Iron Isotopic Compositions of Troilite (FeS) Inclusions from Iron Meteorites

Ten FeS inclusions were analyzed from iron meteorites belonging to the non-magmatic group IAB and from magmatic groups IIAB, IIIAB, and IVA. These are aliquots of the same sample solutions previously analyzed for Ni isotopes (Cook et al. 2008), and the reader is referred to this earlier work for details of the sampling and digestion procedures. Three terrestrial Fe–Ni sulfides were also analyzed to verify that our chemical separation and mass spectrometry protocols are free from analytical artifacts. These samples include violarite (FeNi₂S₄), pentlandite ((Fe,Ni)₉S₈), and pyrrhotite (${\mathrm{Fe}}_{1-{\rm{x}}}$ S); the latter sample was chosen as an external standard. The terrestrial sulfides were digested in Teflon beakers at 120°C in a 2:1 mixture of concentrated HNO₃ and HCl, following the procedure of Cook et al. (2008). In addition to the sulfide samples, two aliquots of the IRMM-014 Fe isotopic standard were processed through the Fe separation chemistry.

Iron was separated from the matrix by adapting protocols reported by Dauphas et al. (2004), using a combination of anion and cation exchange chromatography in HCl media. An aliquot of the digested sample equivalent to ≈2000 μg Fe was dried in a Teflon beaker and re-dissolved in 0.2 ml 6M HCl. Sample solutions were loaded onto columns filled with 1 ml of anion exchange resin (AG1 × 8; 200–400 mesh) that had been cleaned and conditioned with 5 ml MQ, 5 ml 1M HNO₃, 10 ml MQ, and 5 ml 6M HCl. After loading, the columns were washed with 6 ml 6M HCl, and Fe was eluted in 5 ml 0.4M HCl. Iron may not be completely separated from the matrix for samples with abundant transition metal contents using HCl media and anion resin (Schoenberg & von Blankenburg 2005); these elements commonly occur in FeS nodules in iron meteorites (Buchwald 1977). In particular, the separation of Cu and Zn from Fe may be problematic for sulfide samples using only anion resin (Dauphas et al. 2009). Therefore, the Fe fraction from the anion column was further purified by loading onto a column filled with 2 ml of cation exchange resin (AG50W × 8; 200–400 mesh) that had been cleaned and conditioned with 10 ml MQ, 10 ml 3M HCl, and 10 ml 10M HCl. The Fe fraction from the anion column was dried and re-dissolved in 0.1 ml 10M HCl for loading onto the cation column. After loading, the sample was allowed to equilibrate with the resin for 2 hr. Next, the column was washed with 4 ml 10M HCl, and Fe was eluted in 4 ml 3M HCl. The cation column provides a quantitative separation of Fe from all transition metals (cf. Saito 1984). Following the second column, the solutions were evaporated, and the Fe fractions were dried once in a 2:1 mixture of concentrated HNO₃:HCl (0.45 ml) and once in concentrated HNO₃ (0.45 ml). Lastly, all samples were taken up in ≈0.42M HNO₃ for isotopic analysis. The total chemical yield for Fe was ≈100%. The full procedural blank was ≈0.040 μg Fe, which is insignificant compared to the amount of sample Fe (≈2000 μg) processed.

Isotopic measurements were made with a ThermoScientific Neptune Plus MC-ICPMS at the Institut für Geochemie und Petrologie (ETH Zürich), in medium-resolution mode, using standard sampler and skimmer Ni cones. Iron samples were introduced using the ThermoScientific Stable Introduction System (SIS), wet plasma, and an uptake rate of 100 μL/min. Iron analyses consisted of 15 measurements (20 integrations of 8.4 s) bracketed by measurements of the IRMM-014 isotopic standard. All four Fe isotopes (⁵⁴Fe, ⁵⁶Fe, ⁵⁷Fe, and ⁵⁸Fe) were measured simultaneously in static mode along with ⁵³Cr and ⁶⁰Ni, which were used to correct interferences on ⁵⁴Fe from ⁵⁴Cr and on ⁵⁸Fe from ⁵⁸Ni. The ⁵⁶Fe signal was measured using a 10¹⁰ Ohm resistor, and the ⁵³Cr and ⁶⁰Ni signals were measured using 10¹² Ohm resistors. Sample solutions of 10 ppm were analyzed, which provided beams of ≈100 V on the most abundant isotope (i.e., ⁵⁶Fe). A single analysis generally consumed ≈65 μg of Fe. A washout time of 210 s was used after the analysis of a sample or standard solution. Each analytical session was preceded by a single 300 s measurement of the electronic baseline, which was subtracted from all signal intensities. Instrumental mass bias was corrected using the exponential law with ⁵⁷Fe/⁵⁴Fe = 0.36255 (Taylor et al. 1992). Iron isotopic ratios (⁵⁶Fe/⁵⁴Fe and ⁵⁸Fe/⁵⁴Fe) are reported as the parts per 10,000 deviation relative to the bracketing standard, IRMM-014, using the following notation: εⁱFe, where i = 56 or 58. Uncertainties reported for individual samples (Tables 1 and 2) represent 95% confidence intervals based on the 15 repeat measurements of each solution. They were calculated using the following:

