A new model for the origin of Type-B and Fluffy Type-A CAIs: Analogies to remelted compound chondrules


E-mail: aerubin@ucla.edu


Abstract– In the scenario developed here, most types of calcium-aluminum-rich inclusions (CAIs) formed near the Sun where they developed Wark-Lovering rims before being transported by aerodynamic forces throughout the nebula. The amount of ambient dust in the nebula varied with heliocentric distance, peaking in the CV–CK formation location. Literature data show that accretionary rims (which occur outside the Wark-Lovering rims) around CAIs contain substantial 16O-rich forsterite, suggesting that, at this time, the ambient dust in the nebula consisted largely of 16O-rich forsterite. Individual sub-millimeter-size Compact Type-A CAIs (each surrounded by a Wark-Lovering rim) collided in the CV–CK region and stuck together (in a manner similar to that of sibling compound chondrules); the CTAs were mixed with small amounts of 16O-rich mafic dust and formed centimeter-size compound objects (large Fluffy Type-A CAIs) after experiencing minor melting. In contrast to other types of CAIs, centimeter-size Type-B CAIs formed directly in the CV–CK region after gehlenite-rich Compact Type-A CAIs collided and stuck together, incorporated significant amounts of 16O-rich forsteritic dust (on the order of 10–15%) and probably some anorthite, and experienced extensive melting and partial evaporation. (Enveloping compound chondrules formed in an analogous manner.) In those cases where appreciably higher amounts of 16O-rich forsterite (on the order of 25%) (and perhaps minor anorthite and pyroxene) were incorporated into compound Type-A objects prior to melting, centimeter-size forsterite-bearing Type-B CAIs (B3 inclusions) were produced. Type-B1 inclusions formed from B2 inclusions that collided with and stuck to melilite-rich Compact Type-A CAIs and experienced high-temperature processing.


Type-B calcium-aluminum-rich inclusions (CAIs) are multi-millimeter- to multi-centimeter-size coarse-grained spheroidal igneous objects that consist of melilite, fassaite (Al- and Ti-rich Ca-pyroxene), spinel, and minor anorthite (Grossman 1975). Thermodynamic calculations predict that these same minerals formed during the beginning stages of equilibrium condensation from a hot and cooling gas of solar composition (Grossman 1972; Yoneda and Grossman 1995), leading to the idea that the inclusions themselves were condensates. Whatever their origin, it is clear that Type-B CAIs formed very early in solar-system history––about 4.568 Ga ago (e.g., Amelin et al. 2002; Bouvier and Wadhwa 2009; Bouvier et al. 2010), perhaps within an interval as short as 20–50 ka (Bizzarro et al. 2004; Jacobsen et al. 2008).

There are three principal varieties of Type-B CAIs (Wark and Lovering 1982; Wark et al. 1987): Type-B1 inclusions consist of a melilite mantle surrounding a core of pyroxene, spinel, anorthite, metallic Fe-Ni, and noble metal nuggets (e.g., Fig. 10.3.5a of MacPherson et al. 1988). Type-B2 inclusions have no melilite mantle and are mineralogically similar to B1 cores (e.g., Fig. 10.3.5b of MacPherson et al. 1988). Type-B3 inclusions (the least-common variety; e.g., Figs. 1–3 of Wark et al. 1987) contain coarse forsterite grains along with pyroxene, spinel, and melilite; some also include minor anorthite (e.g., Petaev and Jacobsen 2009). These objects are also known as forsterite-bearing Type-B inclusions or FoB CAIs.

Type-B CAIs differ from other CAI types in two principal respects (1) Type-B inclusions are appreciably larger than most CAI types (which typically are of sub-millimeter size) and (2) they occur exclusively in CV chondrites. In this study, I develop a model for the origin of Type-B CAIs that links both these characteristics. I suggest that Type-B CAIs formed directly in the CV region of the nebula (long before chondrule formation) after Compact Type-A CAIs (CTAs) collided, stuck together, incorporated 16O-rich forsteritic dust (and probably some anorthite), and underwent significant melting and partial evaporation. I also suggest that centimeter-size Fluffy Type-A CAIs (FTAs) in CV chondrites formed in a similar manner––by the collision of numerous Compact Type-A CAIs, acquisition of small amounts of 16O-rich mafic dust, and minor melting.


Formation Location of Most Types of CAIs

CAIs in individual chondrites can have variable bulk O-isotopic compositions (e.g., Clayton et al. 1977), but analyses of individual CAI minerals that are moderately resistant to alteration (e.g., spinel, diopside, hibonite) indicate that the most 16O-rich grains in different chondrite groups (OC, EH, CV, CR, CM, CO) tend to have similar Δ17O values (i.e., −19 to −30‰) (e.g., Figs. 24–27 of MacPherson 2005). They resemble the Δ17O value for the solar wind: −29‰ (McKeegan et al. 2011). CAI populations in different chondrite groups tend to plot on or close to the CCAM (carbonaceous-chondrite-anhydrous minerals) line on the standard three-O-isotope graph (Figs. 24–27 of MacPherson 2005). These data suggest that CAIs from different chondrite groups formed in close proximity within the same O-isotopic reservoir.

