Forsterite-bearing type B refractory inclusions from CV3 chondrites: From aggregates to volatilized melt droplets


Corresponding author. E-mail:


Abstract– Detailed petrologic and oxygen isotopic analysis of six forsterite-bearing Type B calcium-aluminum-rich inclusions (FoBs) from CV3 chondrites indicates that they formed by varying degrees of melting of primitive precursor material that resembled amoeboid olivine aggregates. A continuous evolutionary sequence exists between those objects that experienced only slight partial melting or sintering through objects that underwent prolonged melting episodes. In most cases, melting was accompanied by surface evaporative loss of magnesium and silicon. This loss resulted in outer margins that are very different in composition from the cores, so much so that in some cases, the mantles contain mineral assemblages that are petrologically incompatible with those in the cores. The precursor objects for these FoBs had a range of bulk compositions and must therefore have formed under varying conditions if they condensed from a solar composition gas. Five of the six objects show small degrees of mass-dependent oxygen isotopic fractionation in pyroxene, spinel, and olivine, consistent with the inferred melt evaporation, but there are no consistent differences among the three phases. Forsterite, spinel, and pyroxene are 16O-rich with Δ17O ∼ −24‰ in all FoBs. Melilite and anorthite show a range of Δ17O from −17‰ to −1‰.


Forsterite-bearing Type B inclusions (FoBs) are a relatively rare variety of refractory inclusions identified so far only in CV3 and CB carbonaceous chondrites (Clayton et al. 1984; Wark et al. 1987; Davis et al. 1991; Krot et al. 2001a; MacPherson et al. 2005; Bullock et al. 2007). Most FoBs are unambiguously igneous (i.e., solidified from melt droplets), based on their spheroidal shapes, crystallization sequences, and textures consistent with the phase equilibria expected from their bulk compositions, and, in a few cases, vesicles (e.g., MacPherson et al. 1981). The presence of abundant primary forsterite distinguishes FoBs from all other varieties of igneous calcium-aluminum-rich inclusions (CAIs) (e.g., Type B, compact Type A, Type C), and also means that FoBs are “less refractory” than other CAIs owing to their enhanced magnesium and silicon relative to calcium, aluminum, and titanium.

What sets FoBs apart (and makes them so interesting) is that a disproportionately large fraction of them are isotopically anomalous relative to other types of CAIs. One FoB, 1623-5 from Vigarano, is known to have Fractionation and Unidentified Nuclear isotope anomalies (FUN; Wasserburg et al. 1977). 1623-5 contains little or no radiogenic 26Mg (26Mg*) (MacPherson et al. 1995; McKeegan et al. 2005), but has large nucleosynthetic anomalies in elements such as calcium, titanium, barium, and strontium (Davis et al. 1991; Loss et al. 1994). Two other previously characterized FoBs (TE and CG14) have no known nuclear anomalies, but do show significant mass-dependent isotopic fractionation in magnesium, silicon, and oxygen (“F” inclusions), possibly as a result of melt evaporative loss of these elements (Wasserburg et al. 1977; Clayton et al. 1984; Davis et al. 1990; MacPherson et al. 2004; Krot et al. 2008). Finally, several non-FUN (−F) FoBs show only small degrees of mass-fractionation in O and Mg (MacPherson et al. 2008), but contain 26Mg* corresponding to the canonical 26Al/27Al ratio of approximately 5 × 10−5 (e.g., Petaev and Jacobsen 2009). Petrographically and mineralogically, however, the FUN and “F” FoBs are indistinguishable from the non-FUN FoBs, suggesting that the processes responsible for the differences in degree of mass-dependent fractionation did not produce differences in bulk composition or mineralogy.

FoBs are excellent candidates for constraining formation and evolution of igneous CAIs. Their forsterite-rich bulk compositions, in contrast with other types of CAIs, indicate that they formed from precursors that were enriched in MgO and (to a lesser degree) SiO2. They may be hybrids of “normal” CAI-like material plus very olivine-rich materials, similar to amoeboid olivine aggregates (AOAs; Krot et al. 2004) or forsterite-rich accretionary rims (Krot et al. 2001b). Also, as documented herein, many FoBs show clear petrologic evidence for melt evaporation loss of magnesium and silicon (see also Davis et al. 1991).

The petrologic, oxygen isotopic, bulk chemical composition, and mineral chemical data presented herein for six FoB CAIs from CV chondrites provide an unusually complete formation history for this group of objects, evolving from sintered to weakly melted aggregates all the way through to completely melted spherules that experienced melt volatilization from their outer margins.


Bulk compositional data and high-resolution backscattered electron (BSE) images for polished sections of six FoBs (TS35-F1, ALVIN, and SJ-101 from Allende; E60 and E64 from Efremovka; and 3137 from Vigarano) were obtained using the FEI NOVA NanoSEM 600 scanning electron microscope (SEM) at the Smithsonian Institution, operated at 15 keV with a sample current of 2–3 nA. The SEM is equipped with a Thermo-Noran energy dispersive X-ray analytical system, and data are stored and processed using Thermo-Scientific Noran System Six software. The System Six software allows for full-spectrum imaging, in which a complete energy-dispersive spectrum is collected and stored for each pixel within a map. Multi-frame map mosaics with 10–20% frame overlap provide full coverage of large CAIs at sufficient spatial resolution to resolve grains down to approximately 2 μm in diameter. The bulk composition of each frame (corrected for overlap) was determined by summing the compositions for all pixels, summing the frame compositions, and normalizing the totals to 100%. The software allows extraction of data from irregularly shaped areas, so that unwanted pixels (epoxy, cracks, meteorite matrix) can be eliminated. This feature also allows the composition of individual areas within each inclusion to be determined. Spectra were quantified using Gaussian spectrum fitting and standardless Phi-Rho-Z matrix correction. The measured bulk compositions all show only minor secondary components such as sodium and iron. This is consistent with the minor degrees of secondary alteration observed in these objects relative to other varieties of inclusions from CV3 chondrites. Therefore, no corrections were made to the compositions prior to plotting on compositional diagrams.

Mineral compositions were measured with a JEOL JXA-8900 automated five-spectrometer wavelength-dispersive electron microprobe at the Smithsonian Institution, operated at 15 keV accelerating potential, 15–20 nA beam current, and counting times in the range of 10–40 s (depending on the element). Ten elements—Al, Ca, Mg, Si, Ti, Na, Fe, K, Cr, and Mn—were analyzed.

