Structural clues to the origin of refractory metal alloys as condensates of the solar nebula


  • Dennis HARRIES,

    1. Bayerisches Geoinstitut, Universität Bayreuth, D-95440 Bayreuth, Germany
    2. Institut für Geowissenschaften, Friedrich-Schiller-Universität Jena, Carl-Zeiss-Promenade 10, D-07745 Jena, Germany
    Search for more papers by this author
  • Thomas BERG,

    1. Max-Planck-Institut für Chemie, Johannes-Joachim-Becher-Weg 27, D-55128 Mainz, Germany
    2. Institut für Physik, Johannes Gutenberg-Universität, Staudingerweg 7, D-55128 Mainz, Germany
    Search for more papers by this author

    1. Institut für Geowissenschaften, Friedrich-Schiller-Universität Jena, Carl-Zeiss-Promenade 10, D-07745 Jena, Germany
    Search for more papers by this author
  • Herbert PALME

    1. Forschungsinstitut und Naturmuseum Senckenberg, Senckenberganlage 25, D-60325 Frankfurt am Main, Germany
    Search for more papers by this author

Corresponding author. E-mail:


Abstract– Alloys of the refractory metals Re, Os, W, Ir, Ru, Mo, Pt, and Rh with small amounts of Fe and Ni are predicted to be one of the very first high-temperature condensates in a cooling gas of solar composition. Recently, such alloy grains were found in acid-resistant residues of the Murchison CM2 chondrite. We used focused ion beam (FIB) preparation to obtain electron-transparent sections of 15 submicrometer-sized refractory metal nuggets (RMNs) from the original Murchison residue. We studied their crystallography, microstructures, and internal compositional variations using transmission electron microscopy (TEM). Our results show that all RMNs studied have hexagonal close-packed (hcp) crystal structures despite considerable variations of their bulk compositions. Crystallographic superstructures or signs of spinodal decomposition are absent and defect microstructures are scarce. Internally, RMNs are compositionally homogeneous, with no evidence for zoning patterns or heterogeneities due to exsolution. Many RMNs show well-defined euhedral crystal shapes and all are nearly perfect single crystal. Our findings are consistent with a direct (near-) equilibrium condensation of refractory metals into a single alloy at high temperature in the solar nebula as predicted by current condensation models. We suggest that this alloy is generally hcp structured due to an extended ε-phase field in the relevant multicomponent alloy system. The high degree of structural perfection and compositional homogeneity is attributed to high defect energies, high formation temperatures, slow cooling rates, small grain sizes, and rapid internal diffusion.


Submicrometer-sized refractory metal nuggets (RMNs) recently described by Berg et al. (2009) have been demonstrated to match compositions predicted for nebular condensates. The nuggets were recovered from acid-resistant residues of the Murchison CM2 chondrite and may have originally resided in Ca-Al-rich inclusions (CAIs). The particles contain high but variable concentrations of the refractory metals Os, Ir, Mo, and Ru, with subordinate amounts of Pt, W, Rh, Ni, and Fe. The presence of major amounts of Mo (up to 40 wt %) and some W in the alloy distinguishes them from the majority of naturally occurring, terrestrial noble metal alloys. RMNs contain metals that are structurally diverse in their pure forms (see below) and only share low vapor pressures as their common feature. Metals with vapor pressures above that of metallic Fe (e.g., Au and Pd) are absent. Comparison of the compositions of the nuggets with those expected from thermodynamic condensation calculations suggests a formation of these particles by condensation in a nebular environment within a temperature range from 1600 to 1400 K at an assumed pressure of 10−4 bar (Palme and Wlotzka 1976; Campbell et al. 2001, 2003; Berg et al. 2009). The presence of the expected amounts of Mo and W in the alloy excludes major oxidation or sulfidation (forming molybdenite and/or scheelite; Blum et al. 1989) during their residence in the solar nebula, in the chondritic parent body, or during retrieval. Although the original hosting of the nuggets remains unclear, the recent discovery of similar nuggets in CAIs of Allende (Schwander et al. 2012) suggests that the RMNs studied here were once contained in CAIs of the Murchison meteorite.

