ carbon stars and need further investigation.

3.1.2. AB Grains and J- and SC-type Carbon Stars

Type AB grains with higher-than-solar ¹⁴N/¹⁵N ratios are consistent with observations of both J-type carbon stars and a few SC-type carbon stars in Figure 3(a). AB grains were defined as having ¹²C/¹³C ⪅ 10 and a wide range of ¹⁴N/¹⁵N ratios. Based on differences in the ²⁶Al/²⁷Al, Ti, and Mo isotopic ratios, Liu et al. (2017a, 2017b, 2018a) proposed to further divide AB grains into two subgroups, AB1 (with subsolar ¹⁴N/¹⁵N) and AB2 (with higher ¹⁴N/¹⁵N), using the solar ¹⁴N/¹⁵N as a divider. This division scheme is supported by the observed ¹⁴N/¹⁵N ratios of SC- and J-type carbon stars. The B21 cluster analysis also suggested that there are two clusters of AB grains but with some overlap in their ¹⁴N/¹⁵N ratios (ellipses in Figure 3(a)). Liu et al. (2017b) proposed that AB2 grains mostly came from J-type carbon stars based on their general lack of s-process Mo isotopic signatures. However, two of the 12 AB2 grains from Liu et al. (2017b) and one AB2 grain from Stephan et al. (2019) exhibit s-process Mo isotopic signatures and are consistent with an origin in SC-type carbon stars.

The proposal that AB1 grains originated in supernovae that experienced explosive H burning in the He/C zone (Liu et al. 2017a, 2018a) is supported by the stellar observations in Figure 3(a), given the lack of carbon stars with ¹⁴N/¹⁵N < 440 and ¹²C/¹³C ≤ 10. Schmidt et al. (2018) reported large ¹³C, ¹⁷O, and ¹⁵N enrichments in a young, C-rich planetary nebula (K4–47), based on which the authors proposed that the precursor of the nebula was a J-type star that underwent a He-shell flash. The result of Schmidt et al. (2018) raises the possibility that some J-type carbon stars have subsolar ¹⁴N/¹⁵N ratios. However, K4–47 could represent an extended nebula that was ejected by a pair of interacting binary stars during a nova-like explosion (Corradi et al. 2000), which is supported by the observation of high-velocity bullets of material plowing through the surrounding interstellar medium (e.g., Gonçalves et al. 2004). Moreover, such a low ¹⁴N/¹⁵N ratio as observed by Schmidt et al. (2018), 13.6 ± 6.5, has yet to be observed for any J-type carbon stars (Hedrosa et al. 2013). Thus, ¹⁵N-rich J-type carbon stars are relatively rare, if they exist at all. In contrast, AB1 and AB2 grains are about equally abundant, and, unlike AB2 grains, AB1 grains generally exhibit moderate s-process isotopic enrichments (Liu et al. 2018a). Since J-type carbon stars are characterized by the lack of any s-process enrichments, J-type carbon stars are unlikely to be the sources of AB1 grains.

3.2. Comparison with FRUITY Stellar Models for Low-mass AGB Stars

We would like to thank Dr. Diego Vescovi for his help with running some of the FRUITY model tests for this study. We also would like to thank Dr. Ryan Ogliore for pointing out a small mistake in the manuscript. This work was supported by NASA through grants 80NSSC20K0387 to N.L. and NNX10AI63G and NNX17AE28G to L.R.N.