Excesses in 10B have been identified in Type-A CAIs, Type-B CAIs, and one FUN inclusion in CV3 chondrites (McKeegan et al. 2000; Sugiura et al. 2001; MacPherson et al. 2003; Chaussidon et al. 2006) as well as in hibonite platy-crystals from CM2 Murchison (Liu et al. 2009). The 10B excesses correlate with the Be/B ratio of the samples and result from the presence of live 10Be in the precursors of these objects. The 10Be (t½ = 1.5 Ma) subsequently decayed to radiogenic 10B.

Beryllium-10 is not produced by stellar nucleosynthesis. It is produced mainly from 16O in CAI precursors by nuclear spallation reactions, either those engendered by galactic cosmic rays (GCRs) interacting with material in the interstellar medium (Reeves et al. 1970) or those caused by intense radiation in the solar nebula (McKeegan et al. 2000; Marhas et al. 2002; Gounelle et al. 2001, 2006; Chaussidon et al. 2006). Although Desch et al. (2004) concluded that 10Be in CAIs was produced entirely by GCRs, this may not be correct. One Type-B CAI from CV3 Allende (CAI 3529-41) appears to have once contained live 7Be as well as 10Be (Chaussidon et al. 2006). Because 7Be has such a short half-life (t½ = 53 days), this radioisotope must have formed in the nebula at almost exactly the same time as the precursor of the CAI containing it. Furthermore, by analogy to X-ray flaring observed in some young stellar objects (Feigelson et al. 2002; Glassgold et al. 2005) that presumably can produce sufficient radiation to form 10Be from 16O, it is plausible that the early Sun behaved similarly. It thus seems reasonable that the majority of CAIs were produced relatively close to the Sun.

After the CAIs formed, they typically developed Wark-Lovering rims. These rims are a few tens of micrometers in thickness and consist of a series of mono- or bi-mineralic layers of melilite, spinel, and pyroxene along with hibonite and perovskite (Wark and Lovering 1977). The rims appear to have formed by a multistage process that included flash heating of the CAI surface (Wark and Boynton 2001). This model is supported by the observation that spinel and pyroxene in the Wark-Lovering rims have approximately the same level of 16O enrichment as the underlying CAI (Wark and Boynton 2001). Refractory minerals in the Wark-Lovering rims tend to have 16O-rich compositions (e.g., Yoshitake et al. 2005). (However, in some cases, melilite in the rims is poorer in 16O, presumably as a result of parent-body alteration [to which melilite is particularly susceptible; e.g., Nomura and Miyamoto 1998; Wasson et al. 2001]. Such alteration may also be responsible for variations in the V content of pyroxene in the rim around an FTA inclusion from Allende [Zega et al. 2010].)

CAIs are present in every chondrite group. Although it was long assumed that CAIs are absent from CI chondrites, Frank et al. (2011) reported the presence of a single (somewhat altered) CAI in CI Ivuna. In addition, there are a few small 16O-rich grains of hibonite and spinel in CI chondrites that were probably derived from fragmented and altered CAIs (Huss et al. 1995). Russell (1998) and Rout and Bischoff (2008) reported that the modal abundance of CAIs deceases in the order: carbonaceous chondrites > R chondrites > ordinary chondrites (OC) > enstatite chondrites.

The relative formation locations of different chondrite groups were inferred by Wasson (1988) and Rubin (2010, 2011) on the basis of isotopic compositions, fall statistics, and intergroup relationships. The enstatite chondrites probably formed in gross proximity to the Earth, probably within the orbit of Mars (1.5 AU). The OC, which constitute 74% of observed falls (Grady 2000), must be located at a major resonance, plausibly the 3:1 resonance with Jupiter located near 2.5 AU (Wasson 1985, 1988). Carbonaceous chondrites, with their higher abundances of low-temperature phases (e.g., organic material), formed farther from the Sun. R chondrites share similarities with both ordinary and carbonaceous chondrites (e.g., Kallemeyn et al. 1996) and may have formed at intermediate distances. Thus, with increasing heliocentric distance, the chondrite groups may be ordered as EH-EL, H-L-LL, R, CR, CV–CK, CM-CO, CI. Although there is evidence for mixing among chondrite groups (most spectacularly in the Kaidun, Bencubbin, and Almahata Sitta breccias; e.g., Zolensky and Ivanov 2003; Weisberg et al. 1990; Bischoff et al. 2010), this mixing is largely confined to later epochs of solar-system history and is not considered here.