$$\begin{eqnarray*}&&95 \% \ \mathrm{CI}=\sqrt{\displaystyle \frac{1}{n-1}\displaystyle \sum _{k=1}^{n}{({\varepsilon }_{k}-\overline{\varepsilon })}^{2}}\times \displaystyle \frac{{t}_{0.95}}{\sqrt{n}},\end{eqnarray*}$$

where $\overline{\varepsilon }$ is the mean value of the 15 repeats and ${t}_{0.95}$ is Student's t-value for a two-sided 95% confidence interval with $n-1$ degrees of freedom (e.g., Qin et al. 2008).

Over the duration of the study, 14 aliquots of our terrestrial pyrrhotite were processed through the full chemical procedure described in Section 2.2. The isotopic measurements of these replicates define an external reproducibility (2SD) for ε⁵⁶Fe and ε⁵⁸Fe of ±0.05 and ±0.19, respectively (Table 1). This level of analytical resolution represents an improvement compared to previous studies (e.g., Dauphas et al. 2004; Tang & Dauphas 2012). The εⁱFe values measured on the two aliquots of IRMM-014 processed through ion exchange chemistry show no resolvable effects, demonstrating that the analytical procedure is free from artifacts. After chemical separation of Fe, all terrestrial sulfides had Ni/Fe ≤ 2.2 × 10⁻⁵ and Cr/Fe ≤ 6.1 × 10⁻⁷. These ratios are far below the thresholds required to make accurate measurements of Fe isotopes in the presence of isobaric interferences from Cr and Ni (Dauphas et al. 2009). The above data demonstrate the effectiveness of our chemical separation protocol and the accuracy of the overall method.

All FeS inclusions (Table 2) display ε⁵⁸Fe values that are identical to the terrestrial isotopic standard within the overall analytical resolution of ±0.19. In contrast, six of the inclusions show small but clearly resolved deficits in ε⁵⁶Fe (i.e., their ε⁵⁶Fe values differ from the isotopic standard by > ± 0.05). These deficits occur in each of the iron meteorite groups sampled. Like the terrestrial samples, the chemically purified FeS samples have insignificant levels of isobaric interferences (Ni/Fe ≤ 1.4 × 10⁻⁵ and Cr/Fe ≤ 6.9 × 10⁻⁷), which cannot account for the observed ε⁵⁶Fe deficits. We note that it is not currently possible to conclusively identify which Fe isotope is anomalous. One possibility is that the ε⁵⁶Fe values for FeS samples reflect a true variation in ⁵⁶Fe. However, an excess of either ⁵⁴Fe or ⁵⁷Fe, used to correct the instrumental mass bias, can also lead to negative ε⁵⁶Fe values. Thus, we restrict our discussion to ε⁵⁶Fe values internally normalized to ⁵⁷Fe/⁵⁴Fe and do not discuss absolute ⁵⁶Fe abundances.