The SiC grains in this study were extracted from the CM2 carbonaceous chondrite Murchison by using the isolation method described in Nittler & Alexander (2003). The extracted SiC grains have a median diameter of 0.9 μm. Isotopic data for a number of AB, Y, Z, X, putative nova, and ungrouped grains on the prepared mounts have been previously reported (Liu et al. 2017a, 2017b, 2018a, 2018b, 2019). The presolar SiC grains investigated in this study were identified by both scanning electron microscope (SEM)–based energy dispersive X-ray (EDX) analyses and isotopic analyses of C, N, and Si with the Cameca NanoSIMS 50L instrument at the Carnegie Institution. A total of 85 new SiC grains were chosen for NanoSIMS isotopic analyses in this study, including 60 MS, 15 AB, 8 X, and 2 Y grains. Note that grains on mount #4 (named M4-AX-X) were first measured for Mg–Al and Si isotopes, during which 17 of the 40 SiC grains were completely consumed and could not be further measured for their C and N isotopic compositions. While the Si isotopic data clearly indicate that two of the 17 grains are X grains given their large ²⁸Si depletions, the Si isotopic data alone cannot tell MS and AB grains apart. Also, the analytical uncertainties in the measured Si isotopic ratios are relatively large for the other 15 grains, thus precluding us from distinguishing between MS, Y, and Z grains. However, given the dominance of MS grains among presolar SiC grains (∼90%), the inferred stellar sources for these grains are likely robust, considering that the MS, Y, and Z grains all came from low-mass AGB stars and that less than one grain (by assuming 5% AB grains among presolar SiC) is expected to be of type AB. Thus, the grain-type assignment for these 15 grains solely based on Si isotopic ratios is not expected to affect any of the discussions in the main text. In addition, we chose 25 well-isolated (i.e., no adjacent grains), relatively large (0.84 μm, on average) MS and Y grains on mounts #1, 2, and 3 to investigate the problem of C and N contamination by collecting pre- and post-Hyperion C and N isotopic data (see next section for details).

Note that Figure 3(a) suggests that the MS grain M1-A4-G508 (¹²C/¹³C = 13.3 ± 0.3) could be classified as an AB2 grain, as this grain overlaps well with one of the J-type stars. The pre-Hyperion analysis for this grain yielded a ¹²C/¹³C ratio of 19.4 ± 0.1 (Tables A1). Also, note that the AB grain data in Figure 3(a) are all pre-Hyperion data, and the true ¹⁴N/¹⁵N ratios of the AB2 grains are probably a factor of 3 higher on average (see Section 2 for discussion), which explains that the AB2 grains generally show lower ¹⁴N/¹⁵N ratios than observed in J-type carbon stars (Figure A4) and consequently implies that some of the AB1 grains with terrestrial ≤¹⁴N/¹⁵N < solar could be AB2 grains instead.

All isotopic analyses were conducted with the Cameca NanoSIMS 50L instrument at the Carnegie Institution. For 48 grains, a Cs⁺ primary ion beam of ∼1 pA (16 keV energy) was first used for isotopic analyses of C, N, and Si as negative ions (N measured as CN⁻) at a spatial resolution of ∼100–150 nm (pre-Hyperion data for AB and X grains in Table 1 and those for MS/Y grains in Table A1). After that, a primary O⁻ ion beam of ∼3–10 pA produced by the Hyperion source was used here for Mg/Al, Ca, and Ti isotopic analyses, resulting in spatial resolutions of ∼150–300 nm in the obtained ion images. Similar to our previous analytical procedure (Liu et al. 2018b), we conducted Ti isotopic analyses in the multicollection mode instead of the combined analysis mode to improve the analytical stability and reliability. We measured Ca–Ti and Ti–V isotopic ratios in two sets of runs in the multicollection mode: ²⁸Si⁺, ⁴⁰Ca⁺, ⁴⁸Ti⁺, ⁴⁹Ti⁺, ⁵⁰Ti⁺, ⁵¹V⁺, and ⁵²Cr⁺ in the first set and ²⁸Si⁺, ⁴⁰Ca⁺, ⁴⁴Ca⁺, ⁴⁶Ti⁺, ⁴⁷Ti⁺, ⁴⁸Ti⁺, and ⁵¹V⁺ in the second set. For N, the high ¹⁴N/¹⁵N ratios obtained with IMS-3f (Hoppe et al. 1996; Huss et al. 1997) may suggest that extensive presputtering is an effective way to suppress N contamination. This is because, with respect to the corresponding NanoSIMS measurements, these old IMS-3f measurements had to consume more material, as the IMS-3f instrument is equipped with only one detector (five or seven detectors on NanoSIMS) and has a lower transmission efficiency. Thus, after the Hyperion-NanoSIMS analyses, we switched back to the Cs⁺ source and measured the C, N, and Si isotopic ratios in all of the remaining grains (post-Hyperion data for MS/Y grains in Table 1). The hope was that the additional sputtering of the sample and surrounding substrate would have reduced any contamination present. Note that during the post-Hyperion analyses, there was a problem with one of the detectors, and we chose to collect ¹²C¹²C, ¹²C¹³C, ¹²C¹⁴N, ¹²C¹⁵N, ²⁸Si, and ²⁹Si, which means that the ³⁰Si/²⁸Si ratio could not be determined in the post-Hyperion analyses.