The chondrite groups that formed in the inner part of the nebula have few CAIs (EH and EL: approximately 0.01 vol%; OC: approximately 0.02 vol%), R chondrites have somewhat more (approximately 0.04 vol%), and carbonaceous chondrites have the most; the CAI abundance peaks in the CV–CK region and diminishes with increasing heliocentric distance—CR: 0.6 vol%; CV: 3.0 vol%; CK: 4 vol%; CM: 1.2 vol%; CO: 1.0 vol%; CI: 0.0 vol% (Table 2 of Rubin 2011). In addition, some CAI minerals including fassaite, spinel, anorthite, perovskite, and melilite were found among Stardust samples derived from comet 81P/Wild 2 (Zolensky et al. 2006; Simon et al. 2008). Like all Jupiter-family comets, Wild 2 is thought to have formed beyond the orbit of Neptune (e.g., Fernández 1980; Duncan et al. 1988; Quinn et al. 1990; Duncan and Levison 1997).

The widespread distribution of CAIs suggests that after these objects formed near the Sun and developed Wark-Lovering rims, they were transported throughout the nebula by mechanisms that may have included jets, turbulent diffusion, gravitational instabilities, and advection due to viscous expansion of the disk (e.g., Gounelle et al. 2001; Shu et al. 2001; Cuzzi and Hogan 2003; Cuzzi et al. 2003, 2005; Weidenschilling 2003; Boss 2004; Wood 2004; Cuzzi and Weidenschilling 2006; Ciesla 2009, 2010; Jacquet et al. 2011).

In contrast to the conclusion that the majority of CAIs formed in one restricted region of the nebula, it seems likely that chondrules were produced throughout the nebula in different local environments. Chondrules in different chondrite groups have distinct O-isotopic compositions (e.g., Fig. 1 of Rubin 2000; Fig. 10a of Jones et al. 2005). For example, the mean Δ17O values for separated chondrules in the least-equilibrated OC, EH, and CR chondrites are +0.98 ± 0.36‰ (n = 17), +0.07 ± 0.25‰ (n = 10), and −1.40 ± 0.39‰ (n = 9), respectively (Clayton and Mayeda 1985; Clayton et al. 1991; Weisberg et al. 1993). Chondrules in different chondrite groups also differ in mean size, proportions of textural types, percentage of rimmed individuals, and percentage of compound pairs (e.g., Grossman et al. 1988; Rubin 2000, 2010).

Based on the mean sizes of chondrules, the proportion of chondrules with igneous rims, the average thicknesses of igneous rims, the proportions of enveloping compound chondrules, and the abundance of radial pyroxene plus cryptocrystalline (RP + C) chondrules, Rubin (2010, 2011) assigned relative amounts of dust to the nebular regions where the different chondrite groups formed (Fig. 1). Little dust was present in the inner regions of the nebula where enstatite and ordinary chondrites accreted; more dust occurred in the R-chondrite and carbonaceous-chondrite formation locations, peaking in the CV–CK region. The amount of nebular dust gradually declined at greater heliocentric distances where the CM and CO chondrites formed. Variations in dust abundance presumably resulted from aerodynamic forces in the nebula (e.g., Cuzzi and Weidenschilling 2006). If the dust abundance is directly proportional to the amount of fine-grained matrix material in different chondrite groups (cf. Table 2 of Rubin 2011), we can infer that nebular regions differed in their abundance of dust by factors of 3–8.

Figure 1.

 Schematic diagram illustrating the inferred relative amounts of nebular dust at different heliocentric distances in the nebula at the time of CAI formation. The different chondrite groups are shown with their relative formation locations.

Formation of Centimeter-Size Fluffy Type-A CAIs in CV Chondrites

Although Type-A CAIs in most carbonaceous-chondrite groups are typically millimeter-size or smaller, CV chondrites contain some Fluffy Type-A CAIs (i.e., FTAs) that are about 0.3–3 cm in size (MacPherson and Grossman 1984; MacPherson 2005). Most of the centimeter-size FTAs are aggregates of numerous melilite-rich nodules (e.g., Fig. 1 of MacPherson and Grossman 1984; Fig. 8 of MacPherson 2005); as suggested by MacPherson (2005), many of the nodules closely resemble individual, Compact Type-A CAIs (i.e., CTAs) each of which has a Wark-Lovering rim (e.g., Fig. 1 of Simon et al. 1999). In many cases, the nodules in the aggregates are separated by small amounts of fine-grained matrix-like material (MacPherson and Grossman 1984).

I suggest that the individual nodules were independent CTAs (and perhaps a few other CAI types that contained hibonite) that formed near the Sun and were then transported to the CV region. The individual CTAs were subsequently mantled by small amounts of 16O-rich nebular dust (mainly mafic silicate) or were incorporated into small 16O-rich, mafic-silicate-rich dustballs. The CTAs collided with each other at low relative velocity and stuck together (in a manner analogous to the formation of sibling compound chondrules; Wasson et al. 1995). Still early in solar-system history, the aggregated CTAs experienced minor remelting; they were never completely molten (MacPherson and Grossman 1984; Beckett and Stolper 1994). This scenario can account for the recrystallized textures of the nodules (MacPherson 2005) and the deformed shapes of the nodules’ rims. Whether the nodules in individual FTAs are separate or connected in three dimensions is probably a function of the degree of (minor) remelting that the particular FTAs experienced. The occurrence of 16O-rich and 16O-poor melilite in a single FTA in Vigarano suggests that it was subject to multiple melting events (Harazono and Yurimoto 2003), roughly analogous to 16O-poor FeO-rich porphyritic olivine chondrules that contain relict 16O-rich forsteritic olivine grains (e.g., Jones et al. 2000; Kunihiro et al. 2004, 2005).