Oxygen isotopic data were collected using the Cameca ims 1280 ion microprobe at the University of Hawai’i in a technique similar to that described by Makide et al. (2009). A focused Cs+ ion beam of approximately 1.8 nA was rastered over a 25 × 25 μm region for presputtering. Oxygen-isotope compositions were collected from the center of the presputtered region by reducing the raster size to 10 × 10 μm. The 16O and 18O ions were measured at Mass Resolving Power (MRP) of approximately 2000 using multicollector Faraday cups; 17O was measured using a monocollector electron multiplier with MRP of approximately 5500, sufficient to resolve 16OH interference. Typical count rates of 16O and 17O were approximately 4 × 108 and approximately 1.5 × 105 cps, respectively. To measure small grains (<20 μm), some measurements on SJ-101 were made with an approximately 200 pA primary ion beam rastered over a 5 × 5 μm region, and with a combination of multicollection mode and peak-jumping. 16O and 17O were measured simultaneously using multicollection FC and monocollection EM, respectively. Subsequently, 18O was measured via monocollection EM by magnetic field peak switching. Typical count rates for 16O with this protocol were approximately 9 × 107 cps. Instrumental mass fractionation (IMF) effects were corrected by analyzing San Carlos olivine, Burma spinel, Miyakejima anorthite, and Cr-augite oxygen-isotope standards. Measurements of melilite grains were corrected by assuming the melilite IMF to be similar to that of olivine. The standards were analyzed repeatedly before and after each run. Reported errors (2σ) include both the internal measurement precision and the external reproducibility (approximately 1–1.5‰ in both δ17O and δ18O) of standard data obtained during a given session. After analysis, the location of each probe pit was reimaged to check for beam overlap between phases, and to identify large cracks or impurities that may have affected the result.


Mineralogy and Petrography

Figure 1 illustrates paired BSE images and combined Mg-Ca-Al (Kα) X-ray elemental maps for each FoB inclusion studied. The inclusions are presented in order of increasing degree of petrologic complexity, ranging from SJ-101 to TS35-F1. The justification for this interpretation is documented below and explained fully in the Discussion section. In brief, a visual clue to the degree of thermal processing can be gleaned from the presence and thickness of a pyroxene-rich and largely olivine-free outer mantle (see the X-ray maps in Fig. 1). SJ-101 and ALVIN have insignificant mantles, whereas in TS35-F1 and 3137, the mantles are the dominant components of the entire inclusions.

Figure 1.

 Paired back-scattered electron images (top) and Ca-Al-Mg overlay X-ray maps (bottom) for the six FoB CAIs in this study. For this and all following composite X-ray maps, Mg = red, Ca = green, Al = blue. FR and FF indicate the forsterite-rich and forsterite-free areas of each CAI. In each composite image, olivine = red, spinel = lavender, magnesium-rich melilite = yellow green, aluminum-rich melilite = sky blue, pyroxene = dull green/brown, anorthite = dark blue. The bright pistachio-green phase that is prominent in the matrix of ALVIN (esp. filling the large vesicle [V] at lower right) and around TS35-F1 is a mixture of andradite, hedenbergite, and diopside. a) SJ-101. Note how the numerous nodular and sinuous regions that are largely free of forsterite are enclosed in a forsterite-rich matrix. The red box indicates the location of Fig. 2. b) ALVIN. The numerous black holes are vesicles; the large ruptured vesicle at lower right is filled with matrix material and Ca-Fe-rich silicates as noted above. The yellow box indicates the location of Fig. 3, which overlaps onto the large forsterite-free and spinel-rich island described in the text. c) E64. The red box indicates the location of Fig. 4. d) E60. The red and yellow boxes indicate the location of the images in Fig. 5. Note that the forsterite-free mantle is subdivided into inner mantle (IM) and outer mantle (OM), as described in the text. e) 3137. The yellow box indicates the location of Fig. 6. f) TS35-F1. The inclusion is in three separate parts that are joined in the third dimension. The red boxes indicate the location of the images in Fig. 7. As for E60, the inner mantle (IM) and outer mantle (OM) are separately indicated; see text. Note how the extent of the olivine-rich cores diminishes progressively from SJ-101 through TS35-F1: in SJ-101 and ALVIN, olivine occurs very near the margins of the CAIs and there is only a very thin forsterite-free margin, whereas in E60 and 3137 and TS35-F1, olivine-free mantles are quite thick and represent a significant volume fraction of these CAIs.

There is an additional complexity that exists for all but SJ-101, but which is only significant in E60 and TS35-F1. As simply defined above, the mantle of each CAI is the outer forsterite-free portion. But, actually, the mantles are in two parts (see Figs. 1d and f). The inner region, directly overlying the forsterite-rich core, consists mainly of magnesium-rich melilite ± pyroxene ± anorthite ± spinel. Exterior to this “inner mantle” is a thin outer mantle that consists almost exclusively of very aluminum-rich phases: aluminous melilite, spinel, and anorthite.

Below, we briefly characterize each of the inclusions studied.

Allende CAI SJ-101 (Fig. 1a) is a very large (approximately 2.5 × 1.5 cm; the section shown in Fig. 1a does not span the maximum dimensions of the CAI) inclusion described in detail by Petaev and Jacobsen (2009). SJ-101 is similar mineralogically to other, “typical” FoBs, in that it consists of Ca-rich forsterite, Al-Ti-rich diopside, spinel, åkermanitic melilite, and minor anorthite. However, it has a strikingly different internal texture. SJ-101 contains three distinct lithologic units: (1) irregular nodular regions of pyroxene, anorthite, and magnesium-rich melilite, all of which poikilitically enclose densely crowded spinel grains; (2) an enclosing matrix of forsterite and pyroxene that is largely free of spinel; and (3) a thin (<100 μm) outer mantle that is largely free of forsterite and consists of pyroxene, spinel, and anorthite (Fig. 2).

Figure 2.

 Ca-Al-Mg overlay X-ray map of the exterior region of SJ-101. Rounded olivine-free nodules (containing spinel, melilite, pyroxene, and anorthite) are separated by intervening stringers of olivine-rich material. Olivine crystals occur within 100–200 μm of the inclusion margin. An = anorthite; Mel = melilite; Ol = olivine; Pyx = pyroxene; Sp = spinel.

