The origin of chondrules and chondrites: Debris from low-velocity impacts between molten planetesimals?


Corresponding author. E-mail:


Abstract– We investigate the hypothesis that many chondrules are frozen droplets of spray from impact plumes launched when thin-shelled, largely molten planetesimals collided at low speed during accretion. This scenario, here dubbed “splashing,” stems from evidence that such planetesimals, intensely heated by 26Al, were abundant in the protoplanetary disk when chondrules were being formed approximately 2 Myr after calcium-aluminum-rich inclusions (CAIs), and that chondrites, far from sampling the earliest planetesimals, are made from material that accreted later, when 26Al could no longer induce melting. We show how “splashing” is reconcilable with many features of chondrules, including their ages, chemistry, peak temperatures, abundances, sizes, cooling rates, indented shapes, “relict” grains, igneous rims, and metal blebs, and is also reconcilable with features that challenge the conventional view that chondrules are flash-melted dust-clumps, particularly the high concentrations of Na and FeO in chondrules, but also including chondrule diversity, large phenocrysts, macrochondrules, scarcity of dust-clumps, and heating. We speculate that type I (FeO-poor) chondrules come from planetesimals that accreted early in the reduced, partially condensed, hot inner nebula, and that type II (FeO-rich) chondrules come from planetesimals that accreted in a later, or more distal, cool nebular setting where incorporation of water-ice with high Δ17O aided oxidation during heating. We propose that multiple collisions and repeated re-accretion of chondrules and other debris within restricted annular zones gave each chondrite group its distinctive properties, and led to so-called “complementarity” and metal depletion in chondrites. We suggest that differentiated meteorites are numerically rare compared with chondrites because their initially plentiful molten parent bodies were mostly destroyed during chondrule formation.


Chondrules are igneous-textured grains that make up 50% or more by volume of most chondrites. Many of them are frozen droplets of magma, typically 0.1–2 mm across, rounded to lobate in shape, and composed largely of the Mg-rich silicate minerals olivine, (Mg,Fe)2SiO4, and pyroxene, (Mg,Fe)SiO3 (Zanda 2004; Scott and Krot 2007). They tend to be either FeO-poor (type I chondrules) or FeO-rich (type II chondrules). Their igneous textures suggest cooling from near-liquidus temperatures and solidifying over a matter of hours. Evidently, they were present in the solar nebula, or protoplanetary disk (the disk of gas and dust that surrounded the infant Sun, from which the planets later developed) prior to accreting to the surfaces of growing chondrite parent asteroids. They co-accreted with other disk materials including small grains and droplets of Fe-Ni metal and sulfide, refractory objects called calcium-aluminum-rich inclusions (CAIs), mineral fragments, and dust including micron-scale grains of stardust surviving from the presolar molecular cloud. As chondrites account for five out of six meteorites falling to Earth, the formation of chondrules was apparently a process that affected a substantial fraction of the solid material in the solar nebula.

There is no consensus on how chondrules were made: two fundamentally different approaches have come to dominate current discussion. The first contends that clumps of dust in the disk were transformed directly to chondrules by rapid melting, probably as a result of shock-induced heating in the nebula. This idea was suggested half a century ago (Wood 1963, p. 165) in a seminal paper whose title we adopt here, and the melting of dust-clumps, whether by shock or by some other means, has since become the prevailing theory for chondrule formation. It has been advocated, tacitly if not overtly, by many authors, including Taylor et al. (1983), Wood (1988), Grossman (1988), Wasson (1993), Rubin (2000), Shu et al. (2001), Boss and Durisen (2005), Lauretta and McSween (2006), Scott (2007), Alexander et al. (2008), and Ruzicka et al. (2012a).

The second approach imagines that chondrules were produced when large volumes of molten rock became splashed and dispersed into the nebula as showers of spray. Specifically, it has come to embrace a hypothesis, dating from about 30 yr ago, that chondrules originated in great plumes of droplets launched by collisions between planetesimals that had been intensely heated and melted by the decay of 26Al (Zook 1980, 1981; Wänke et al. 1981, 1984). This second hypothesis lay dormant for many years, almost completely overshadowed by the first, and its re-awakening has been slow. Sanders (1996) attempted to revive interest, and the idea has also been promoted by LaTourrette and Wasserburg (1998), Chen et al. (1998), Lugmair and Shukolyukov (2001), and Hevey and Sanders (2006). Sanders and Taylor (2005) reviewed the hypothesis in detail. While the production of chondrules from molten planetesimals has never had a majority following, few today would dismiss it out of hand, and recently, Asphaug et al. (2011) enhanced its credibility with a computer simulation of how chondrules might form from magma released as a result of collisions.