All of the isotopic data were collected in imaging mode, which allowed for the selection of smaller ROIs for calculating isotopic ratios and thus reduced the likelihood of contamination (see Figure A1 for an example). The measurements were all done with an entrance slit width of 30 μm and exit slit widths of 50 (minor isotopes, e.g., ¹³C) and 80 (major isotopes, e.g., ¹²C) μm, and all of the secondary ions were counted with electron multipliers. The achieved mass resolving powers (5000–8000) were sufficient to suppress all potential molecular isobaric interferences, e.g., resolving ¹²CH from ¹³C at mass 13. The detector background was found to be between 0.001 and 0.01 cps, which compares with the lowest ion signal intensity of 0.1 cps for ²⁵Mg⁺.

Zoom In Zoom Out Reset image size

To demonstrate the effectiveness of our two approaches (extensive presputtering and reduced ROIs) in suppressing potential contamination, the two sets of ¹⁴N/¹⁵N (pre-Hyperion versus post-Hyperion) and ²⁶Al/²⁷Al (all Al signals included versus Al contamination excluded) ratios are compared in Figure A2 for MS/Y grains from this study. The Mg–Al isotopic systematics of our MS and Y grains are also compared with those from Huss et al. (1997) for their distributions in Figure A3, from which we infer significant Mg and Al contamination sampled in the study of Huss et al. (1997). Although presolar SiC grains are generally free of Ti contamination (Figures A1 and A6), the presence of Cr can result in an isobaric interference at mass 50, thus compromising the analysis of δ⁵⁰Ti. Since Ti and Cr signals do not always overlap in ion images of presolar SiC grains (Figure A1), the improved spatial resolution here allowed us to suppress the Cr contribution by choosing small ROIs to exclude Cr contamination and thus to derive more reliable δ⁵⁰Ti values (the ⁵⁰Cr contribution is generally <10%, with an average of 5%). The high spatial resolution achieved by the Hyperion source also enabled direct observation of Ti- and V-rich subgrains within the SiC host in NanoSIMS ion images (Figure A6), which had mainly been previously achieved by transmission electron microscopic (TEM) observations. The 1σ errors reported in all tables and figures include both analytical (estimated based on 1σ standard deviations of all standard measurements) and counting statistical uncertainties (1σ Poisson errors).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