The Allende inclusion known as ALVIN (ALlende Vesicular INclusion; a.k.a. TS45-F1) was described briefly by MacPherson et al. (1981). It is spherical, approximately 1 cm in diameter, and its most striking features are the numerous vesicles that testify to its once having been a molten droplet (Fig. 1b). Vesicles in the interior contain needles of wollastonite, and a large ruptured vesicle at the edge has been in-filled with matrix material. Enclosing one of the vesicles is a fine-grained spinel-rich region, approximately 2.5 × 2.5 mm2, which is entirely free of forsterite (the mainly blue region in the upper right of the X-ray area map of Fig. 1b). The interior of ALVIN consists predominantly of Al-Ti-rich diopside that poikilitically encloses abundant small grains of spinel and forsterite (Fig. 3). Melilite is rare, occurring as irregular grains between pyroxene grains, and is very åkermanite-rich (Åk66–90). Within approximately 0.5 mm of the rim of the inclusion, olivine greatly decreases in grain size and abundance; the outer 300 μm is the aluminum-rich and mostly olivine-free outer mantle that consists of pyroxene, melilite, spinel, and anorthite. Some secondary nepheline replacing anorthite occurs in the outer Al-rich rim.

Figure 3.

 Ca-Al-Mg overlay X-ray map of the interior of ALVIN. Irregularly shaped olivine-free regions are enclosed by olivine-rich material. Abbreviations are as in Fig. 2.

Efremovka CAI E64 is an ovoid object, approximately 0.5 cm in maximum dimension, having a clear core-mantle structure (Fig. 1c). The core contains forsterite and spinel poikilitically enclosed within Al-Ti-rich diopside, plus minor quantities of highly altered magnesium-rich melilite (Fig. 4). The mantle, up to 0.5 mm thick, consists predominantly of pyroxene; forsterite grains are rare. There is only a hint of an outermost aluminum-rich mantle that consists mostly of aluminum-rich melilite and spinel.

Figure 4.

 Backscatter electron image of the interior of E64, showing an altered magnesium-rich melilite crystal within a matrix of pyroxene + olivine + spinel. Olivine and spinel both occur as inclusions within the melilite crystal. The white phase is Fe, Ni metal. Abbreviations are as in Fig. 2.

Efremovka CAI E60 (Fig. 1d) was previously studied for magnesium and aluminum isotopes by Amelin et al. (2002) and Galy et al. (2004) and found to have initial 26Al/27Al ∼ 5 × 10−5 at the time of formation. The inclusion is heart-shaped, approximately 1 cm in diameter, and has a distinct core-inner mantle-outer mantle structure. The core consists of åkermanitic melilite and Al-rich diopside that poikilitically enclose spinel and forsterite. Surrounding the forsterite-rich core is a forsterite-free, melilite- and pyroxene-rich inner mantle (Fig. 5). External to this inner mantle is an outer mantle that is composed of spinel, aluminous melilite, and anorthite. Both the inner and outer mantles intrude into the interior of the inclusion along the bifurcation of the “heart” (see X-ray map in Fig. 1d). Melilite shows a range of composition (Åk20–80) with the aluminous compositions being restricted to the outer mantle. Anorthite is partly replaced by nepheline.

Figure 5.

 Paired backscatter electron image (top) and Ca-Al-Mg overlay X-ray map of Efremovka E60.

Vigarano inclusion 3137 (Fig. 1e) was previously studied by MacPherson and Davis (1992), who measured its magnesium isotopes and found no evidence for extinct 26Al (26Al/27Al < 1 × 10−6) at the time of formation. Those authors also found elevated mass-dependent isotopic fraction (heavy) of magnesium, approximately 10‰/amu. The CAI is several mm in size and irregular in shape, but only a portion of it is contained within the studied thin section. There is a general concentric zonation, although not as clearly seen as in inclusions such as E60. The outer zones are forsterite-free, pyroxene- and melilite-rich, and the outermost margins are plagioclase-rich. Interior forsterite is poikilitically enclosed within Al-Ti-rich diopside and melilite (Fig. 6). Spinel occurs throughout the CAI, enclosed within Al-Ti-rich diopside and melilite grains; spinel grains toward the rim of the inclusion contain up to 6 wt% FeO. Melilite grains toward the outer edge of 3137 are zoned, with Mg-rich cores (approximately Åk86) and Al-rich rims (approximately Åk50).

Figure 6.

 Detailed Ca-Al-Mg overlay X-ray map of Vigarano 3137. The inclusion margin is visible in the lower left corner; note that anorthite occurs only within the outermost approximately 0.5–1 mm of the CAI, and olivine is absent from the anorthite-bearing region. Abbreviations are as in Fig. 2.

Allende TS35-F1 (Fig. 1f) was previously described with the designation AL6S3 by Clayton et al. (1977). It is highly irregular in shape, and prior to extraction from the meteorite was observed to branch into three “arms” through which the thin section was cut. Thus, the branches appear as three separate regions in the figure. This CAI has extensive concentric zoning and, like E60, clearly exhibits three distinct zones: (1) a core consisting of forsterite + pyroxene + minor Mg-rich melilite (approximately Åk84–86), with spinel mainly concentrated in coarse-grained spinel + pyroxene “islands” (Fig. 1f); (2) around the forsterite-rich core (Fig. 7a) is an approximately 1 mm thick forsterite-free and pyroxene-rich inner mantle containing little spinel; (3) a very aluminum-rich outer mantle consisting of spinel, aluminous melilite (Åk6–43), and anorthite (Fig. 7b). Sodium is concentrated in this outermost mantle, although minor quantities of nepheline and sodalite occur within the core of the inclusion.

Figure 7.

 Backscattered electron images of the core (top) and rim of Allende TS35-F1. Abbreviations are as in Fig. 2.

Mineral Chemistry

The composition of melilite in FoBs as a group spans the range Åk6–90 (Fig. 8), and several individual inclusions span this entire range. Most importantly, in all but one case (SJ-101), the range within individual inclusions spans the liquidus minimum in the åkermanite-gehlenite binary system at approximately Åk72. Thus, for these objects, the melilite composition range cannot be explained by fractional crystallization alone. Also, the melilite composition within individual FoBs correlates with the thickness of the forsterite-free mantle. SJ-101 and ALVIN, which have very thin pyroxene mantles, show limited ranges of melilite composition, from Åk80–90 and Åk65–90, respectively, with no gehlenitic melilite present. This contrasts with inclusions that have a thick mantle, such as TS35-F1 and 3137, in which melilite can be as aluminous as Åk6. Finally, melilite composition closely correlates with petrologic setting. All melilite having aluminum-rich compositions (lying on the Al-rich side of the binary minimum composition) is restricted to inclusion mantles. All melilite having magnesium-rich compositions (lying on the Mg-rich side of the binary minimum composition) is restricted to inclusion cores. This observation is fundamental to understanding the evolution of FoBs.