The morphology of the particles suggested mixing of all metals in a single alloy despite differences in the crystal structures of the pure individual metals, which are hexagonal close-packed (hcp) for Os and Ru; cubic close-packed/face-centered cubic (fcc) for Ir, Ni, Pt, Rh; and body-centered cubic (bcc) for Mo, W, and Fe (also fcc). Sylvester et al. (1990) predicted condensation of refractory metal alloys into three different crystal structures: hcp alloys of Ru and Os; bcc alloys of Mo and W; and fcc alloys of Ir, Fe, and Ni. Later mixing produced the observed opaque assemblages (OAs) or Fremdlinge in CAIs. Palme et al. (1994) favored condensation into a single refractory metal alloy. This prediction was based on the observation of micrometer-sized OAs, which contain oxidized, sulfidized, and exsolved phases. In these assemblages, Mo occurs as molybdenite, W as scheelite, Os and Ru within hcp alloys, and Pt and Ir within Fe-Ni metal. In the view of Palme et al. (1994), OAs originated from processing of originally homogeneous refractory metal within CAIs under high fS2 and fO2 conditions, either by thermal events in a nebular environment or through fluid-assisted metasomatic processes within the chondritic parent bodies (Blum et al. 1989; Palme et al. 1994). As sulfide- and oxide-rich OAs appear to occur predominantly in CAIs of the oxidized CV subgroup (e.g., Allende; El Goresy et al. 1978; MacPherson 2003), the parent-body scenario seems to be the more likely. The population of the submicrometer-sized particles extracted from Murchison and studied by Berg et al. (2009) apparently escaped oxidation and sulfidation and presumably retained their original structures and compositions, probably due to low alteration temperatures in the CM parent body and/or good shielding from reactive fluids.

Here, we report the results of a transmission electron microscopy (TEM) study including quantitative chemical analyses of 15 submicrometer-sized particles from the original sample of Berg et al. (2009). Our study aims at unraveling the structure and microchemistry of RMNs and obtaining clues to their origin. For this purpose, we address the following questions: What are the crystal structures of the RMNs? Are they single alloys or mixtures of several different alloys? Are the nuggets compositionally uniform or are they zoned? Is the structure of these grains similar to what can be expected for nebular condensates?

Samples and Methods

We used the focused ion beam (FIB) technique to sample 15 compositionally diverse RMNs from the SiC-rich, acid-resistant residue previously studied by Berg et al. (2009). The sample is an aliquot obtained during the preparation process (cf. Amari et al. 1994) and corresponds to an initial amount of several grams of the Murchison chondrite. We specifically focused on including chemically diverse grains into our selection for TEM preparation to search for structural variations. Subsequent TEM-EDX analysis and fitting to the condensation model of Berg et al. (2009) (based on equilibrium condensation calculations at 10−4 bar described by Palme and Wlotzka [1976] and Campbell et al. [2001]) showed that the selected grains indeed span the complete range of previously reported condensation temperatures from 1400 to 1600 K (Fig. 1; Table 1).

Figure 1.

 Calculated (curves) and measured/fitted (points with bars) temperature-composition relationships for condensation of the RMNs investigated in this study (TEM-EDX). The model condensation temperatures for RMNs were obtained using the method of Berg et al. (2009) assuming a solar gas at Ptot = 10−4 bar; their data points from the initial study of the RMN population are shown as gray dots (SEM-EDX). Error bars bracket the minimum and maximum values for different EDX spectral deconvolution techniques and accordingly derived condensation temperatures.

Table 1. Chemical compositions (wt%) and hexagonal lattice parameters of RMNs studied.
Gr  Tc (K)WOsIrMoRuPtNiFe a (nm) c (nm)
  1. Platinum was used during FIB preparation; measured values might be slightly too high from fluorescence contributions. Uncertainties of lattice parameters are 1σ precision (see text).


The sampled RMNs range in size from 200 to 900 nm and show equant (almost spherical) to rarely platy morphologies. To securely fixate particles on the graphite substrate and aid targeting in the ion beam image, selected grains were covered with platinum by electron beam deposition (at 5 keV) prior to FIB processing (FEI Quanta 3-D FEG dual beam FIB-SEM). This approach also preserves external grain surfaces free of ion beam damage. After in situ lift out and attachment to copper grids, final thinning to electron transparency was done with lowered beam currents down to 10 pA at 30 keV energy. Final thicknesses were in the range of approximately 60–80 nm and sufficient to record clear selected area electron diffraction (SAED) patterns from these highly dense materials.