A potential objection to this model is that some FTAs in Allende have reversely zoned melilite grains (MacPherson and Grossman 1984), whereas most CTAs do not (Simon et al. 1999). Reverse zoning could potentially develop if the FTAs were experiencing evaporation of Mg. Alternatively, it seems possible that the significant aqueous alteration experienced by Allende (e.g., Kojima et al. 1995; Krot et al. 1995, 1998) is responsible for reverse zoning in FTA melilite; many of the inclusions with reversely zoned melilite are highly altered (MacPherson and Grossman 1984). Åkermanite-rich melilite apparently can form along with pyroxene from gehlenitic melilite by reaction with perovskite (Fig. 4 of MacPherson and Grossman 1984). Melilite is particularly susceptible to alteration (Nomura and Miyamoto 1998) and fractures in the melilite grains could provide conduits for water. Vigarano, which is appreciably less altered than Allende (e.g., Weisberg et al. 1997; Krot et al. 1998), contains FTAs that were not noted to have reversely zoned melilite (MacPherson 1985).

The higher hibonite content of FTAs relative to that of many CTAs suggests that the CTAs may have collided with a few hibonite-bearing CAIs to form the FTA precursors.

If nebular dust became poorer in 16O with time, this model would be consistent with the inference that FTAs underwent multiple heating episodes in the nebula resulting in, for example, the coexistence of 16O-rich and 16O-poor melilite in an FTA in CV3 Vigarano (Harazono and Yurimoto 2003). Late remelting of some CAIs in 16O-poor environments is consistent with the data of Aléon et al. (2002) who found that most CAIs in CR chondrites have 16O-rich compositions, but those that exhibit petrographic evidence of melting are appreciably poorer in 16O. This scenario is also consistent with the occurrence in CV chondrites of compound chondrule-CAI objects that lack excesses in 26Mg and have relatively 16O-poor compositions (e.g., Krot et al. 2007).

The cooling rates inferred for Type-B CAIs that contain melilite phenocrysts with reverse zoning are 0.5–50 K hr−1 (MacPherson et al. 1984), too slow to have been caused by direct radiative cooling into space. These inclusions may have been embedded in large dustballs.

If this scenario is correct, then the large FTAs in CV chondrites can be considered compound CAIs that are broadly analogous to adhering sibling compound chondrules (e.g., Wasson et al. 1995). They are not products formed directly in the nebula by gas-solid condensation during a period of decreasing pressure as proposed by MacPherson and Grossman (1984).

Formation of Type-B CAIs

Because accretionary rims around CAIs contain abundant 16O-rich forsterite (Cosarinsky et al. 2008), nebular dust was probably forsterite-rich (and 16O-rich) at the time and place these rims were formed. This inference leads to a model similar to that of MacPherson and Huss (2000) who suggested that Type-B CAIs formed from altered Type-A CAIs after they acquired olivine.

I suggest that, in contrast to other CAI types (which formed near the Sun), Type-B CAIs formed directly in the CV region of the nebula by a three-step process (1) the clumping of numerous CTAs (the same basic process that formed the centimeter-size FTAs), (2) the incorporation of significant amounts of 16O-rich mafic dust (mainly forsterite) into the assemblage (somewhat analogous to the formation of enveloping compound chondrules; Wasson et al. 1995), and (3) extensive processing at high temperatures. (Note that the model calls for the dust in the CV region to have been rich in 16O during this period, long before chondrule formation.) During heating, Type-B CAIs were melted and experienced some evaporation, depleting them in bulk MgO and SiO2 (Richter et al. 2006) and causing enrichment in the heavy isotopes of Mg and Si (Clayton et al. 1988).

Mass-balance calculations (wt%) using published CAI bulk compositions (Grossman et al. 2000, 2008) show for illustrative purposes that a mixture of 87% mean Type-A CAIs and 13% forsterite yields a composition similar to that of mean Type-B CAIs (Table 1). Slightly closer to the mean Type-B CAI composition is a mixture of 80% mean Type-A CAIs, 10% forsterite, and 10% anorthite (Table 1). The bulk compositions of these hypothetical mixtures are well within the range of Type-B CAIs for all major elements.

Table 1.   Mass-balance calculations of CAI compositions (wt%) regarding the transformation of mean Type-A CAIs into mean Type-B CAIs.
Mean Type-A CAIEndmember forsteriteEndmember anorthiteMean Type-B CAIMixture:
87% col. 1, 13% col. 2
80% col. 1, 10% col. 2, 10% col. 3
  1. Mean CAI compositions calculated from data in Grossman et al. (2000, 2008).