Figure 8.

 Histograms illustrating the range of åkermanite contents of melilite in FoB CAIs, taken together (top) and individually. Overall, the melilite becomes increasingly gehlenitic from SJ-101 to TS35-F1. The vertical dashed bar in each diagram marks the composition of the gehlenite–åkermanite binary minimum.

Olivine in the FoB inclusions is generally almost pure forsterite, with minor amounts of iron present in a few grains (Fo93–100; Fig. 9a), and between 0.7 and 1.9 wt% CaO (Fig. 9b). There is no difference between the CaO contents of olivine grains in the different FoBs.

Figure 9.

 Histograms illustrating the forsterite (a; mole %) and CaO (b; weight %) contents of FoB olivine (90 analyses).

Similar to pyroxene in Type B CAIs, that in FoBs is highly variable in composition. Al2O3 and TiO2 are in the range 0.5–23 wt% and approximately 0–10 wt%, respectively, which mostly are lower than in pyroxenes from Type B CAIs (Fig. 10). As with melilite, there is a progressive change in pyroxene composition that correlates with the thickness of the olivine-free mantle of the FoB. In SJ-101 and ALVIN, the pyroxene is sector-zoned, and shows a clear bimodal distribution of compositions. In the more evolved inclusions such as 3137 and TS35-F1, pyroxene composition overall is more Al2O3- and TiO2-rich, and shows fewer signs of sector zoning. The titanium content of the pyroxene grains is generally too low to obtain accurate Ti3+/Ti4+ ratios based on stoichiometry of electron microprobe analyses. Nevertheless, those analyses with greater than approximately 5 wt% indicate that approximately 60–80% of the titanium is present as Ti3+ (Fig. 11), meaning that FoBs formed under reducing conditions similar to their Type B counterparts (e.g., Grossman et al. 2008).

Figure 10.

 Al2O3 versus TiO2 (weight %; all Ti calculated as Ti4+) compositions of pyroxene in all FoBs taken together (top) and in individual FoBs. All analyses by wavelength-dispersive spectrometry, as described in the Analytical Methods section. Bimodal titanium distributions such as in SJ-101 and ALVIN reflect the presence of sector zoning in the pyroxene crystals. The outlined regions show the range of pyroxene compositions in Type B1 CAIs.

Figure 11.

 A plot of total titanium (calculated as TiO2) versus total cations per formula unit, for all FoB CAIs in this study. Apparent cation deficits are due to the presence of trivalent titanium in the pyroxene structure, and the dotted lines mark the loci of specific Ti3+/Ti4+ ratios. Values at low total titanium are meaningless, but at elevated titanium contents, the ratio of Ti3+/Ti4+ within FoB pyroxene mostly is in the range 0.4–0.8.

Most spinel in the FoBs is low in FeO, with a peak at 0.1–0.2 wt% (Fig. 12). Some grains, particularly toward the rims of the inclusions, contain a higher abundance (up to 17 wt% FeO) that likely is due to secondary alteration.

Figure 12.

 Histogram of FeO contents in FoB spinel.

Bulk Chemical Compositions

Our measured bulk compositions for all six FoB CAIs are given in Table 1 and shown graphically (Fig. 13) on a ternary projection from spinel onto the liquidus plane Al2O3–Ca2SiO4–Mg2SiO4 (see MacPherson and Huss 2005), along with published literature data for other FoBs. Also plotted on this diagram are the bulk compositions of two other groups of objects whose mineralogy is broadly similar to FoBs: AOAs and Al-rich chondrules. All compositions plot on or above the spinel saturation surface (shown in MacPherson and Huss 2005), meaning that melts of such composition are saturated with respect to spinel. Therefore, the phase relationships shown in Fig. 13 (also 15, 16) are valid for these FoBs, which as a group all plot within the forsterite primary liquid phase volume, as expected for strongly olivine-phyric objects. Our objects plot at the forsterite-poor end of an array that extends roughly from the pyroxene primary phase volume toward the composition of forsterite. This FoB array approximately parallels, but mostly does not overlap with, the array defined by Al-rich chondrules. However, the FoBs mostly lie at the magnesium-poor end of the array defined by AOAs. Indeed, AOAs have both the physical and chemical characteristics required for the FoB precursors. This idea is strengthened by the internal structure of SJ-101.

Table 1. Bulk chemical compositions of FoB CAIs.
  ALVIN E60 E64 SJ-101a SJ-101b TS35-F1 Vig 3137
  1. aThis work.

  2. bPetaev and Jacobsen (2009).

  3. n.d. = not detected.

Figure 13.

 Bulk composition of six FoBs, from Allende (TS35-F1, ALVIN and SJ-101), Vigarano (3137) and Efremovka (E60 and E64), plotted on an Al2O3–Ca2SiO4–Mg2SiO4 plot projected from spinel (after Huss et al. 2001). The bulk compositions of the six inclusions are compared with compositional ranges of FoBs and Type B CAI (Beckett 1986), amoeboid olivine aggregates (Komatsu et al. 2001), and Al-rich chondrules (Bischoff and Keil 1983). An = anorthite; CA2 = grossite (CaAl2O4); Cor = corundum; Di = diopside; Geh = gehlenite; Hib = hibonite; Mo = monticellite; Mel = melilite; Fo = forsterite; Pyx = pyroxene; L = liquid.

Oxygen-Isotope Compositions

All oxygen-isotope data collected for each of the six FoBs are shown in Fig. 14 and summarized in Table 2. Data were collected in situ from multiple mineral phases within each inclusion, and are compared with bulk data, where available. Some analyses overlapped cracks or contaminants within mineral grains, which may have affected the result—these data are highlighted and discussed below. Data for all six inclusions plot slightly off the carbonaceous chondrite anhydrous mineral (CCAM) line; in addition, oxygen isotopic compositions of spinel, forsterite, and pyroxene show a small degree (up to approximately 5‰) of mass-dependent fractionation and plot along a slope approximately 0.5 line.

Figure 14.

 Oxygen-isotope compositions of minerals within individual FoBs. Shown on each plot is the terrestrial fractionation line and the carbonaceous chondrite anhydrous minerals line. Mineral abbreviations are Mel = melilite; An = anorthite; Px = pyroxene; Ol = olivine; Sp = spinel; Ne = nepheline. Other abbreviations: c = core of inclusion; r = rim of inclusion. Bulk composition data for Vigarano 3137 from Clayton, unpublished data; bulk data from Allende TS35-F1 from Clayton et al. (1977).