TEM study of the RMNs was carried out using a Philips FEG CM20 TEM (operation voltage: 200 kV) equipped with a Thermo Noran energy-dispersive X-ray (EDX) spectrometer using a high-purity germanium detector. Quantification of EDX spectra was based on the Cliff-Lorimer method using precalibrated (factory default) k factors due to unavailability of suitable reference alloys. Sample thicknesses were estimated from FIB-SEM images. The resulting, independently derived, TEM-EDX compositions are in good general agreement with previously measured SEM-EDX results of Berg et al. (2009) (Fig. 1). The results of both measurements can, however, not be compared directly, because it was not possible to exactly relocate individual grains due to their small sizes. The defect microstructure of RMNs was characterized by conventional bright-field and dark-field imaging and high-resolution (HR-TEM) imaging. SAED and convergent beam electron diffraction (CBED) were used to identify the crystal structures of the particles. If possible, lattice parameters of individual grains were determined from multiple SAED patterns (recorded on film and scanned at high resolution). The uncertainties given in Table 1 are based on many (usually >15) measurements of d-values across the patterns. The camera constant of the TEM was regularly calibrated using standards such as gold particles dispersed on carbon-coated grids.


Crystal Structures

All 15 grains studied are monophase single crystals. The SAED patterns of all grains can be unequivocally indexed in terms of hcp metal structures and Fig. 2 (and Table 1) shows that the derived lattice parameters closely match those of hcp structured, binary Mo55Ru45 (Anderson and Hume-Rothery 1960), which has been described as the mineral hexamolybdenum from the Allende meteorite (Ma et al. 2009). As shown in Fig. 2, all grains cluster densely and we determined average hexagonal lattice parameters of a = 0.276 ± 0.004 nm and = 0.445 ± 0.013 nm (uncertainty given as standard deviation of all determinations, cf. Table 1). A CBED pattern obtained from a very thin edge of grain F1 along the [100] zone axis (Fig. 3) additionally shows systematic absences of reflections in accordance with the hcp structure (space group P63/mmc), namely those for which h + 2= 3n and = odd (= integer).

Figure 2.

 Hexagonal lattice parameters of investigated RMNs based on TEM-SAED patterns. Error bars represent 95% coverage (approximately 2σ). Dashed lines indicate the values for hexagonal Mo55Ru45 (Anderson and Hume-Rothery 1960; Ma et al. 2009). Compositionally unusual RMNs are marked, Gr 02 is Fe-rich and Mo-poor, Gr 10 is Mo- and Pt-rich.

Figure 3.

 Convergent beam electron diffraction (CBED) pattern of RMN grain F1. Systematically absent reflections (arrows) and intensity differences are visible due to small sample thickness at the edge of the grain. Absences and intensities are consistent with a superstructure-free hcp structure (space group P63/mmc).

The group of solely hexagonal grains also includes all those that are compositionally dominated by metals that are not hcp structured at ambient conditions in their pure forms, such as Fe, Mo, Ir, and Pt (amounting up to 60–67 atomic %, cf. Table 1). Grains 02 and 10 are particularly rich in Fe and Mo + Pt, respectively (Table 1), and fall at the lower and upper ends of the range defined by our measured a lattice parameters. However, none of the diffraction patterns of any grain studied shows evidence for exsolution, compositional modulation (e.g., spinodal decomposition leading to diffuse satellite reflections), or superstructure formation (Fig. 4).

Figure 4.

 TEM bright-field images and SAED patterns for selected RMNs of this study. Orientations of images and diffraction patterns are corresponding. Zone axes (ZA) are indicated as a) Gr 02: Euhedral, platy shape. b) Gr 05: Euhedral, isometric shape (based on additional SEM observations). c) Gr 06: Anhedral shape. d) Gr 08: Euhedral, almost isometric shape. e) Gr 11: Subhedral, isometric shape. f) Gr 15: Euhedral, isometric shape. The smallest RMN sampled.

Grain Morphologies

Most of the grains studied in cross-section show straight edges that correspond to crystal faces based on recorded SAED patterns (Fig. 4). Some grains are fully or almost fully euhedral (e.g., Gr 02, Gr 15; Figs. 4a and 4f), others appear to have no crystallographically defined morphology at all (e.g., Gr 06; Fig. 4c). Commonly, those grains with partially euhedral morphology have equant shapes characterized by more or less equal contributions of prismatic {010} and {110}, pyramidal {011}, and basal {001} faces. A remarkable exception is the Fe-rich grain 02 (Fig. 4a), which is dominated by the {001} form and hence displays a platelet shape. There seems to be no size preference for euhedral development, both the largest (Gr 02, Fig. 4a) and the smallest grain (Gr 15, Fig. 4f) studied show well-developed faces.

