Formation of Wüstite
Wüstite has been reported in the inner parts or closely related to several meteorite fusion crusts (e.g., El Goresy and Fechtig 1967; Ramdohr 1967; Genge and Grady 1999; Kim et al. 2009). However, in all these instances, it is neither the only nor the dominant Fe-oxide. Rather magnetite is most prevalent. Thus, the dominance of wüstite needs further consideration. Generally, troilite inclusions within iron meteorites are preferentially destroyed during atmospheric entry as troilite is melted more rapidly than the main mass comprised of metal and/or silicates (Ramdohr 1967). Furthermore, iron meteorites are known to possess broader fusion crusts than stony meteorites because of the high heat conductivity of the metal phase (El Goresy and Fechtig 1967; Ramdohr 1967 and references therein). In this respect, the survival of MS-166 sulfide-metal assemblage is remarkable. However, as will be shown, considerable amounts of predominantly troilite were lost—either ablated or they reacted to FeO.
The main portions of MS-166 are free of any wüstite (Fig. 2a). Consequently, wüstite must be of secondary origin. The presence of wüstite in fusion crusts of iron meteorites is commonly explained by oxidation of metallic Fe along with fusion during atmospheric passage. Oxygen is supplied by the Earth's atmosphere. It is reasonable to assume that material similar to the nonfused, FeO-free portion of MS-166 was the precursor of the wüstite-bearing fusion crust. Comparing the modal abundances of metal and sulfide in the fusion crust (Fig. 2b) and the main rock (neglecting the minor ∼5 vol% of silicate in the main mass; Fig. 2a), considerable amounts of FeS (and probably also minor metal) were lost during fusion. FeS was determined to constitute ∼65 vol% and metal ∼30 vol% of the MS-166 main mass below the fusion crust. However, these proportions have inevitably changed (cf. Fig. 2b) when looking at the fusion crust where the metal grains are the dominant constituent and the wüstite-troilite material is of lower modal abundance (∼75 vol% metal, ∼25 vol% FeO-bearing matte). Troilite in the fusion crust makes up only some 10 vol% (Fig. 2b). Metal grains in the fusion crust and those in the unaffected interior of MS-166 are similar in size, shape, and chemical composition (both cores and Ni-rich rims) with the only significant difference that the Ni-rich portions in the fusion crust indicate onset of melt resorption. From this textural and chemical point of view, it is unlikely that all the metal was completely melted as observed in other meteorite fusion crusts (e.g., El Goresy and Fechtig 1967; Genge and Grady 1999). Thus, it seems plausible that wüstite formed in large part from troilite and that considerable amounts of FeS on the order of more than 50% were lost. This ablated material could have given rise to S-rich (and probably even some metal-rich) meteorite ablation spherules (Genge and Grady 1999a).
Formation of wüstite from FeS requires the replacement of S by O as iron in both wüstite and troilite is ferrous. Nevertheless, two possible reactions of importance for MS-166 wüstite formation have to be considered that were given in Asaki et al. (1974) who reported on the oxidation of molten ferrous sulfide:
As the metal cores appear texturally mainly unmolten, reaction (1) was likely the dominant process taking place in MS-166. However, the outer Ni-rich portions of the metal grains indicate (partial) melting. Thus, reaction (2) might also have contributed to FeO although in minor proportion than FeS. The precipitation of wüstite from an FeS melt undergoing oxidation is furthermore suggested by the eutectic texture of the fusion crust matte. At the surface of meteorite fusion crusts, temperatures of some 1500 °C are likely. Partial melting with generation of an S-rich metallic melt would start at ∼988 °C taking the Fe-FeS system as a basis or would start some 50 °C lower in the Fe-Ni-S system (Villars et al. 1995). 65 vol% FeS corresponds to ∼20 wt% S in MS-166. Transferring this to the Fe-S binary would lead to complete melting of MS-166 at around 1300 °C. Taking the amount of Ni into consideration, the temperature might be somewhat lower. Ni-rich metal (taenite) has a lower melting temperature than low-Ni metal (kamacite). Consequently, Ni-rich metal resorption by the metallic melt starts at lower T, which is in textural accordance with what is observed in the MS-166 fusion crust. Taking the Fe-Ni binary system (Schwartzendruber et al. 1991) and the core (7–9 wt% Ni) and rim (∼33 wt% Ni) compositions of MS-166 metal grains, the corresponding liquidus temperatures are about 1500–1510 °C and 1460 °C, respectively. This suggests that the temperatures prevailing at the MS-166 surface during fusion were somewhat lower, probably on the order of the above-mentioned ∼1300 °C, because the Ni-rich portions show an onset of resorption by the melt and because one is facing a more complex multicomponent system. However, if such elevated temperatures occurred deeper inside the MS-166 rock, fusion would have completely melted the rock. Nevertheless, apparently troilite and only minor metal were effectively molten during atmospheric passage with the majority of FeS material being ablated, some being transformed into FeO, and minor amounts being preserved. Temperatures were either not high enough to cause complete metal melting or the duration of the heating event was too short. The conclusion that the metal cores were largely unmolten (only the outer Ni-rich portions indicate some mobilization) is supported by the siderophile element systematics of the low-Ni cores.