Table 2. Oxygen isotopic compositions of individual minerals in FoBs.
  1. aThese melilite analyses were made using a much smaller rastered area, owing to the fine grain size. See the Analytical Methods section.

Allende SJ-101
 A3 mel#2mel–18.10.9–22.31.0–12.81.1
 A3 mel#1mel–14.00.8–18.61.0–11.41.0
 A2a mel#1mel8.–3.21.1
 AH01a fo#1ol–44.70.8–47.20.9–24.01.0
 A9b fo#1ol–46.90.8–47.70.9–23.31.0
 AH05a sp#1sp–44.21.8–46.81.3–23.81.6
 AH05b sp#1sp–43.01.8–46.61.3–24.31.6
 AH05c fo#1ol–45.50.8–47.21.0–23.51.1
 020 mel#1mel1.00.8–4.61.0–5.11.1
 AH02b mel#1mel–0.40.8–5.41.0–5.21.1
 020 cpx#1px–41.51.2–45.30.8–23.71.1
 AH02a an#1an–13.11.1–17.70.8–10.91.0
 AH02b an#1an–25.31.1–30.40.9–17.31.1
 A16 cpx#1px–44.01.2–46.80.9–23.91.1
 A16 an#1an3.21.1–3.40.9–5.11.0
 A16 an#2an–9.01.1–14.50.9–9.81.1
 A02 cpx#1px–43.81.2–46.80.9–24.01.1
 A02a an#1an3.11.1–3.30.9–4.91.0
 A02b an#1an3.91.1–1.90.9–3.91.1
 A09b cpx#1px–44.91.2–47.10.9–23.81.1
 A23 an#2an2.41.1–3.40.8–4.61.0
 A3 mel#1bmela–13.91.4–19.01.7–11.81.9
Allende ALVIN
Efremovka E64
 px#3 (WLR)cpx–36.81.6–38.71.2–19.61.4
Efremovka E60
 px#3 (WLR)px–35.31.5–37.91.0–19.51.3
Vigarano 3137
Allende TS35–F1

Allende SJ-101

This inclusion is far more heterogeneous in its oxygen isotopic compositions than the other FoBs in this study. Spinel, forsterite, and pyroxene have 16O-rich compositions (Δ17O ∼ −24‰); the pyroxene and spinel show small degrees of mass-dependent fractionation (Table 2; Fig. 14). Melilite and anorthite are variably depleted in 16O relative to spinel, pyroxene, and forsterite, and disperse along the CCAM line: Δ17O ranges from −15‰ to −17‰ to −3‰ (Petaev et al. 2010).

Allende ALVIN

Forsterite, spinel, and pyroxene are 16O-rich (Δ17O ∼ −24‰; Table 2), and cluster along a mass-dependent fractionation line (Fig. 14). Only one analysis each of melilite and anorthite was obtained: these are 16O-depleted and plot slightly below the terrestrial fractionation line (TF) line (Δ17O ∼ −3‰), just off the CCAM line. Both of these analyses overlapped minor cracks.

Efremovka E64

Spinel, forsterite, and pyroxene are 16O-rich (Δ17O ∼ −23‰; Table 2), and cluster along a mass-dependent fractionation line (Fig. 14). The single melilite analysis is 16O-poor and plots on the CCAM line just below its intersection with the TF line (Δ17O ∼ −2‰). No anorthite grains were analyzed.

Efremovka E60

Pyroxene, spinel, and forsterite are 16O-rich (Δ17O ∼−24‰; Table 2) and exhibit a small degree of mass-dependent fractionation. One exception is a pyroxene analysis from the Wark-Lovering rim sequence, and which is less 16O-rich than interior pyroxenes. In contrast to other FoBs, anorthite in E60 is 16O-rich (Δ17O ∼ −24‰) (Nagashima et al. 2010); one analyzed grain shows mass-dependent fractionation. Melilite is 16O-poor and plots just below the TF line (Δ17O = −1.5 ± 1.5‰).

Vigarano 3137

Forsterite, spinel, and pyroxene are 16O-rich (Δ17O ∼ −24‰; Table 2), and plot along a mass-dependent fractionation line (Fig. 14). There is no correlation between isotopic composition and location of the analyzed spots (i.e., core or rim) within the inclusion. Melilite is 16O-poor and mostly plots just below the TF line (Δ17O ∼ −1.0 ± 1.7‰). Anorthite is intermediate between melilite and pyroxene/spinel. The bulk oxygen isotopic composition of this inclusion (Clayton, unpublished data) also plots between olivine-pyroxene-spinel and melilite, just to the right of the CCAM line.

Allende TS35-F1

Forsterite, spinel, and pyroxene are 16O-rich (Δ17O ∼ −24‰; Table 2) and tightly cluster along the mass-dependent fractionation line. Anorthite and melilite are 16O-poor (Δ17O ∼ −3.0 ± 1.9‰) and plot on the CCAM line. Bulk oxygen-isotope compositions measured for the core and rim of TS35-F1 (Clayton et al. 1977) are intermediate between the spinel–olivine–pyroxene data and the anorthite–melilite data collected in this study, with the core being much more 16O-rich.

In summary, forsterite, spinel, and pyroxene are 16O-rich with Δ17O ∼ −24‰ in all FoBs we analyzed. Melilite and anorthite show a range of Δ17O from −17‰ to −1‰ because these phases are more susceptible to oxygen isotope exchange with an external reservoir. More work is needed to understand what part of this range is due to exchange with nebular gas of different composition (Simon et al. 2011) or interaction with an aqueous fluid in an asteroidal setting (cf. Krot et al. 1995).


Deciphering the origin and evolution of FoBs requires understanding the separate origins of the cores and mantles of the CAIs, and the relationship between them. To this end, the bulk compositions of the forsterite-rich cores and those of the forsterite-free surrounding mantles (in all diagrams, inner and outer mantles are combined into a single average composition) were separately determined for each CAI. These are plotted along with the overall bulk compositions in Fig. 15, which is a ternary projection from spinel onto the liquidus plane CaAl2Si2O8 (anorthite)—CaAl2SiO7 (gehlenite)—Mg2SiO4 (forsterite) (Stolper 1982). This diagram does not cover the compositional breadth shown in Fig. 13, but is better suited to examine in more detail the phase relationships within these (spinel-saturated) CAIs. Also plotted in Fig. 15 are the representative vectors showing the effect of Mg and SiO evaporation. Neither SiO nor Mg plots within the confines of this diagram nor even in the same plane. For any CAI bulk composition, the effect of Mg evaporation is to project through the bulk composition away from a point far below the forsterite apex, and for SiO loss to project through the bulk composition away from a point far off to the right of the diagram. Each of the plotted CAIs defines a clear evolutionary trend from forsterite-rich core to forsterite-depleted mantle. The most striking features of Fig. 15 are (1) the core-mantle trends of all the CAIs cross phase-petrologic boundaries; and (2) the trends of the six objects are not parallel to one another. Indeed, the core-mantle trends do not in general parallel the trend defined by the bulk inclusions as shown in Fig. 13. Feature (1) means that the core and mantle of each are not in equilibrium with each other and must have somewhat separate origins. Feature (2) indicates that the mantles of different CAIs evolved in different ways from one another. It is also clear from Fig. 15 that the precursor objects for these FoBs had a range of bulk compositions and must therefore have formed under varying conditions if they condensed from a solar composition gas.