Compositional Variations

The compositions of RMNs studied here span the whole range of calculated condensation temperatures of Berg et al. (2009) from 1406 to 1601 K (Table 1). Results of EDX analysis in scanning TEM mode at high spatial resolution (<10 nm) indicate the absence of compositional zoning across individual grains (Fig. 5a–d). For all studied grains, the compositional variations between core and rim regions are low and typically range between 1% and 10% (Fig. 5e). In few cases, particularly in the unusually Fe-rich grain 02, we detected small Fe (and likely Ni) deficits at the outermost rims (less than 20 nm from the surface). As the depletion is limited to Fe and Ni and confined to the outermost margins of the otherwise uniformly composed grains, these observations likely represent artifacts of the measurement (decreased X-ray self-absorption at the grain edge) or sample preparation (oxidation and acid leaching of Fe) rather than truly primary features.

Figure 5.

 Compositional variations within and among the studied RMNs based on TEM-EDX. X-ray intensity ratios are used for better accuracy in data reduction and statistical treatment. a) TEM-EDX profile from rim to center of Gr 05 (Fig. 4b), live counting time was 200 s per point. Error bars are 95% coverage (approximately 2σ) based on counting statistics; the shaded areas are the expected ranges of scatter about the means if only counting statistics contribute to the measured variations. b) Difference between center and rim intensity ratios (vertical) plotted against the mean (bulk) intensity ratio of the same grain (horizontal). The differences between centers and rims of individual RMNs are typically smaller than 10% of the respective bulk composition (dashed diagonal lines). Hence, internal variations within individual RMNs are small compared with variations among the sampled RMNs. Error bars are 95% coverage based on counting statistics. Note that vertical error bars are large due to logarithmic scaling. Error bars with arrows cross zero; hence, there is no significant difference between centers and rims.


Microstructures within individual RMN grains are scarcely observed. As indicated by clearly defined diffraction patterns and conventional bright-field and dark-field TEM imaging, the grains are nearly perfectly crystalline with few defects. High-resolution (HR) TEM imaging is commonly hampered due to the high density of the metal and the relatively large thickness of FIB-prepared sections. In case of grains F1 and 04, we were, however, able to obtain useful HR-TEM images. Grain F1 shows very few planar defects parallel to (001), which apparently terminate at dislocations and do not extend far within the grains (Fig. 6). The displacement of lattice planes across these defects indicates that they are stacking faults. Grain 04 showed no readily discernable defect structures, which appears to be representative for most grains.

Figure 6.

 High-resolution TEM image of grain F1 parallel to zone axis [100] (SAED pattern as inset). Shown is one of the few defect microstructures observed in the sampled RMNs (the edge of the thinned sample is seen in the lower left part of the image, this does not correspond to the original edge of the grain due to material loss during FIB preparation). The structure resembles a pair of stacking faults, which terminate in a dislocation. However, the true nature of this planar defect is not fully clear.


The uniformly hexagonal structures of the 15 Murchison RMNs studied strongly suggest that the alloys initially formed as a compositional continuum of single, hcp structured phases. The Mo-dominant grains of our study match the compositions, structure, and approximate lattice parameters of the hexamolybdenum phase previously described from CAIs in the Allende and NWA 1934 CV3 chondrites, where it occurs in association with Sc-Zr-rich ultrarefractory oxides (allendeite), perovskite, and Os-rich alloys (Ma et al. 2009) and krotite (CaAl2O4)/grossite (Ma et al. 2011), respectively. Particularly interesting is that the hexagonal structure in these samples appears to be preserved even for compositions with just about 20–30 atom % hcp metals (Ru + Os). Our findings of solely hexagonal metal alloys are also in accord with previous TEM observations of Eisenhour and Buseck (1992), who investigated five Mo- and W-bearing RMNs hosted in spinel of a fluffy type-A CAI from the Allende CV3 meteorite. Also, these grains were euhedral, hcp structured, and without detectable internal compositional heterogeneity. Recently, Croat et al. (2012) reported three hcp structured, Mo- and W-bearing RMNs from graphite grains of likely presolar origin, supporting the direct condensation from a gas phase. As mentioned above, low vapor pressure is the only common property of all the diverse metals in these metal alloys. This requires high temperatures for the formation of the metal nuggets, either as condensates from a gas of solar composition or as residues from evaporation. We will first describe to what extent evidence from the structure of the alloys constrains their origin and then discuss their possible formation as condensate and as evaporative residues in more detail.