Before discussing the results for the Fe isotopes in detail, we first present a re-evaluation of the stellar component responsible for the observed nucleosynthetic Ni isotopic variations (Cook et al. 2008) in the FeS inclusions. These results are then used to explore the potential causes for the observed Fe isotopic variations. Cook et al. (2008) presented trends for isotopic mixtures between terrestrial Ni and that produced in three stellar environments, as well as for potential spallation effects on FeS from GCRs. In the current study, we take a similar approach. In order to constrain the possible stellar source that produced the component responsible for the observed εⁱNi variations, isotopic mixtures were calculated between terrestrial Ni (Gramlich et al. 1989) and the yields from various nucleosynthetic stellar models, including type Ia supernovae (SN Ia) (Iwamoto et al. 1999; Travaglio et al. 2004; Maeda et al. 2010; Travaglio et al. 2011; Seitenzahl et al. 2013), type Ib,c supernovae (SN Ib,c) (Woosley et al. 1995), a core collapse type II supernova (SN II) for a 25M_⊙ progenitor (Rauscher et al. 2002), and asymptotic giant branch (AGB) stars (Domínguez et al. 2011; Pignatari et al. 2014). For the SN II mixtures, the yields from each of the eight individual nuclear burning zones within the pre-supernova star, as well as the yields for the bulk ejecta from the explosion (Rauscher et al. 2002), were considered. In all, 65 stellar models were considered, and mixing lines between each stellar source and terrestrial Ni were calculated.

Figure 1 shows selected trends expected for mixtures between terrestrial Ni and individual stellar models. Eight mixtures provide isotopic compositions that are within uncertainty of the most extreme observed variations (i.e., Mundrabilla). For visual clarity, only four of these are shown: SN Ia, SN Ib,c, SN II (O/Ne zone), and AGB (5M_⊙ Z = 0.02). These represent the mixtures that provide the closest match from each of the four types of stellar environments (SN Ia, SN Ib,c, SN II, and AGB). Three additional SN Ib,c mixtures and one AGB (5M_⊙ Z = 0.01) mixture give similar results to those shown in Figure 1. These trends confirm that the Ni isotopic variations in FeS are consistent with a nucleosynthetic origin and that a variety of stellar environments could represent the source composition needed to explain the observations.

Zoom In Zoom Out Reset image size

Previous studies of iron isotopes from a wide variety of meteorite types (Dauphas et al. 2004, 2008; Tang & Dauphas 2012) did not observe any resolvable variations and concluded that Fe isotopes were homogeneously distributed in the protoplanetary disk. These previous studies placed a particular emphasis on the homogeneity of ε⁵⁸Fe because the low abundance of ⁵⁸Fe (0.28%; Taylor et al. 1992) makes it a sensitive proxy for mixing between sources with disparate Fe isotopic compositions; in general, our calculations reveal that mixtures containing a small (≈0.0001%) amount of stellar Fe from any of the 65 sources considered would have resolvable ε⁵⁸Fe values relative to pure terrestrial Fe. The lack of resolved ε⁵⁸Fe variations in FeS is consistent with previous studies of bulk meteorites, and this ratio by itself supports a homogeneous distribution of Fe isotopes within the early nebula.