Another argument that predominantly troilite was oxidized to FeO comes from observations made during oxidation experiments of the Gibeon IVA iron meteorite by Visscher and Lodders 2002 and references therein). They found, as other authors before, that Ni in meteoritic metal obviously suppresses wüstite formation, but favors magnetite-spinel formation instead. Nickel is strongly enriched in the outer portions of the metal grains. However, no complete oxidation of FeO into magnetite occurred as observed in the experiments by Asaki et al. (1974). Nevertheless, EELS spectra of wüstite (Fig. 4) in the fusion crust argue for the presence of some Fe3+ suggesting at least some oxidation of Fe2+ to Fe3+. Darken and Gurry (1945) found a correlation between the Fe3+ abundance, temperature, and the composition of the surrounding gas in equilibrium with wüstite. For an Fe3+ content on the order of 18% as inferred from the EELS spectra, this leads to a temperature in the range of ∼1200 °C (assuming a CO2/CO of unity). However, this is rather a rough estimate as experimental conditions and procedures differ from what is thought to have taken place during formation of wüstite from FeS in MS-166. The tiny chromite grains found inside wüstite might have formed along with the reaction of troilite to wüstite during fusion as troilite in the main mass was found with minor Cr. However, the abundances were quite low (up to ∼0.3 wt% Cr). Similarly, P might have been oxidized and formed the observed Fe-phosphate. The source of P was likely mainly melted Fe,Ni metal from the outer Ni-rich portions from which it was dissolved into and then oxidized in the matte precipitating as the Fe-phosphate upon cooling. LA-ICP-MS analyses on unmelted metal grains gave some 4000 ppm P in the low-Ni cores and can reach even higher values of ∼7000 ppm P (Table 2). Trace element analyses on the large FeO-free sulfide masses give P abundances below 100 ppm (unpublished data).
The depth to which temperature was effective in altering the MS-166 meteorite is given by the low-oxygen portions (Fig. 2c). They appear in direct contact with the FeO-free main mass of MS-166 and represent a transition zone in which the effect of fusion and oxidation was less pronounced. Decreasing effects of fO2 with increasing depth into the meteorite are a typical feature of meteorite fusion crusts (e.g., El Goresy and Fechtig 1967; Genge and Grady 1999). Roundish troilite with tiny droplets of FeO might point to incipient melting or at least recrystallization under elevated temperatures with some oxidation. Partial melting would be supported by Ni-rich nuggets surrounding the metal grains in these areas. The relatively sharp contact of the altered with the unaffected material clearly marks the depth to which heat effectively induced mineralogical changes and is about ∼1.5 mm in the portions where the fragile fusion crust was preserved around the cut through the main mass.
The well-ordered state of wüstite close to the P″-type superstructure suggests that cooling took place within a somewhat prolonged time interval. Anderson and Sletnes (1977) report that at 225 °C the P′ structure transforms completely within 30 min to the ordered P′′ structure. As our TEM observations indicate a structure close to P′′ with some residual disorder, it appears that temperatures in the range between ∼200 and 320 °C (below which the ordering phenomena occur) prevailed for some minutes after peak heating. The residual heat might be plausibly explained by the high metal abundance in MS-166 and the correspondingly high heat conductivity, which allowed large volumes of the meteorite to heat up (only partially melting the outer portions) and transfer heat back to the fused, outer portions during dark flight.