Structural Constraints on the Origin of Refractory Metal Nuggets

Considering the large number of constituents and the highly variable compositions, it is surprising that without exception, all yet sampled RMNs from either Allende or Murchison are hcp structured, homogeneous alloys. One would rather expect to find a variety of different structures and exsolution features. Hence, to constrain the formation of RMNs, it is important to understand whether the hcp structure formed stably or metastably in the relevant temperature interval. Unfortunately, no experimental data exist on the phase relation in the quaternary system Mo-Ru-Os-Ir, and less so in even more extensive systems involving additional Fe, Pt, W, etc. The Mo-Ru binary system clearly shows a fairly wide miscibility gap between bcc and hcp solid solutions at temperatures below 1600 K. Furthermore, the system contains an intermetallic compound (tetragonal σ-phase, approximately Mo5Ru3) at temperatures between 1416 and 2188 K (Okamoto 2000). Also, the Ir-Os, Ir-Ru, and Mo-Os binary systems show miscibility gaps between fcc and hcp alloys at temperatures below 2000 K (Massalski et al. 1986; Okamoto 1992, 1994). Consequently, based on these binary phase diagrams alone, one could expect to find structural diversity among Mo-Ru-Os-Ir alloys. However, the Mo-Ir, Mo-Pt, Mo-Rh, and Fe-Ir binary systems of bcc and fcc metal endmembers show relatively wide homogeneity ranges of nonordered (i.e., superstructure-free) hcp phases at intermediate compositions (Massalski et al. 1986). These phases, known as ε-phases, occur, for example, as complex alloys in spent nuclear fuel, where the hcp structure can prevail over large compositional ranges (e.g., in the Mo-Pd-Ru system; Rand and Potter 1981). Although the temperature ranges of hcp fields in the binary bcc-fcc alloy systems; are either too high (>1880 K for Mo-Ir, >1750 K for Mo-Pt) or too low (<900 K for Fe-Ir) with respect to RMN condensation temperatures, this perspective offers at least a possibility that rather extended, stable ε-phase fields might exist in the multicomponent systems and compositional ranges relevant to RMNs (Fig. 7). Grain 02, which is unusually poor in Mo and essentially composed of Fe, Os, Ru, and Ir (Table 1) likely adopted a hcp structure due to extended hcp fields in the Fe-Ru and Fe-Os binary systems (Swartzendruber and Sundman 1983) and the occurrence of the ε-phase in the Fe-Ir system (Massalski et al. 1986).

Figure 7.

 Pseudoternary plot of the RMN compositions measured in this study by TEM-EDX. Elements are grouped based on their crystal structure in pure form. Gr 02 is Fe-rich and appears anomalous. However, if Fe is grouped to the bcc metals, Gr02 will fall to the center and little will change among the positions of the other grains. On the left and lower edges, selected 2-phase fields of relevant binary phase diagrams are shown (for references see text); on the right edge, 1-phase fields of hcp structured ε-phases within bcc-fcc systems are shown. Unless noted, the boundaries are for approx. 1500 to 1600 K.