In those cases where greater amounts of forsterite were incorporated into centimeter-size aggregates of Type-A CAIs prior to melting, a centimeter-size forsterite-bearing Type-B CAI (i.e., a B3 inclusion) could have been produced (cf. MacPherson and Huss 2000). For purposes of illustration, a hypothetical mixture of 76% of a particular Compact Type-A CAI from CV3 Leoville (inclusion 3536-2; Grossman et al. 2000) and 24% forsterite would yield a composition similar to that of a particular Type-B3 CAI from CV3 Efremovka (inclusion E60; Grossman et al. 2008) (Table 2). Even closer to the composition of inclusion E60 is a hypothetical mixture of 74% of inclusion 3536-2, 24% forsterite, and 2% anorthite (Table 2). This mixing model is compatible with that of Bullock et al. (2011) who proposed that forsterite-bearing Type-B CAIs formed from sintered, refractory-mineral-bearing “proto-CAIs” that acquired forsterite-rich rims prior to being melted by intense flash-heating events. In some cases, melting may have been rather limited, leading to the preservation of precursor materials (e.g., Petaev and Jacobsen 2009).

Table 2.   Mass-balance calculations of CAI compositions (wt%) regarding the transformation of a particular compact Type-A CAI into a particular Type-B3 CAI.
Compact Type-A CAIEndmember forsteriteEndmember anorthiteType-B3 CAIMixture:
76% col. 1, 24% col. 2
74% col. 1, 24% col. 2, 2% col. 3
  1. 1 = inclusion 3536-2 from CV3 Leoville (Grossman et al. 2000); 4 = forsterite-bearing Type-B inclusion E60 from CV3 Efremovka (Grossman et al. 2008).


Because some accretionary rims around CAIs in Allende contain Ti-, and Al-bearing Ca-pyroxene (Allen et al. 1978; MacPherson et al. 1985), it seems likely that some of this pyroxene was included along with forsterite (or forsterite plus anorthite) in the dust that mixed with the aggregates of CTAs prior to melting. (This is also consistent with the occurrence of pyroxene in Type-B3 CAI SJ101; Petaev and Jacobsen 2009.) Addition of forsterite, anorthite, and Ca-pyroxene to pure gehlenite (or particular Type-A CAIs that are very rich in gehlenite) could approximate the phase-diagram positions of Type-B CAIs (Fig. 4 of MacPherson et al. 2005) on the plane Al2O3–Mg2SiO4–Ca2SiO4 projected from spinel (MgAl2O4) (cf. Huss et al. 2001). The other Type-A CAIs on the MacPherson et al. diagram correspond to mixtures of gehlenite and forsterite and could produce Type-B CAIs only with the addition of diopside and (especially) anorthite; Type-A CAIs of these less gehlenitic compositions may not have served as precursors of Type-B CAIs.

The TiO2 content of pyroxene in CTAs is appreciably higher than that in most B1 mantles (e.g., Grossman 1975; Simon et al. 1999). Nevertheless, the bulk TiO2 contents of the two types of inclusions are roughly comparable (i.e., there is no dearth of bulk TiO2 in Type-B inclusions): The mean TiO2 content of CTA pyroxene listed in Table 1 of Simon et al. (1999) is 17.6 wt%; pyroxene constitutes variable amounts of the inclusions ranging from trace abundances to approximately 20% (Simon et al. 1999). If the mean pyroxene abundance is 12 wt%, then, on average, CTAs contain 2.1 wt% TiO2. On the other hand, the mean TiO2 content of Type-B pyroxene listed in Tables 3 and 4 of Grossman (1975) is 7.9 wt%; pyroxene constitutes 35–60% of the Type-B inclusions (Grossman 1975). If the mean pyroxene abundance is 45 wt%, then, on average, Type-B CAIs contain 3.6 wt% TiO2. The comparable amounts of bulk TiO2 in the CTA and Type-B inclusions may indicate that the CTA precursors of Type-B CAIs were partly melted, that many pyroxene nuclei remained in the melt, and that numerous pyroxene grains crystallized; TiO2 was partitioned among the larger number of pyroxene grains in the Type-B CAIs. In addition, a small amount of TiO2 was partitioned into other crystallizing phases such as hibonite (which, in some cases, forms needles projecting inward from the outer parts of the mantle of the B1 inclusions; Blander and Fuchs 1975; Wark and Lovering 1978). Melilite in these Type-B1 CAIs likewise grew inward from the rims.

If substantial evaporation of MgO and SiO2 occurred during the heating of Type-B CAI precursors (Richter et al. 2006), then the mass-balance calculations (Tables 1 and 2) are lower limits on the amount of forsterite that was added to Type-A CAI aggregates to produce Type-B compositions. (Endmember forsterite consists exclusively of MgO and SiO2.)