The modal abundance of wüstite in the MS-166 fusion crust is remarkable. Although Ramdohr (1967) reported wüstite being prevalent in Fe-rich chondrites, iron meteorites, and mesosiderites, it is commonly found as very small rather rare grains and often intergrown with magnetite. The same author concluded that wüstite can only form and can only be preserved when the supply of FeO increases rapidly, something that was the case with MS-166 via oxidation of large masses of predominantly FeS leaving behind a highly porous framework of metal grains interconnected with Fe-S-O matte.
Partitioning of Siderophile Elements
The unaltered, i.e., nonfused portions of MS-166 (Fig. 2a) were proposed to have formed from a fractionally crystallizing S-rich metallic melt (Horstmann et al. 2011) related to ureilite petrogenesis (Horstmann et al. 2012b). Although the analyzed metal grains are from the fusion crust, they all fall onto the same trend observed for MS-158 metal (Horstmann et al. 2011). Metal from both samples spans nearly three orders of magnitude variation in Ir abundance similar to IIIAB iron meteorites. A negative correlation between Ir and Au is suggestive of a fractional crystallization origin of the metal (Horstmann et al. 2011) with the surrounding troilite-dominated material representing the remaining S-rich liquid. The preserved Ni- and trace element-zoning argues for fast cooling of the material. However, it should be noted that recent, more detailed examinations of MS-166 metal indicate a more complex formation history than only simple fractional crystallization, but this will be the issue of an upcoming article.
Chabot and Humayun (2011) reported on experimental investigations that had the goal of deciphering the influence of oxygen on siderophile trace elemental partitioning in metallic liquids. Oxygen is of considerable importance in planetary differentiation as it is a potential light element in metallic melts probably even in the Earth's core (e.g., Hillgren et al. 2000). Chabot and Humayun (2011) found that elements with affinities to form stable oxides like W and Ga preferentially partition into the oxygen-bearing liquid rather than into the crystallizing metal. This behavior is different from S-rich metallic liquids in which the compatibility in metal of both elements increases with increasing S-content of the melt (e.g., Chabot and Jones 2003; Chabot et al. 2003, 2009). The presence of abundant oxygen preserved as wüstite allows insights into the influence of oxygen on partitioning in a natural sample. Horstmann et al. (2012b) first reported on the presence of wüstite in Almahata Sitta, but reinspection of the entire MS-166 specimen (Horstmann et al. 2012c) revealed that the fragments studied by Horstmann et al. (2012b) sampled mainly fusion crust material, and were thus unrepresentative of the main mass. Therefore, the previous conclusions drawn with respect to ureilite petrogenesis were incorrect. Nevertheless, trace element compositions of metal and surrounding matte could indicate some influence of oxygen on siderophile trace element partitioning. Previously, the effect of oxidation on some elemental abundances has been reported in I-Type cosmic spherules and meteorite fusion crusts (Nozaki et al. 1999 and references therein; Genge and Grady 1999). Here, the first LA-ICP-MS analyses of siderophile trace element partitioning influenced by oxygen in a metallic melt will be presented.
Calculated (apparent) distribution coefficients (D values) from analyses on metal and matte (Fig. 7) are generally comparable to the experimental results of Chabot and Humayun (2011). It has to be pointed out here that the largely unmelted metal cores were measured and used for calculation of the D values, which were likely not in equilibrium with the surrounding matte. This is allowed for by adding “apparent” to the D values. The best indicator for the influence of oxygen on elemental partitioning in metallic melts is, as already mentioned, the behavior of W and Ga (Chabot and Humayun 2011), which show decreasing concentrations in the solid metal when oxygen is present in the liquid. As shown in Fig. 7, measurements of the O-rich portion of MS-166, which has ∼7 wt% O, show a lower D value for W than the O-poor (∼2.5 wt% O) MS-166 measurement. This difference is consistent with the elemental partitioning signature expected due to O in a metallic melt. The presence of O in the metallic melt would also impart a distinctive signature on Ga, and the D(Ga) value for the O-rich measurements of MS-166 is lower than the O-poor ones, consistent with the behavior due to the presence of O in a metallic melt.