In contrast to ε⁵⁸Fe, six FeS inclusions (Table 2) show small but resolvable deficits in ε⁵⁶Fe. These are the first reported non-mass-dependent Fe isotopic variations for samples other than CAIs (Völkening & Papanastassiou 1989). Unlike Ni isotopes in the same samples, which show excesses and deficits in ε⁶⁴Ni, only deficits are observed in ε⁵⁶Fe (Figure 2). This raises the question of whether additional constraints on the possible stellar source of the Ni isotopic variations could be provided by comparing the measured Fe isotopic compositions to collateral effects on Fe isotopes predicted by the mixtures that match the Ni isotopic data. For this comparison, the most anomalous Ni isotopic composition (i.e., Mundrabilla) was chosen. First, mixtures were calculated that reproduced the magnitude of the observed ε⁶⁴Ni deficit in Mundrabilla. Next, the expected collateral effects on Fe isotopes were calculated. A dilution factor was included that accounted both for the Fe/Ni ratio of the nucleosynthetic material for the relevant stellar model and the Fe/Ni ratio for average solar system material (Lodders 2003). The predicted collateral effects on Fe isotopes in Mundrabilla FeS are large (Figure 3) and would be well-resolved from the terrestrial reference standard with our current analytical precision; they are clearly inconsistent with the measured values. Similarly, matching the Fe isotopic composition of FeS from Gibeon to mixtures of stellar and terrestrial Fe predicts collateral effects on Ni isotopes that do not equal those observed. The absence of the predicted effects, in combination with the lack of correlation (r² = 0.09) between ε⁵⁶Fe and ε⁶⁴Ni (Figure 2), indicate that the Ni and Fe isotopic anomalies are decoupled. Thus, the Fe isotopes do not help to constrain the possible stellar source responsible for the nucleosynthetic Ni variations.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Decoupling of nucleosynthetic variations have been reported in several studies. For example, Wang et al. (2011) reported an absence of Fe isotopic variations in acid leachates prepared from whole rock chondrites that contained large nucleosynthetic Cr isotopic variations. Similarly, Fehr et al. (2006) did not observe Te nucleosynthetic isotopic anomalies in bulk carbonaceous chondrite leachates for which large variations in Zr isotopes were reported (Schönbächler et al. 2005). In addition, Walker (2012) also showed that nucleosynthetic variations are absent in Os isotopes in iron meteorites that are anomalous for isotopes of other siderophile elements like Ru (Chen et al. 2010) and Mo (Burkhardt et al. 2011). One possible explanation for the lack of correlation between nucleosynthetic effects on Ni and Fe isotopes is that Ni and Fe resided in different presolar carriers. Condensation calculations for products from a SN II (Fedkin et al. 2010) showed that the redox environment can affect the nature of the host phase(s) for Fe and Ni. In oxidizing environments (e.g., O/Ne zone), Ni is predicted to be hosted in Fe-poor metal, whereas Fe will primarily condense into silicates and oxides. Thus, fractionation of Fe from Ni may occur during the condensation of some stellar products and could provide a mechanism to separate these two elements into different carrier phases. Subsequent incorporation of varying amounts of each carrier (i.e., metal versus silicate) into the iron meteorite parent bodies could lead to a decoupling of the nucleosynthetic signal. Alternatively, thermal processing and preferential destruction of certain carriers within the protoplanetary disk could have led to reservoirs with different isotopic compositions for one element but not another (e.g., Trinquier et al. 2009; Larsen et al. 2011; Walker 2012; Akram et al. 2015).

An alternative possibility is that the ε⁵⁶Fe deficits do not represent isotopic variability in the nebula, but rather are the result of processes acting on the meteorite parent bodies, which began with homogeneous Fe isotopic compositions. Such an explanation may account for the seemingly contradictory occurrence of ε⁵⁶Fe deficits accompanied by homogeneous ε⁵⁸Fe values. For example, while in outer space, meteoroids are bombarded by high-energy GCRs. This interaction can induce nuclear reactions within the meteoroid, leading to changes in its isotopic composition; such effects are well-documented in meteorites (e.g., Masarik & Reedy 1994; Masarik 1997; Leya et al. 2003; Leya & Masarik 2013). Primary spallation reactions by GCR lead to the production of secondary neutrons; these neutrons may then be captured by some nuclei if they are slowed down sufficiently by collisions as they travel through the target. The modification of isotopic compositions due to neutron capture reactions have been reported in iron meteorites for multiple elements, including Ag (Matthes et al. 2015), Ru (Fischer-Gödde et al. 2015), Os (e.g., Walker 2012; Wittig et al. 2013), Pt (e.g., Kruijer et al. 2013), and W isotopes (e.g., Markowski et al. 2006; Kruijer et al. 2013). Furthermore, GCR effects are not uniform because their magnitude changes as a function of depth below the surface (e.g., Masarik 1997; Leya & Masarik 2013). Thus, this depth dependency can lead to isotopic differences among irons within a particular group and also between different pieces of the same meteorite (e.g., Walker 2012; Kruijer et al. 2013; Qin et al. 2015).

To assess potential neutron capture effects on Fe isotopes, the neutron capture resonance integrals for Fe isotopes from Mughabghab (2006) were used to calculate the possible influence of GCRs on ε⁵⁶Fe and ε⁵⁸Fe. Neutron capture effects shift ε⁵⁶Fe and ε⁵⁸Fe to lower values (Figure 4). The magnitude of the effect on ε⁵⁸Fe is larger than on ε⁵⁶Fe (ε⁵⁸Fe/ε⁵⁶Fe = 1.68). However, for the small ε⁵⁶Fe deficits observed (≥−0.12), the corresponding effects on ε⁵⁸Fe are nearly equal to or less than our analytical resolution (±0.19), and the measured ε⁵⁸Fe values agree with this expectation. A GCR origin for the ε⁵⁶Fe deficits can also explain the variability among members of the same group (e.g., IIIAB), which experienced different degrees of irradiation due to origins at varying depths within the parent body. Overall, the predicted and measured values are consistent, which supports the potential modification of an originally homogeneous Fe isotopic composition due to GCR effects. Future studies that combine measurements of Fe isotopes with those of Ti and/or Cr may help to identify GCR effects in FeS inclusions from iron meteorites. Neutron capture effects have been shown to occur on Ti isotopes for lunar and meteoritic samples (Zhang et al. 2012), and significant GCR effects on Cr isotopes are predicted for samples with long exposure ages and elevated Fe/Cr ratios (Leya et al. 2003).