As reviewer Tim Fagan pointed out, if Type-B CAIs formed by mixing forsterite with Type-A CAIs, then Type-B CAIs should be depleted in bulk Ti relative to Type-A CAIs. Although some Type-A inclusions are richer in Ti than most Type-B inclusions, this is not generally the case: the data of Sylvester et al. (1992, 1993) show that the mean TiO2 content of Type-A CAIs is 1.44 ± 0.69 wt% (n = 6) and that of Type-B CAIs is 1.70 ± 0.51 wt% (n = 8). The values in these small data sets are the same to within one standard deviation. Even if 13% Ti-free forsterite were added to the mean Type-A composition (Table 1), the resultant Type-A TiO2 content would be 1.25 wt%, still within one standard deviation of the Type-B value. Thus, the model is not invalidated by the respective, statistically indistinguishable, average TiO2 contents of Type-A and Type-B CAIs.

There is geochemical support for the model proposed here: (1) The compositions of Type-B CAIs are less refractory than those of Type-A CAIs (MacPherson 2005). This is because mafic dust was incorporated into the Type-B precursors. (2) Type-B CAIs show more evidence than Type-A inclusions of a late disturbance in the 26Al-26Mg and 10Be-10B systematics (e.g., MacPherson et al. 1995, 2003; Kita et al. 2010; Connolly et al. 2011; Mishra et al. 2011). This is plausibly due to aggregates of Type-A CAIs acquiring mafic dust and experiencing remelting. Although both types of inclusions may have experienced remelting, only those that became Type-Bs acquired significant amounts of mafic dust. (As pointed out by Tim Fagan in his review, the evidence for Al-Mg disturbances in Type-B CAIs could be an artifact caused by the much greater abundance of anorthite in Type-B inclusions; anorthite, with its high Al/Mg ratio, is a more sensitive recorder of disturbances than other common CAI phases. The late disturbances in the 10Be-10B systematics are still plausibly due to aggregates of Type-A CAIs experiencing remelting.) (3) A Type-B CAI in Leoville appears to be 30–40 ka younger than the oldest refractory inclusions (Kita et al. 2010); some of these old inclusions are CTAs (e.g., Vigarano F9; MacPherson et al. 2010). The 30–40 ka interval may approximate the timescale for the transportation of CTAs to the CV region of the nebula, collisions with other CTAs at this location, and the time expended there prior to melting. (4) Type-B1 CAIs appear to contain relict phases in melilite (such as perovskite and clinopyroxene) indicating that the inclusions were remelted (Kennedy et al. 1997; Paque et al. 2011). (5) Several Type-B CAIs have petrologic features indicative of late-stage remelting at temperatures below those of the initial melting event (Simon et al. 2005); such features include two sets of melilite laths, relict gehlenite, complexly zoned melilite, and vastly different textures in cores and mantles.

Although other models for the formation of Type-B CAIs include evaporation of MgO and SiO2 from liquid droplets followed by crystallization (e.g., Grossman et al. 2000; Richter et al. 2006), these models do not explain why these inclusions occur exclusively in CV chondrites or why Type-B CAIs show more evidence of late disturbance in their 10Be-10B (and, if not an artifact, their 26Al-26Mg) systematics.

Relationship of Type B1 and B2 CAIs

Wark and Lovering (1982) tabulated the petrologic differences between Type-B1 and -B2 CAIs: B1 inclusions are coarser grained and have more strongly zoned crystals, they have a greater compositional range in melilite and fassaite, and they are richer in bulk Ca and Al. The B1 inclusions consist of a melilite mantle (accounting for their relatively high bulk Ca and Al) and a core of pyroxene, spinel, anorthite, metallic Fe-Ni, and noble metal nuggets; Type B2 inclusions are similar in mineralogy to B1 cores.

It is clear that Type-B1 and -B2 CAIs are closely related. Wark and Lovering (1982) suggested that both subtypes formed as evaporative residues; in their scenario, the B1 inclusions were completely melted and devolatilized at high temperatures, but the B2 inclusions were heated to lower temperatures and experienced lower degrees of evaporation. However, Stolper and Paque (1986) undertook cooling-rate experiments of Type-B CAI compositions and concluded that Type-B CAIs were never completely molten; they inferred that these inclusions were only partly molten and cooled “relatively slowly.”

Richter et al. (2002, 2006) suggested that the occurrence of melilite mantles around Type-B1 inclusions and their absence around B2 inclusions were due to differences in the evaporation rates of Mg and Si from the surfaces of the precursor droplets compared to the diffusion rates of these species within the droplets. Relatively rapid diffusion rates would result in homogeneous droplets and random nucleation of melilite; the end result would be a Type-B2 inclusion. Slower diffusion would cause the outer parts of the droplet to become depleted in Mg and Si via evaporation; melilite would crystallize near the droplet surface and produce a mantle, characteristic of Type-B1 inclusions. Crystallization and evaporation experiments by Mendybaev et al. (2006) confirmed the geochemical viability of this model. A prediction of this model would be that Type-B1 inclusions would tend to be the same size or smaller than B2 inclusions. The observations of Wark and Lovering (1977) appear to show that B1 inclusions are larger than B2 (1–2 cm versus 0.5–1 cm), contrary to the implicit predictions of the Richter et al. (2002, 2006) model. However, Wark and Lovering (1982) later stated that the inclusion sizes reported in the literature are generally too imprecise to determine systematic differences between B1 and B2.