A variety of standards with terrestrial isotopic compositions were used as isotopic standards and also for determining the relative sensitivity factor (RSF), defined as (A/B)_true = (A/B)_measured/RSF, in which A and B represent two elements of interest. Synthetic SiC and Si₃N₄ grains were used as standards for C, N, and Si isotopic analyses. The NIST glass standard, SRM 610, was used as a standard for Ca–Ti and Ti–V isotopic analyses, while the Burma spinel was used as a standard for Mg–Al isotopic analysis. For Ti isotopic analysis, synthetic TiC was used as an additional standard. The sensitivity factors of Mg, Ca, and Ti relative to Si were determined by measuring SRM 610, and the obtained values were 3.10, 9.78, and 4.02, respectively. The RSF of Mg/Al was determined to be 1.20 in the Burma spinel for deducing ²⁶Al/²⁷Al. The ²⁶Al/²⁷Al ratios were determined from the collected ion counts by use of the equation ²⁶Al/²⁷Al = [²⁶Mg_measured – ²⁴Mg_measured × (²⁶Mg/²⁴Mg)_standard]/(²⁷Al_measured × RSF), where (²⁶Mg/²⁴Mg)_standard is the measured ²⁶Mg/²⁴Mg ratio of the Burma spinel standard (e.g., Nittler et al. 2003). However, it is questionable whether O-rich standards such as SRM 610 are appropriate for determining RSF values for C-rich samples like SiC, and we examine this problem in Appendix B by comparing the NanoSIMS elemental data of presolar SiC grains with the corresponding EDX data obtained with an SEM and TEM when available.

Although the SIMS ionization efficiency for positive ions shows a strong correlation with the ionization potential, variations are expected depending on the sample matrix and the element itself. For example, the presence of oxygen in the sample enhances positive ion yields for most elements, and an O₂ flooding gun was therefore used in old 3f-IMS analyses to boost the yields of Mg⁺, Al⁺, and Ti⁺ ions for presolar SiC measurements (e.g., Hoppe et al. 1996; Huss et al. 1997). However, NanoSIMS is not equipped with such an O₂ flooding gun. The contrasting matrix chemistries between the standard SRM 610 (silicon glass) and SiC raise the question of whether RSF values determined based on SRM 610 measurements can be directly applied to calculate trace-element abundances (Mg, Al, K, Ca, Ti, V, and Cr) for SiC samples. In comparison to NanoSIMS analyses, SEM-EDX trace-element analyses are less affected by differing sample-standard matrix chemistries but more affected by other factors, such as topography. We therefore compared the RSF-corrected NanoSIMS Mg and Al data for SiC grains from this study and Liu et al. (2018b) with their SEM-EDX data when available. The EDX data were obtained with a JEOL 6500F field emission SEM equipped with an Oxford Instruments silicon drift detector in autoparticle analysis mode (Liu et al. 2017c). The data were quantified by using pyrope (Mg, Al garnet) as the standard for Si and ENAL20 (enstatite aluminum 20 wt%, an Al-bearing enstatite glass) for Mg and Al. In addition, EDX data were also obtained with a TEM on electron transparent thin sections that had been extracted from X grains M2-A1-G674 and M2-A2-G1036 by using a focused ion beam instrument (Singerling et al. 2021). It is noteworthy that the Mg/Si ratios in MS SiC grains are determined by NanoSIMS to be <0.003 (Table 1; also see Hoppe et al. 2015), and Mg in single presolar SiC grains is, therefore, below the detection limit of SEM-EDX analyses. The high Mg abundances reported for X grains in Figure A7(b) mainly result from ²⁶Mg from ²⁶Al decay, since most X grains are characterized by extremely high ²⁶Al/²⁷Al ratios (≳0.1; see Liu et al. 2017c for details). Errors in the trace-element data are not reported in Figure A7 because the associated statistical errors are negligible, and uncertainties in the data are probably mainly caused by analytical artifacts as mentioned above, which cannot be accurately assessed based on our obtained data.