The possibility that the hcp structure was retained metastably during the growth of RMNs appears rather unlikely: Os-rich and therefore hexagonal metal is the first condensate at high temperatures (>1600 K) expected from equilibrium condensation calculations, and therefore, the nucleation of hcp structured alloy is reasonably the first step of RMN formation. In the hypothetic scenario that a cubic structure would be the thermodynamically stable alloy form at lower temperatures, hexagonally structured RMNs may continue to grow metastably by topotaxy on the previously condensed grains. This process could be favored by hindered nucleation of new cubic alloy crystals in the ambient gas and the resulting oversaturation of condensable metal species. However, such a mechanism would likely work only if the actually stable alloy is bcc structured, because a fcc structure might easily heterogeneously nucleate on the hexagonally close-packed (001) faces of the hcp alloy. Furthermore, in any of such cases, strong compositional zoning and/or deviation from the calculated condensation sequence would be expected, which is not observed in our RMNs. Diffusion calculations based on extrapolated interdiffusion coefficients of hexagonal Mo-Os alloys (Erley and Wagner 1973) show that homogenization of 400 nm sized RMNs at 1500 K is likely to occur within hours or days (D ∼ 3 × 10−17 m2 s−1). Considering the cooling rate of 0.5 K yr−1 suggested by Berg et al. (2009), homogenization as well as equilibration with the surrounding gas would be well feasible in the RMN forming environment. Under such conditions of high atomic mobility within the alloys, it is unlikely that a metastable phase or zoning pattern would persist or that exsolution would be kinetically inhibited. Furthermore, it is probable that any defects within the metal structures would be obliterated almost instantaneously, adding to the argument of high stacking fault energies as discussed above. All in all, it seems plausible that the hcp-structured RMNs formed as thermodynamically stable ε-phase alloys.

From a microstructural perspective, it can be expected that condensation from a gas phase will result in extended planar defects, such as stacking faults or microtwins, which originate from misarranged addition of atoms to the dense-packed atomic layers on the surface of the growing grains. Defect densities are expected to be high when fault energies are low and growth rates are rapid. In the case of presolar silicon carbide and meteoritic nanodiamond grains, abundant stacking faults and microtwins are microstructural indicators of their condensation origin (e.g., Daulton et al. 1996, 2003). The formation of stacking faults in SiC is particularly favored by rather low stacking fault energies of <40 mJ m−2 (Ning and Ye 1990; Hong et al. 2000). In contrast, metals like Mo and Ru have stacking fault energies in the range of 200–400 mJ m−2 (Hirschhorn 1963; Igarashi et al. 1991). This fact may explain the observed scarcity of planar defects in the RMNs studied, considering that cooling and, therefore, condensation rates, were likely relatively low (Berg et al. 2009). In addition, microstructures might have been erased by annealing through secondary thermal processing, e.g., during heating of the CAIs that hosted RMNs.

In summary, we find that (1) a hexagonal ε-phase alloy of refractory metals appears to be a thermodynamically stable phase. (2) Microstructural indicators expected for condensates are lacking. (3) Elemental zoning patterns, which may be expected during fractional condensation, are not present. The latter two points may simply reflect high-temperature formation and processing, which would erase structural defects and compositional gradients by diffusion.

Refractory Metal Nuggets as Condensates

The uniform and likely thermodynamically stable structures of RMNs are compatible with a condensation origin, although there are no definitive signs that would point to condensation as the only possible mechanism of RMN formation. As outlined above, high temperatures have likely erased or prevented such evidence. The main arguments for condensation come from the chemical composition of RMNs. The compositions of RMNs closely follow predictions by condensation calculations, assuming ideal solid solution of all metals involved, as shown by Berg et al. (2009). Particularly noteworthy is the correlation of the hcp metal Ru with the bcc metal Mo and the anticorrelation of the two hcp metals Os and Ru (Fig. 1). Such correlations are predicted by condensation calculations, assuming condensation into a single alloy, which has been confirmed by the TEM work described above. Single alloy condensation has been assumed by Palme et al. (1994) in explaining the patterns of opaque assemblages in some CAIs. Sylvester et al. (1990) suggested condensation of refractory metals in three distinct phases according to the structure of the individual metals. The final metal assemblage is then a mechanical mixture of hcp-, bcc-, and fcc-metal alloys. The results of this study and of the Berg et al. (2009) calculations refute such a model.

The presence of Mo and W in the proportions predicted by condensation calculations in RMNs also supports condensation models. Both elements can be easily oxidized and are then lost from the alloy as both Mo and W oxides are highly volatile. The calculations of Fegley and Palme (1985) show that at an H2O/H2 ratio of 5 × 10−4, which is a factor of ten higher than the canonical solar ratio, a significant Mo-anomaly is present in condensing metal. This oxygen fugacity is still 6 orders of magnitude below the IW-buffer, reflecting the very reducing conditions in the H2-dominated early solar nebula. As a Mo-anomaly is not present in our suite of nuggets, formation at extremely reducing conditions is required. The fully condensed W and Mo in the RMN grains studied here also excludes losses of W and Mo by parent-body processes, probably because nuggets were fully enclosed in mineral grains.