Figure 15.

 Compositions of the structural components of individual FoBs projected from spinel onto the ternary plane forsterite-gehlenite-anorthite (after Stolper 1982). Arrows extend from the core through the bulk to the mantle compositions for each CAI. The wide gray arrows show the separate effect of evaporative loss of Mg and SiO. Note that these arrows do not emanate from any mineral composition, but rather from SiO and Mg, neither of which plots on this diagram. Thus, the effect on any specific CAI will be somewhat different owing to the differing relative effects of Mg and SiO on a particular bulk composition. The core-mantle trends for each CAI are the vector sums corresponding to the actual ratios of Mg and SiO evaporated.

We will discuss the CAI cores first, then the mantles and their relationships to the cores. In the final section, we will return to Fig. 15 and its implications.

Formation of the Forsterite-Rich Cores

An assumption throughout the previous sections of this article is that FoBs solidified from partial or near-complete melts. In part, this follows from many previous studies of these objects, which have consistently noted two common properties suggestive of melt solidification: (1) many FoBs are spheroidal (ALVIN, SJ-101, E60, E64); and (2) some are even vesicular (ALVIN, 1623-5; Davis et al. 1991). Nevertheless, we explore that assumption in detail here with particular respect to the inclusion cores, which must represent the earliest recorded state of the CAIs. The compositional data for the forsterite-rich cores (only) are plotted separately for each CAI in Fig. 16, in the same manner as in Fig. 15. A test for melt crystallization is whether the texturally observed mineralogy and crystallization sequences for each bulk composition (CAI) are consistent with experimentally determined phase equilibria. In this case, we take that bulk composition to be that of the forsterite-rich core for each CAI. On the individual charts of Fig. 16 are shown the predicted crystallization path for each CAI core, indicated by the hollow gray arrows showing the predicted path during cooling. The predicted crystallization paths lead away from the actual composition of FoB olivine, which (as noted earlier) has a significant calcium content and thus plots slightly below and to the left of pure forsterite at bottom right on the charts.

Figure 16.

 Olivine-rich core compositions of FoBs projected from spinel (as in Fig. 15) onto the ternary plane forsterite–gehlenite–anorthite. For each core composition, the thick gray arrow shows the expected liquid line of descent during igneous crystallization. Point “A” is a distributary reaction point at which anorthite + forsterite react with the melt to form pyroxene. Point “B” is a tributary reaction point at which forsterite reacts with the melt to form pyroxene + melilite. Note that the composition of FoB olivine (small circle at lower right) does not plot exactly at the forsterite apex, due to the presence of significant CaO.

All CAI cores do plot in the forsterite + spinel + liquid stability field, which is consistent with the observation that the textures of all the CAI cores except SJ-101 are dominated by large and generally euhedral olivine grains accompanied by small spinels. In the case of SJ-101, only the olivine-rich bands contain any olivine at all, so this texture is confined to them. The diagrams do not distinguish which CAIs might have spinel as the liquidus phase as opposed to forsterite, but in either case the forsterite will be the dominant textural phase.

Four of the inclusions—E64, ALVIN, SJ-101, and TS35-F1—have crystallization paths that lead to the anorthite field. Anorthite ought to be third-crystallizing phase, followed by pyroxene as the fourth phase. But E64, ALVIN, and TS35-F1 contain no anorthite in their cores. There are several possible reasons for this seeming discrepancy. One is the well-known difficulty of anorthite nucleation (e.g., Grove and Bence 1979). Another is that for E64 and ALVIN, the expected crystallization paths intersect the anorthite phase field only very briefly prior to encountering the pyroxene field. A minor error in determining the core bulk compositions could easily account for the discrepancy. But a more fundamental reason is that the invariant point marked “A” on the diagrams is a distributary reaction point. A liquid evolving down the anorthite-forsterite (+spinel) reaction curve will begin reacting with both anorthite and forsterite at constant temperature until one of the two phases is exhausted. For a liquid whose bulk composition lies in the forsterite field (e.g., a FoB core), the first phase to entirely react away will be anorthite. At that point, the liquid evolves down the forsterite-pyroxene (+spinel) curve, crystallizing both phases (+spinel) toward another reaction point at “B.” The essential point is that there is no anorthite discrepancy for these four inclusions: anorthite should not be expected (preserved) in the FoB cores at all except in the event that a crystal becomes totally enclosed within another phase and is thereby isolated from further reaction. SJ-101 is different from the other three because it does contain abundant anorthite in its core, but only in the olivine-free regions of this heterogeneous CAI. Clearly, the degree of melting in SJ-101 was so low that overall melt equilibrium was never approached. Thus, the modal mineralogy of each local region was controlled by the local bulk composition. Following progressively down the liquid-line-of-descent for all four of these CAIs, pyroxene universally encloses both olivine and spinel as expected. All have minor interstitial magnesium-rich melilite, which is consistent for melts that reach invariant point “B” and terminate crystallization there. So, for these four FoBs, the texturally derived crystallization sequences are indeed consistent with melt solidification.

Inclusions E60 and 3137 are different. Their core bulk compositions are such that melilite, not anorthite, should follow olivine and spinel in the crystallizations sequence. Both contain abundant interior melilite as well as pyroxene that poikilitically encloses both spinel and olivine. Based on textures, the relative sequence of melilite and pyroxene is ambiguous, but nevertheless, the expected versus observed crystallization sequences of E60 and 3137 are broadly consistent.

We conclude that the cores of all six of these FoBs are indeed likely to have solidified from melt droplets.