I offer an alternative model. Igneous rims around chondrules (Rubin 1984) and secondary shells around primary chondrules in enveloping compound chondrules (Wasson et al. 1995) formed after individual chondrules acquired dusty mantles (or were incorporated into dustballs). The dusty shells were then melted. Type-B1 inclusions may have formed in an analogous manner: B2 inclusions (which are mineralogically similar to B1 cores) acquired a melilite-rich mantle and were then extensively melted. The additional melilite made the B1 inclusions richer in bulk Ca and Al. Relatively slow cooling allowed the minerals to grow coarser in the B1 inclusions; non-equilibrium crystallization from the melt produced compositional zoning. Differences in cooling rate and degree of melting among different B1 inclusions caused inclusion-to-inclusion variations in mineral composition. Most of the clumps of spinel grains (spinel framboids) and the spinel palisades common in B2 inclusions (Wark and Lovering 1982) were ultimately derived from Wark-Lovering rims around small CAIs and were melted during this process; this can account for their paucity in B1 inclusions. (Spinel framboids occur in 23% of the B1 and 83% of the B2 inclusions studied by Wark and Lovering [1982]; spinel palisades occur in 15% of the B1 and 33% of the B2 inclusions [Wark and Lovering 1982].) Consistent with this interpretation is the occurrence in some palisades of perovskite, a phase also present in many Wark-Lovering rims (Wark and Lovering 1982).

Where could the melilite-rich mantles that surrounded some B2 inclusions have come from? They came from CTAs, which are very rich in melilite; in fact, a few CTAs are composed of nearly monomineralic melilite (Simon et al. 1999). Just as the CTAs are modeled here as having collided and stuck together in the CV region of the nebula to form the centimeter-size FTAs, the CTAs also collided with and stuck to some B2 inclusions (after these latter objects formed in the CV region). The collisions between CTA and B2 inclusions were followed by high-temperature processing of the compound assemblages. This is consistent with the observations of Simon et al. (2005) who showed that significant remelting of Type-B CAIs was common.

Petrographic evidence consistent with this model (and irrelevant to the model of Mendybaev et al. 2006) is the occurrence of a few Type-B CAIs located within some Type-A inclusions (Wark and Lovering 1982). These are Type-B inclusions that collided with Type-A CAIs. The spinel palisades in some Type-A CAIs that enclose Type-B material are remnants of the Wark-Lovering rims around the Type-B CAIs that experienced subsequent thermal processing. For illustrative purposes, I derived a rough approximation of the average melilite composition of the mantle of a Type-B1 CAI (Table 3) by determining the mean of the seven listed melilite analyses from Vigarano inclusion USNM 1632-8 (Table 1 of MacPherson and Davis 1993). As shown in Table 3, this approximate mean melilite composition is quite similar to the bulk composition of a particular CTA (3536-2) from Leoville (Table 1 of Grossman et al. 2000). This indicates that, in principle, the melilite-rich mantles around B1 inclusions could have been derived from CTAs.

Table 3.   Comparison of bulk composition of a Compact Type A CAI in Leoville with mean melilite from a Type B CAI in Vigarano (wt%).
Compact Type-A CAI 3536-2 from
CV3 Leoville
Mean melilite in Type B CAI 1632-8 from
CV3 Vigarano
  1. 1 = data from Table 1 of Grossman et al. (2000); 2 = average of all seven analyses from Table 1 of MacPherson and Davis (1993) normalized to 100 wt% (after round-off error) after subtraction of small amounts of FeO and Na2O.


An implicit prediction of this model is that Type-B1 inclusions would tend to be larger than B2 inclusions. The initial observations of Wark and Lovering (1977) suggest that this is the case: 1–2 cm for B1; 0.5–1 cm for B2; however, the lack of hard data (Wark and Lovering 1982) precludes us from considering the result definitive.

CAIs and Chondrules

CAIs appear to be 1–2 Ma older than chondrules (Huss et al. 2001; Mostefaoui et al. 2001; Amelin et al. 2002; Krot et al. 2005, 2007; Russell et al. 2005; Villeneuve et al. 2009), consistent with the very low initial 87Sr/86Sr ratio in CAIs (Gray et al. 1973; Podosek et al. 1991). The CAIs may have been prevented from drifting into the Sun prior to chondrite accretion by nebular turbulence (e.g., Cuzzi et al. 2003). Alternatively, it is possible that CAIs were trapped in a “dead zone” in the nebula after gravitational instabilities diminished, in regions where the magnetorotational instability was not operating, and the advection velocity was reduced (Jacquet et al. 2011).