A great deal more is known today about the early solar system than was known 30 yr ago. The past 5–10 yr in particular have witnessed major improvements in mass spectrometry and meteorite chronology, significant developments in the modeling of planetesimal melting and of conditions in the solar nebula, and a substantial increase in the number and variety of meteorites available for study. The same recent period has also seen much exchange of ideas between meteorite researchers, astronomers, and planetary scientists. Observations of disks around young stars, discoveries of extra-solar planetary systems, and computer simulations of orbital dynamics have revolutionized our thinking on the formation of the asteroid belt. Together, these advances are bringing into focus a new picture of the young solar system that is significantly different from the one that has been conventionally portrayed.

In this article, we examine that new picture and how it might be extended to embrace the formation of chondrules and the assembly of chondritic asteroids. We begin by showing how precise meteorite chronology now indicates that planetesimals first melted long before chondrules were made, consistent with 26Al having been an internal planetesimal heat source. We argue that molten planetesimals dominated the population of bodies in the inner solar system for the first 2 Myr, and we propose that the inevitable collisions and mergers between them led to “splashing” with ejecta plumes bearing swarms of chondrule droplets and other debris, which later accreted to chondritic parent bodies. We evaluate the established petrographic, compositional, and experimental evidence for chondrule formation, and find it to be consistent with this collision scenario. Importantly, we find some of the evidence hard to reconcile with the conventional view that chondrules began as dust-clumps. In the last section of the article, we discuss how the “splashing” hypothesis may relate to the broader picture of planetesimal evolution in the young disk, speculating on the nature of the molten precursor planetesimals, and on the origin of some hitherto poorly understood features of chondrites.

A New Paradigm for Chondrites Anchored in 26AL Heating

The Conventional View of Chondrites

Chondrites have conventionally been interpreted as aggregates of primitive materials that were assembled at the very start of the solar system from the same reservoir of nebular dust as went to make the Sun (e.g., Wood 1988). This view stems from the remarkable similarity between the chemical composition of chondrites and that of the Sun’s photosphere for all elements other than a few that normally occur in gases (e.g., H, He, C, N, Ar). The view is reinforced by the presence in chondrites of CAIs, which are the oldest dated objects with a solar system isotopic signature (Amelin et al. 2002, 2010). Indeed, the time of CAI formation has now been widely adopted as defining the start of the solar system (t = 0). In addition, chondrites contain pristine grains of stardust that are even older than the solar system (e.g., Hoppe 2008). Thus, conventional thinking holds that chondrules, along with CAIs, were created directly from clumps of nebular dust at the outset, before the first planetesimals (presumed to be the chondrite parent bodies) had accreted. It further holds that after their accretion, some chondritic planetesimals became overheated, melted, and differentiated into molten metal cores and basaltic crusts, the sources, respectively, of iron and basaltic meteorites (e.g., Lauretta and McSween 2006).

However, this intuitive and long-established interpretation of meteorites has recently been challenged by chronological evidence, which suggests that chondrules were made after, and not before, the parent bodies of differentiated meteorites had melted.

The Chronology of Molten Cores and Chondrules

Most iron meteorites now appear to come from planetesimals that had accreted and melted extremely early, perhaps within 1 Myr of CAI formation (t < 1 Myr) because their ε182W values are very low and within error of the initial ε182W of CAIs (Burkhardt et al. 2008). In the early solar system, ε182W was rising due to the radioactive decay of 182Hf to 182W (half-life 8.9 Myr). Tungsten is a siderophile (iron-loving) element, whereas hafnium is lithophile (silicate-loving). The ε182W values of iron meteorites became fixed, therefore, when molten metal segregated into planetesimal cores. At that stage, the radioactive hafnium, being lithophile, was removed from close proximity to metal and transferred to the silicate mantle of each planetesimal where subsequent radiogenic 182W accumulated. A recent estimate of the time of core formation based on ε182W in magmatic iron meteorites is 0.3 ± 1.2 Myr after CAIs (Kruijer et al. 2011). Even more recently, Burkhardt et al. (2012) revised the value of initial ε182W in CAIs, and they inferred that while the cores of some parent bodies (notably the IVB group) formed at t approximately 0.3 Myr, the cores of others continued to form up to about t = 2 Myr. The timing of core separation is critical to constraining early disk evolution, and is the culmination of painstaking efforts to understand and refine the Hf-W chronometer by many workers (Horan et al. 1998; Kleine et al. 2005; Markowski et al. 2006; Scherstén et al. 2006; Qin et al. 2008; Burkhardt et al. 2008). In particular, Markowski et al. (2006) recognized the need to correct ε182W values for the effects of long-term exposure to cosmic rays; uncorrected values had previously given erroneously old ages.