Melt Evaporation and Formation of the Forsterite-Free Mantles

The observation that the core-mantle trends cross-cut phase boundaries (Fig. 15) is another manifestation of the earlier observation that melilite compositions in the cores and mantles tend to lie on opposite sides of the thermal minimum in the gehlenite-åkermanite binary system. Taken together, these two observations virtually dictate that the cores and mantles cannot have crystallized from a single homogeneous melt in either sense: the mantles cannot have crystallized from melts of the core compositions, and the cores cannot have crystallized from melts of the mantle compositions. Put simply, they are incompatible in a closed system—but not in an open system. If we assume that any open system behavior must occur at the CAI margin, then the mantles must be derived from the cores. There are only two possibilities: calcium and aluminum were added (condensation), or magnesium and silicon were lost (evaporation). The two processes are distinguishable, and the evidence suggests that melt evaporation of magnesium and silicon is the more likely here. Condensation at very low (i.e., nebular) pressures proceeds from vapor to solid and therefore the evolutionary trend will follow a path governed by the stoichiometry of each successive condensing phase. Melt evaporation, in contrast, follows a path governed by the stoichiometry of evaporating molecules, not minerals. For CAIs, the dominant volatile species were SiO and Mg (Davis and Richter 2003). The core-mantle trends shown in Fig. 15 mostly do not extrapolate to any mineral composition within the system. Rather, they more closely match the effects of melt evaporation involving varying proportions of SiO and Mg. ALVIN has a trend suggesting evaporation of primarily Mg. SJ-101 is similar to ALVIN, although the orientation of the core-mantle trend suggests the possibility (see also Petaev and Jacobsen 2009) that some SiO might actually have been added to the mantle. 3137 and E60 have trends suggesting evaporation of dominantly (but not exclusively) SiO, and the other two are intermediate. Davis et al. (1991) studied the FUN CAI Vigarano 1623-5 and, assuming that the core-mantle evolutionary trend defined by that object was due to melt volatilization, showed that the slope of the evolution vector is purely a function of the evaporating ratio Mg/Si. Our study supports that conclusion, but also greatly expands on it by demonstrating that different CAIs evaporated different proportions of SiO to Mg. And indeed, Richter et al. (2002) showed experimentally that the relative evaporation rates of silicon and magnesium from a CAI melt are controlled by bulk composition, fO2, and the partial pressure of H2 in the gas. Wark et al. (1987) proposed an evaporative origin for the mantle on one FoB. Our work makes the more general case for all FoBs.

In all cases shown in Fig. 15, the effect of the evaporation was to drive the composition of the residual melt out of the forsterite primary phase volume and into either the anorthite field (ALVIN, SJ-101, TS35-F1, E64) or the melilite field (3137, E60). Nevertheless, once the melt crossed out of the forsterite field, olivine became unstable and reacted with the residual melt to form anorthite or melilite. This is the reason that the mantles are forsterite-free.

While molten, each FoB underwent (as shown above) evaporative loss of Mg and (except perhaps SJ-101) SiO from the surface of the inclusion, resulting in the formation of a forsterite-free mantle surrounding a forsterite-rich core. The existence of the highly aluminous outer mantles and less aluminous but still forsterite-free inner mantles demonstrates that radial concentration gradients were established, persisted, and finally frozen in upon melt solidification. If the CAIs were totally melted, then evaporation from the surface proceeded more rapidly than did melt diffusion and equilibration. Alternatively, the outermost mantles represent a separate and later stage of surficial melting and evaporation. In either case, with greater degrees of melting and evaporation (either a prolonged period, or multiple episodes), both the inner and outer mantle layers increased in thickness. SJ-101 was subjected only to minimal melting, resulting in little evaporation and only a very thin (<100 μm) forsterite-free mantle (Fig. 2). In ALVIN and E64, evaporation may have been intense (the outermost mantles are very aluminous) but also must have been brief because the forsterite-free mantles are quite thin—generally <300 μm (Figs. 1b and c). E60 represents the next stage of melt evolution, with a forsterite-free mantle that is up to 0.5 mm thick (Fig. 5). The inclusions that have undergone the most extensive evaporative loss of Mg and SiO—3137 and TS35-F1—have forsterite-free mantles that are over a millimeter thick (Figs. 1e and f).

The compositions of individual minerals—melilite, olivine, and pyroxene—also reflect the evaporative history of each FoB. Melilite is a solid solution between two end-members: åkermanite (Ca2MgSi2O7) and gehlenite (Ca2Al2SiO7). The range of melilite compositions seen within a given inclusion generally correlates with the thickness of the forsterite-free mantle present in the inclusion (Fig. 8). With progressive evaporation, and loss of magnesium, melilite trends toward more gehlenitic compositions (i.e., becomes more Al-rich). Melilite composition also correlates with location within the inclusion. Melilite in the cores is very åkermanitic, while that found in the outermost mantles is more gehlenitic. In SJ-101, which has a very thin forsterite-free mantle, there is no melilite in the mantle and that in the core is uniformly magnesium-rich, Åk80–90. ALVIN, which has a slightly thicker mantle, shows a wider compositional range (Åk66–90), again with a peak at Åk80–90. At the opposite end of the scale, TS35-F1 contains melilite with a range of Åk6–88, with no peak at approximately Åk80–90, and with the most åkermanitic melilite preserved only within the very core of the inclusion. The exception to this trend is E64. The melilite composition of E64 shows a range from Åk10–90, consistent with the values we would expect for an inclusion with an olivine-free mantle; however, this inclusion does not show a clear peak at approximately Åk86, and instead displays a range more similar to that of TS35-F1. This can be accounted for by studying the backscattered electron images of the E64 (Figs. 1d and 4). Although there are numerous small grains of åkermanitic melilite, these grains are simply too cracked and porous to be able to obtain reliable electron probe data. Analyses from these grains result in low totals, and so were rejected. In contrast, E60 contains many clean grains of melilite in the core, which allowed high-quality analyses of each grain and a resulting well-defined histogram peak at peak at Åk80–90.

Pyroxene grains within the FoBs show extensive variation in their Al2O3/TiO2 ratios, and are generally less refractory than pyroxene in Type B CAIs (Fig. 10). As with melilite, there is a correlation between pyroxene composition and the thickness of the forsterite-free mantle. In SJ-101, pyroxene is poor in TiO2, and shows a clear bimodal distribution. With increasing thickness of the olivine-free mantle, pyroxene shows a more continuous range in composition, and also extends to higher abundances of Al2O3 and TiO2 content, probably as a result of magnesium loss from the inclusion.