Zoom In Zoom Out Reset image size

Fujii et al. (2006a) suggested that effects due to the NFS could provide an alternative explanation for isotopic variations in CAIs that had initially been attributed to a nucleosynthetic origin. Hence, it is useful to consider NFS effects when non-mass-dependent isotopic variations are observed. The NFS effect on isotope fractionation was described by Bigeleisen (1996) and Schauble (2007). The NFS is expressed during chemical exchange reactions, and it causes isotopic variations that deviate from a mass-dependent fractionation trend. These effects arise from differences in the sizes and shapes of nuclei among isotopes of a given element, which are in turn due to the different numbers of neutrons within the nucleus. In general, NFS effects are largest for isotopes of heavy elements (e.g., U), and they decrease with atomic mass. The effects are predicted to be small for light elements (A < ≈ 100). Because NFS effects are proportional to the nuclear charge radii, they do not follow a linear dependence on atomic mass. This nonlinear behavior is evident in diagrams of nuclear charge radii versus atomic mass (e.g., Aufmuth et al. 1987; Fujii et al. 2009) and is known as odd–even staggering. Figure 5 shows such a diagram for Fe isotopes. The nuclide ⁵⁷Fe, the only Fe isotope with an odd number of neutrons, plots slightly off the linear trend between its even-number neutron neighbors, which suggests it may be prone to NFS effects.

Zoom In Zoom Out Reset image size

Nuclear fields shift effects are particularly important during chemical reactions that involve a change in redox state. For redox reactions that involve the transfer of s-orbital electrons, the heavy isotopes will be preferentially concentrated in the more oxidized phase (Bigeleisen 1996; Schauble 2007). Qualitatively, NFS effects predict that the heavy Fe isotopes should show an enhanced preference for the oxidized phase (FeS) over the reduced phase (metal) in iron meteorites because the redox reaction (Fe⁰ to Fe²⁺) involves a change in s-orbital electrons. A slight excess of ⁵⁷Fe in the FeS would lead to apparent deficits in ε⁵⁶Fe and ε⁵⁸Fe because ⁵⁷Fe is used in the ratio to correct the instrumental mass bias. Experimental work by Fujii et al. (2006b) did not find conclusive evidence for NFS effects on Fe isotopes, but some results show slight departures from the trend expected for a purely mass-dependent fractionation. Additionally, theoretical predictions for the maximum NFS effects in Fe-group elements of around 0.2–0.3ε per amu (Moynier et al. 2013) are comparable to the observed ε⁵⁶Fe deficits in FeS.

Values for nuclear charge radii (Angeli 2004) were used in conjunction with Equation (1) from Fujii et al. (2006a) to calculate NFS effects on Fe isotopic ratios. The expected trend (ε⁵⁸Fe/ε⁵⁶Fe = 0.97) is distinct from that for neutron capture effects (Figure 4), but like those for neutron capture effects, the predicted shifts in ε⁵⁸Fe are too small to resolve. The NFS trend provides a slightly better fit to the data, but a lack of correlated effects (r² = 0.08) in ε⁵⁶Fe and ε⁵⁸Fe preclude differentiating between the two trends. The FeS inclusion from Gibeon displays the largest ε⁵⁶Fe deficit. Comparing the effects from the NFS to GCR-induced effects (Figure 6) shows a somewhat better overall match between the NFS and the measured values. However, the effects from neutron capture reactions are nearly at the limit of our ability to resolve a difference in ε⁵⁸Fe from the terrestrial standard, so GCR effects cannot be ruled out.

Zoom In Zoom Out Reset image size