At the time CAIs formed, most of the chemical components of chondrules were probably still in the gas phase. Forsteritic dust was generally rich in 16O (e.g., Cosarinsky et al. 2008). As time passed and the nebula cooled, more-oxidized materials condensed (e.g., Grossman and Larimer 1974) and parcels of gas and dust with distinct O-isotopic compositions may have fallen into the nebula from the surrounding molecular cloud (e.g., Wasson 2000). Existing dustballs in the nebula incorporated and reacted with some of this infalling material to become more ferroan and relatively depleted in 16O. Chondrules were produced throughout this period; type-I (low-FeO, relatively 16O-rich) chondrules formed prior to type-II (high-FeO, relatively 16O-poor) chondrules (e.g., Jones et al. 2000; Kunihiro et al. 2004, 2005; Wasson et al. 2004; Kita et al. 2010).

Chondrules were probably produced by flash heating (e.g., Boss 1996; Rubin 2000). This mechanism is consistent with (1) the inferred rapid cooling rates of chondrules (perhaps approximately 5000 K hr−1 at high temperatures; Yu et al. 1996); (2) the rounded, but nonspherical shapes of many chondrules (Rubin and Wasson 2005); and (3) the presence of microchondrule-bearing fine-grained rims around some normal-size chondrules (Krot and Rubin, 1996; Krot et al. 1997).

Flash heating of the CAI surface was involved in producing Wark-Lovering rims (Wark and Boynton 2001); analogous rims were produced around some chondrules. These include the microchondrule-rich rims around some normal-size chondrules in ordinary chondrites; these microchondrules were formed by the flash-melting of pyroxene phenocrysts at the chondrule surface (Fig. 1 of Krot et al. 1997).

Chondrules may have formed at the interface between the dusty midplane and a gas-rich zone above it (e.g., Wasson and Rasmussen 1984). Although there are a number of compound chondrule-CAI objects (e.g., Misawa and Nakamura 1988; Krot et al. 2001, 2005, 2007), their relative paucity might be due to CAIs mainly having settled to the nebular midplane during periods of low nebular turbulence, prior to chondrule formation. Chondrite accretion occurred sometime after chondrules also settled to the midplane.

Unlikelihood of Hot Accretion of CAIs

Because of the ≥1-Ma lag time between the formation of Type-B CAIs and chondrite accretion, it seems highly unlikely that a deformed Type-B CAI from CV3 Leoville (USNM 3537-1) could have accreted from the nebula to the parent asteroid while “still very hot” as suggested by Caillet et al. (1993). It would still be unlikely even if the CAI had been reheated during the epoch of chondrule formation (Rubin and Brearley 1996).

How could Leoville inclusion 3537-1 have been heated and deformed? This inclusion has a length/width aspect ratio of approximately 6 and is aligned along the meteorite’s plane of foliation (Fig. 1 of Caillet et al. 1993). Chondrite petrofabrics are associated with collisional processes on asteroids (Nakamura et al. 1995; Tomeoka et al. 1999; Gattaceca et al. 2005; Rubin and Swindle 2011). During an impact, matrix pores collapse under shock pressure (Scott et al. 1992) causing chondrule and CAI flattening, as these objects squeeze into pores. Inclusion 3537-1 was probably heated and deformed by the same collision that produced the petrofabric in Leoville. Because shock effects are heterogeneously distributed in targets (e.g., Bischoff and Stöffler 1992), it is not unexpected that some of the refractory inclusions in Leoville (e.g., inclusion 3537-2; Caillet et al. 1993) remained relatively undeformed.


Sub–millimeter-size CTAs formed near the Sun early in solar-system history developed Wark-Lovering rims at this location and were then transported throughout the nebula. Some of the CTAs that were deposited in the CV–CK region acquired small mantles of 16O- and forsterite-rich nebular dust and collided with one another (and perhaps a few other CAI types); these objects stuck together and were partly melted to form centimeter-size FTAs. Other aggregates of small gehlenite-rich Type-A CAIs incorporated substantially more forsterite-rich dust (on the order of 10–15%) (and probably comparable amounts of anorthite) and experienced more extensive melting and partial evaporation; they formed centimeter-size Type-B CAIs. A few aggregates of small Type-A CAIs incorporated appreciably higher amounts of forsterite-rich dust (on the order of 25%) (and minor anorthite); upon melting, these objects formed centimeter-size forsterite-bearing Type-B CAIs (B3 inclusions). Some Type-B2 inclusions collided with and stuck to melilite-rich CTAs; after high-temperature processing, these compound objects became Type-B1 CAIs.

Acknowledgments–– I thank K. D. McKeegan and J. T. Wasson for discussions; A. Ruzicka and S. B. Simon for responding to queries; and Tim Fagan, two anonymous reviewers, and associate editor E. R. D. Scott for useful constructive criticism. This work was supported by NASA Cosmochemistry grant NNG06GF95G.

Editorial Handling— Dr. Edward Scott