Wark (1986) reported data on RMN in a CAI of the Allende meteorite. Nuggets in the core of the CAI have a more refractory composition with enhanced Re, W, Os, Ir, and Mo, whereas nuggets near the rim of the CAI are higher in the less refractory metals, such as Ru, Pt, Rh, Fe, and Ni, clear signatures of condensation. Moreover, evidence for condensation is not limited to refractory metals. Rare earths elements and other refractory lithophile elements provide evidence for condensation of refractory elements at high temperatures, recorded in CAIs (Boynton 1975; Davis and Grossman 1979).

Refractory Metal Nuggets as Residues from Evaporation

The calculation of RMN composition was carried out by assuming thermodynamic equilibrium between gas and solid. A given solid metal–gas equilibrium can be either reached by condensation (cooling) or by evaporation (heating) in a very reducing environment. As pointed out by Cameron and Fegley (1982), several cycles of evaporation and recondensation of refractory metals might have occurred. The preservation of Mo and W in the metallic state within RMNs requires low oxygen and sulfur fugacities as expected for a solar gas (Blum et al. 1989), because otherwise these elements would volatilize in the form of their oxides (Palme et al. 1998) or convert to sulfides (particularly MoS2).

RMNs could represent the last solid after evaporation of lower temperature phases. Prime candidates for such materials would be Fe-Ni metal grains with solar siderophile trace element abundances and sizes in the order of several tens of micrometers. CAIs as evaporation precursors appear less likely, because RMNs can occur abundantly in single CAIs (amounting to a hundred or more, e.g., Schwander et al. 2012), which would imply that either the precursor CAIs were extremely metal-rich (which is not observed in preserved CAIs), or somehow residues of many CAIs accumulated in few remaining CAIs (which is rather difficult to envision as other residual solids would accumulate as well in those CAIs). Anyhow, the evaporation model works realistically only if a canonical solar gas is present:

  • 1 Evaporation of RMN precursors in dust- or ice-enriched regions by shock-wave or radiation heating will occur at higher oxygen fugacities compared with the canonical solar gas and will likely lead to extensive or complete loss of W and Mo from the metal (Fegley and Palme 1985).
  • 2 Heating of chondritic material after partial dissipation of the H2-rich nebula would imprint a much higher oxygen fugacity on the system compared with the solar nebula and, hence, would inevitably lead to losses of W, Mo, and eventually Re and Os (Palme et al. 1998).
  • 3 Equilibrium between solid-metal grains and gas requires that metals must not evaporate into vacuum. For example, a Ru atom evaporating from a refractory metal alloy must subsequently contribute to the partial pressure of Ru above the solid metal. Thus, evaporation into vacuum would not be sufficient to account for the apparent equilibrium compositions observed.

In the condensation scenario, equilibration with the surrounding gas is assumed to have ceased at different temperatures, presumably by the incorporation of RMNs into contemporaneously forming silicate or oxide phases, e.g., spinel as shown by Eisenhour and Buseck (1992), or spinel and hibonite as described by Wark (1986). This eventually resulted in the observed compositional variations. It is not clear how the sequence of alloys reflecting a continuum of formation temperatures (Berg et al. 2009) could be produced by evaporation instead of condensation. For example, if the RMNs originated as residues from evaporation of individual, and necessarily tens of μm large, Fe-Ni metal grains, then this process must have happened in a region characterized by a relatively stable temperature gradient bracketing the calculated equilibrium temperatures, or a mechanism must have existed that provided isolation at some point during locally rising temperatures (obviously, it cannot have been the condensation of a host phase). Both cases are very difficult to reconcile with realistic scenarios.

One could also envision a model where refractory metals form solid alloys by precipitation from liquid silicates during evaporative mass loss. This model could be attractive, because it is now clear that many CAIs were once molten and kinetic isotope fractionations indicate rapid evaporation (e.g., Shahar and Young 2007). Although fO2 was low during CAI melting (Grossman et al. 2008) and probably favorable for keeping Mo in the alloys (most oxidation to form opaque assemblages likely occurred in the chondritic parent bodies), a contemporaneous melting of RMNs during CAI melting or precipitation from a CAI melt is improbable for several reasons:

  • 1 (Near-) equilibrium crystallization would result in single or only narrowly variable alloy compositions, unlike what is observed among the RMNs sampled.
  • 2 Within the multicomponent system, fractional crystallization (e.g., by entrapment) is unlikely to mimic the volatility-controlled, calculated compositional trends, which fit our data well. In such a case, it is expected that the geochemical distribution coefficients control the partitioning of refractory siderophile elements between metal and melt.
  • 3 The platelet shape of the Fe-rich grain 02 suggests that it has never been molten at any time, despite its melting point is likely among the lowest of the observed RMN compositions. If it had melted, a droplet shape or at least a more isometric crystal habit could be expected due to the high interface energy between metal and silicate melt.