Figure 7. Comparison of partition coefficients from the Chabot and Humayun (2011) Fe-S-O experiments (black and gray dashed lines correspond to ∼10 and ∼12 wt% O, respectively, each with ∼20 wt% S; dotted line corresponds to ∼3 wt% O, ∼20 wt% S) and data obtained from MS-166 (black solid lines correspond to ∼2.5 wt% O, 30 wt% S, open lines indicate ∼7 wt% O, 21 wt% S). See text for details.
Download figure to PowerPoint
Both the S and O contents of a metallic melt will influence the partitioning behavior, and the experimental results presented in Fig. 7 have slightly different S and O contents than measured in the MS-166 analyses. The solid metal-liquid metal partitioning behavior in the Fe-(Ni)-S system has been parameterized (Chabot and Jones 2003) such that one can apply it to the specific S contents measured in the MS-166 analyses. Using the inferred bulk chemical abundances of S for the high-O and low-O portions (∼21 wt% S and ∼30 wt% S, respectively) surrounding the metal grains analyzed, the apparent partition coefficients between solid metal and liquid metal for the given compositions in the Fe-S system were calculated (Fig. 8). As mentioned above, the best indicator for the influence of oxygen on elemental partitioning in metallic melts is the behavior of W and Ga (Chabot and Humayun 2011). Figure 8 shows that Ga and W, which are expected to show deviating behavior from the modeled D values in the pure Fe-S system corresponding to the shaded area in Fig. 8, do not exhibit clear anomalies. For the low-oxygen portion, it can be concluded that the partitioning is consistent with partitioning in the pure Fe-S system without any obvious indication for influence of oxygen. In the case of the oxygen-rich portions, however, one can see a deviating behavior for both W and Ga relative to Ge and Rh. Both elements show lower apparent D values by about an order of magnitude than in the low-oxygen portion (Fig. 8a). This is also clearly depicted in Fig. 8b, in which the ratio of the apparent D values for the high-O and low-O portion is shown. Gallium in the wüstite-rich portion shows a lower D value than in the low-oxygen domain, but apparent D values for Ga and Ge are much higher than expected from the modeled values using the estimated S abundances for the respectively analyzed portions of MS-166 fusion crust and plot above the shaded area in Fig. 8a, while the Ga value for the high-O portion ends up in the shaded area. Consequently, there is trace elemental evidence for the influence of oxygen on elemental partitioning in the MS-166 fusion crust, although conditions in the fusion crust were certainly nonequilibrium. However, the observed textures in the large metal grains (Fig. 3a) are seemingly not in accordance with complete melting of the fusion crust metal, but rather imply partial dissolution of the Ni-rich rim metal leaving the low-Ni cores largely unaffected, with obvious implications for apparent partition coefficients. As noted earlier, the original texture of the fusion crust material was likely similar to the unaltered main mass of MS-166. Thus, the Ni-rich nuggets surrounding the metal cores might indicate early stages of metal resorption by the S,O-rich melt. LA-ICP-MS measurements of the tiny Ni-rich nuggets surrounding the low-Ni cores were precluded by the small grain size. Thus, no conclusion based on data from these small nuggets can be drawn as to whether the outer metal portions experienced elemental equilibrium partitioning in an oxygen-bearing liquid. Nevertheless, the apparent D values (Fig. 8) indicate that some elemental redistribution took place. Minor amounts (compared with the low-Ni cores) of W and Ga were present in the outer portions of the metal grains that were (partially) resorbed by the Fe-S-O liquid during fusion crust formation, and thus W and Ga could have preferentially partitioned into the matte upon solidification accounting for the deviant behavior of both W and Ga (Fig. 8). Such a scenario would not require complete melting of the metal grains. In contrast with Fe-sulfide, it appears that the metal cores originally present in the outer portions of MS-166 largely survived the fusion process.
Figure 8. a) Compilation of modeled and measured (or apparent) Dsolid metal/liquid metal values for the FeO-rich and low-FeO portions of MS-166. The D values were modeled for partitioning in the Fe-S system with liquid S-contents that ranged from 21 to 30 wt%, as measured in the matte of MS-166 (shaded area). b) Ratio of apparent D values for the FeO-rich and low-FeO portion illustrating the deviant behavior of W and Ga.
Download figure to PowerPoint