The data comparisons shown in Figure A7 suggest that the Al/Si and Mg/Si ratios in SiC grains determined based on NanoSIMS measurements of SRM 610 are generally overestimated, i.e., the RSF values are underestimated, by factors ranging from 2 to 5. First of all, it is noteworthy that the X grain data from Liu et al. (2018b) in Figure A7 were obtained with a duoplasmatron O source at a worse spatial resolution than the grain data from this study. As discussed in the main text, this probably means that the X grain data sampled an average of 45% Al contamination (Section 2), which roughly accounts for the larger slope (by a factor of ∼2) determined by the X grain data in Figure A7(a). We excluded MS grain M1-A4-G476-1 and X grain M2-A2-G1036 (highlighted by black arrows) in obtaining the linear fits because (1) the MS grain was quite small (0.47 μm) and had significant Al contamination that could not be completely excluded by selecting a small ROI (Figure A5), and (2) the NanoSIMS data suggest that the X grain had an Al/Si ratio of 0.6, which is too high for SiC. The high Al/Si ratio of the X grain was probably caused by Al contamination, since the ²⁶Al/²⁷Al ratio inferred based on its NanoSIMS data (0.21 ± 0.01) is a factor of 3–4 lower than those based on its SEM-EDX (0.93 ± 0.11) and TEM-EDX (0.77 ± 0.08) data. The lower ²⁶Al/²⁷Al ratio inferred by the NanoSIMS data suggests that ∼70% of the sampled Al signals by NanoSIMS were from contamination instead of being intrinsic signals. In other words, the intrinsic Al/Si_NanoSIMS ratio for the X grain should be a factor of around 3 lower, and the corrected data point falls within the 95% confidence band for the linear fit in Figure A7(a). By excluding these two grains, the respective Pearson's correlation coefficients (R) in Figure A7(a) increase from 0.85 to 0.91 for the MS grains and 0.88 to 0.95 for the X grains. Second, the NanoSIMS X grain data in Figure A7(b) are less likely to have suffered from significant Mg contamination. This is because, for instance, Hoppe et al. (2015) reported that the Mg/Si ratios of all of their studied MS grains lie below 0.003 (i.e., the maximum level of Mg contamination) based on NanoSIMS analyses with a duoplasmatron source. In comparison, all but one X grain from Liu et al. (2018b) show Mg/Si_NanoSIMS above 0.003, thereby suggesting that the Mg signals are dominated by ²⁶Mg and that the effect of Mg contamination should be negligible in skewing the correlation shown in Figure A7(b). The X grain data in Figure A7(b) show a strong linear correlation, with a correlation coefficient R value of 0.95 (gray dashed line). However, the linear fit is mainly controlled by the two grains with the highest Mg/Si ratios (highlighted by black arrows), and many grains with low Mg/Si ratios lie below the 95% confidence band for the linear fit. As a test, we excluded these two grains and obtained a new linear fit (blue dotted line; R = 0.90), which includes more grains in the lower range within the confidence band. Given that the two slopes agree with each other within the uncertainties, to be conservative, we adopted the shallower slope (2.43) for correcting the NanoSIMS Mg/Si data for presolar SiC (Table 1, Figure 2(b)). In summary, according to the MS grain data in Figure A7(a) and the X grain data in Figure A7(b), the RSF values of Al/Si and Mg/Si should be revised to 5.6 ± 1.6 (95% confidence) and 7.5 ± 2.4, respectively. The revised Mg/Si RSF value, 7.5, was adopted for calculating the Mg/Si ratios reported in Table 1 and Figure 2(b). For calculating the ²⁶Al/²⁷Al ratio, we still adopted the RSF value of 1.2 for Mg/Al based on the NanoSIMS analyses of the Burma spinel because this value agrees with the revised RSF value (1.3), which, however, has a much larger error due to the uncertainties in the linear fits in Figure A7.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Finally, note that the SEM-EDX Mg/Si and Al/Si data of grain M2-A1-G674 are both 40% lower than its corresponding TEM-EDX data. The differences for grain M2-A1-G674 appear to be caused by some SEM-EDX analytical artifact, e.g., topography, as the TEM-EDX Mg/Si data agree better with the linear fit shown in Figure A7(b). However, it is noteworthy that the TEM-EDX data for grain M2-A1-G674 show large Mg and Al heterogeneities across the TEM thin section (Singerling et al. 2021). The observed heterogeneities could be responsible for the inconsistencies between the EDX data and the linear fit in Figure A7(b) and also between the TEM-EDX and SEM-EDX data, because the SEM-EDX data represent the grain's bulk composition while the NanoSIMS and TEM-EDX analyses only sampled a thin slice of the grain.