As outlined above (including the difficulty of isolation), evaporation of Fe-Ni metal grains could produce the same sequence of RMN compositions during rising temperatures, if the starting material had a solar relative abundances of refractory metals. However, if the evaporating material had been fractionated previously, particularly by oxidation and volatilization of Mo and W (Palme et al. 1998) or, more applicable to molten/sintered CAIs, by partitioning between metal and sulfides, oxides, or silicates, an evaporation origin of the observed compositional variations would be unlikely to comply with condensation calculations invoking solar element ratios in the gas phase.

In principle, it is possible to postulate a perfect equilibrium evaporation scenario under reducing conditions of a canonical solar gas that is the exact reversal of the equilibrium condensation assumed in the calculations of Berg et al. (2009). In this case, ignoring the difficulties discussed above, the resulting RMNs would be exactly the same and it would be impossible to test one scenario against the other. However, if we assume that RMNs are indeed formed by such evaporation, then we have to assume that their precursor (e.g., μm-sized Fe-Ni metal grains), before evaporation, are formed by equilibrium condensation—otherwise evaporation would not end up with the same results as indicated by equilibrium condensation calculations (e.g., any metal processed in chondrule-forming environments or parent bodies would have likely lost substantial amounts of Mo and W due to oxidation). All the evaporation scenarios outlined above are rather complex and appear less likely to happen than a direct formation of RMNs by condensation from a solar gas.


On the basis of our TEM observations, we find that all 15 studied RMNs from the original sample of Berg et al. (2009) have hexagonal crystal structures consistent with a superstructure-free, hexagonal close-packed solid solution among the constituting elements. All RMNs are single crystals without any evidence of exsolution, superstructures, or internal compositional variations. Most of the grains show subhedral to euhedral external shapes and are internally free of defect microstructures. Among the possible formation scenarios of RMNs (condensation, evaporation, precipitation from melt), condensation from a gas of solar composition appears to be the most plausible, an evaporation origin is very difficult to explain and unlikely, and precipitation from a melt can be excluded. Our findings of structurally uniform grains support previous condensation models assuming the formation of a single alloy phase. Obvious signs of a condensation origin, such as extended stacking faults or compositional zoning have likely been erased or never existed due to high defect energies, high formation temperatures, slow cooling rates, small grain sizes, and rapid internal diffusion. However, whether the last step in solid-gas reaction of RMNs was condensation or evaporation cannot be determined at present with certainty due to missing or yet unrecognized nonequilibrium effects. Investigations on the correlation between the compositions of individual RMNs and their position within the host CAIs and their host minerals can potentially provide important clues.

Based on the evaluation of binary phase diagrams, the observed phase homogeneity is surprising, but may be explained by extended fields of the hcp structured ε-phase in the relevant (4+n)-component systems. The broad homogeneity range of ε-phase alloy is suggested by its formation in several binary systems of involved bcc and fcc metals and the occurrence of multicomponent ε-phase alloys in spent nuclear fuel. We therefore argue that the observed hexagonal ε-phase is indeed the thermodynamically stable form of refractory metal alloys. Future experimental investigations of phase relations and refined modeling of condensation, e.g., invoking realistic activity coefficients and nonequilibrium effects (e.g., Tanaka et al. 2002), are needed and expected to be helpful to better understand the formation of RMNs, particularly the last step, and the conditions and processes within the high-temperature regimes of the solar nebula.

Acknowledgments— We gratefully acknowledge financial support by the Leibniz program of the German Research Foundation (DFG; LA 830/14-1 to F.L.) and the ENB program of the Bavarian State Ministry of Sciences, Research and the Arts (to D. H.). Reviews by A. Campbell and K. Croat and comments by Sasha Krot helped to improve the manuscript and are gratefully acknowledged.

Editorial Handling— Dr. Alexander Krot