Olivine grains within each FoB are forsteritic (Fo93–100), and consistently contain 0.7–1.9 wt% CaO, with most grains containing just over 1 wt% (Fig. 9). There is no difference between olivine compositions in different FoBs, indicating that all grains crystallized from similarly CaO-rich melts. Elevated CaO contents in forsterite also characterize other FoBs (Clayton et al. 1984; Wark et al. 1987; Davis et al. 1991).

The oxygen isotopic data for five of the six FoBs measured in this work show small degrees of mass-dependent fractionation (Fig. 14), broadly consistent with melt volatilization (e.g., Richter et al. 2002; MacPherson et al. 2004). In each inclusion, spinel, pyroxene, and olivine grains are uniformly 16O-rich, and show up to approximately 5‰ variation in δ18O. However, the degree of mass-dependent fractionation does not in general correlate with the thickness of the pyroxene mantle (and therefore the amount of melt volatilization experienced by an inclusion). Neither is there any correlation with degree of mass fractionation and core versus mantle lithology: there are no consistent isotopic differences between mantle and core. The degree of mass-dependent fractionation observed in TS35-F1 is comparable to that seen in ALVIN. Because the petrologic features described herein virtually demand that melt evaporation did occur, we have to conclude that the isotopic compositions of the evaporating melts mostly remained equilibrated with that of the surrounding gas and within the melt droplets themselves. SJ-101 shows almost no mass-dependent fractionation of oxygen (or magnesium or silicon; Petaev and Jacobsen 2009; Petaev et al. 2010), which fits with the petrographic evidence for it not having undergone sustained melting and volatilization. Measurement of magnesium and silicon isotopic fractionation in the other FoBs will be the subject of future work.

FoB Precursors and the Vigarano 1623-5 Puzzle

We have argued above that the complex internal structure of the inclusion SJ-101 strongly suggests that its precursor consisted of fine-grained olivine-spinel-pyroxene-rich CAIs with a banded structure: spinel-pyroxene cores mantled by abundant forsterite. Specifically, one group of primitive refractory inclusions that have the physical and chemical characteristics to be FoB precursors are amoeboid olivine aggregates. AOAs consist mostly of fine-grained forsteritic olivine, and have numerous small banded and nodular refractory nodules (typically spinel- and pyroxene-rich) distributed within them. A typical AOA that experienced partial melting to form a coherent whole inclusion would look exactly like SJ-101. Supporting evidence for such a model comes from the 16O-rich composition of the core forsterite. 16O-rich forsterite is not common in chondritic meteorites, but two places where it is ubiquitous are AOAs (Krot et al. 2004) and forsterite-rich accretionary rims around CAIs (Krot et al. 2001a; Cosarinsky et al. 2008). Indeed, other than mode of occurrence, there is no clear difference between forsterite-rich accretionary rims and AOAs. In both cases, forsterite encloses refractory nodules consisting mostly of spinel and pyroxene. We infer that SJ-101, and ultimately all FoBs, began as AOA-like objects in which nodular CAIs were enclosed within thick mantles of forsterite.

The FoB inclusion 1623-5 from Vigarano is a bona fide FUN CAI, showing not only endemic nuclear anomalies (Loss et al. 1994) but also extreme degrees of mass-dependent isotopic fractionation in magnesium, silicon, and oxygen (Davis et al. 1991; Marin-Carbonne et al. 2012). Although the mantle lithology tends to have a greater degree of mass fractionation, even the interior olivines are fractionated (Marin-Carbonne et al. 2012). The latter authors interpreted this to mean that even the starting bulk composition of 1623-5 was itself the result of melt evaporation. This interpretation in turn requires that the starting composition of 1623-5 was even more forsterite-rich than what is now observed (e.g., Mendybaev et al. 2009). Thus, 1623-5 would seem to present a puzzle in light of the present work: none of the CAIs in our work show anything like the degree of mass-dependent isotopic fractionation observed in 1623-5, yet their core compositions are broadly similar, albeit somewhat more refractory. Moreover, the inclusion SJ-101 demonstrably experienced only minor partial melting, and thus can hardly be considered as the product of extreme melt evaporation.

Nevertheless, Mendybaev et al. (2009) demonstrated that 1623-5 plausibly evolved—both chemically and isotopically—from a starting material that was broadly similar to many amoeboid olivine aggregates. Overall, our FoBs are somewhat more refractory in composition (higher Al2O3, CaO) than the putative precursor of 1623-5. Thus, if 1623-5 is in fact related to “normal” FoBs and did not arise from some completely different precursor, then (1) 1623-5 formed from a much less refractory (more olivine-rich) AOA than did the FoBs we studied and (2) it experienced more intense and prolonged melt evaporation than did our FoBs, possibly under different conditions.


The forsterite-bearing Type B CAIs are a rare class of inclusion, that show a clear evolution from aggregated objects through to examples that have undergone extensive melting and evaporation. This evolution is reflected not just in the petrography of the inclusion, but also in the mineral chemistry and isotopic signature of each inclusion. The most primitive (least melted) FoBs, such as SJ-101, have composite textures inherited from that of their precursors and have undergone essentially no evaporative loss. More extensively melted examples underwent moderate evaporation and developed thin but clearly defined forsterite-free mantles surrounding a core that still contains abundant forsterite. Finally, some FoBs underwent extensive evaporation (e.g., E60, 3137, and TS35-F1), resulting in the creation of a thick forsterite-free mantle within which the aluminum-rich mineralogy (e.g., gehlenite-rich melilite) formed out of a melt that was incompatible with that from which the core minerals formed.

A small degree of mass-dependent fractionation of oxygen isotopes, consistent with loss during evaporation, is seen in every inclusion except SJ-101. There is no correlation between the amount of evaporation and the degree of mass-dependent fractionation, however, indicating partial re-equilibration of oxygen isotopes after the volatilization event.

The isotopic differences between the CAIs studied herein and the FUN FoB 1623-5 are puzzling, yet are explainable provided that 1623-5 started out with a more olivine-rich bulk composition and experienced more prolonged melt evaporation. All FoBs have in common the fact that they are case studies in the effects of melt evaporation on CAI melts.

Acknowledgments— This work was supported by NASA grants NNX11AD43G and NNX07AJ05G (GJM), NNX10AH76G (ANK), NNX08AH79G (MP), and NNX11AK82G (SBJ). We thank Drs. Andrew Davis, Kevin McKeegan, and Frank Richter for many useful discussions that helped us to understand melt evaporation and “the Vigarano 1623-5 problem.” Finally, we are indebted to Drs. Steve Simon, Shoichi Itoh, and Associate Editor Ed Scott for helpful and constructive reviews that greatly improved the manuscript.

Editorial Handling— Dr. Edward Scott