By contrast, 26Al-26Mg dating suggests that chondrules are mostly 1.5–2.5 Myr younger than CAIs. Following the pioneering work of Hutcheon and Hutchison (1989), about 100 chondrule dates based on 26Al-26Mg internal isochrons have now been published. A recent review of them by Kita and Ushikubo (2012) shows that more than 90% of those from unequilibrated (type 3.0) chondrites (64 determinations) fall within the 1.5–2.5 Myr age range, with just three dating from < 1.5 Myr. The same age difference between chondrules and CAIs of about 2 Myr has been measured independently by 207Pb-206Pb dating (e.g., Amelin et al. 2002, 2010; Connelly et al. 2008), and by 182Hf-182W dating (Kleine et al. 2008). As the chondrite parent asteroids were assembled after the youngest chondrules within them had formed, i.e., probably more than about 2.5 Myr after CAIs, then far from being the first bodies to have accreted, as is conventionally assumed, they were perhaps among the last to have done so.

The late accretion of chondritic asteroids was already suspected on petrographic grounds long before the recent chronological evidence became known. Fragments deemed to be of planetary igneous rock were identified in chondrites by Kurat and Kracher (1980), Hutchison et al. (1988), and Kennedy et al. (1992). The last of these authors reported a 2 mm chip of high Mn/Fe basalt in the Parnallee chondrite, which they interpreted as being derived from a high Mn/Fe planetesimal that had already melted and broken up before Parnallee’s parent body had accreted. In addition, Ruzicka et al. (1995) reported an unusual silica pyroxenite clast in Bovedy (L3), which they regarded as having a planetary igneous origin. More recently, Sokol et al. (2007) reviewed the occurrence of a wide variety of differentiated igneous rock fragments in chondrites, and Ruzicka et al. (2012b) reported further silica-rich clasts of supposed igneous origin. On a related note, Libourel and Krot (2007) discovered small pieces of texturally equilibrated olivine rock inside chondrules, which they interpreted as tiny fragments of earlier planetesimals that had been metamorphosed and then broken up by impacts before being incorporated into chondrules. Although Whattam et al. (2008) questioned that interpretation, the petrographic evidence, like the chronological evidence, clearly points to high-temperature planetesimal processing prior to chondrite accretion.

Heating by 26Al: The Key to the New Chronology

The cause of early melting and core formation, some 1–2 Myr before chondrules were made, is not hard to find. Ever since evidence for live 26Al was discovered in CAIs, it has been realized that the decay energy from this short-lived isotope (half-life 0.72 Myr) would have been more than sufficient to melt the fully insulated interiors of planetary bodies that accreted early enough, while radioactive heating was intense (Lee et al. 1977). The corollary is that the chondrite parent bodies, as they did not melt, accreted later, after the 26Al had largely decayed and lost its capacity to cause melting. This explanation is reinforced in the following paragraphs by a simple quantitative analysis of the likely effects of 26Al heating on the timing of initial planetesimal meltdown, the timing of chondrule formation, and the timing of chondrite accretion.

Our estimate of the energy available in 26Al to heat the first crop of planetesimals at t = 0 is about 6.6 kJ per gram of dry dust. This estimate requires knowledge of the concentration of Al in the dust, of the initial ratio of 26Al/27Al, and of the heat released by each decaying atom of 26Al. The concentration of Al is not known precisely, but would presumably have been more than 0.85 wt.% (the level in CI chondrites, which are extensively hydrated; Lodders and Palme 2009). We conservatively, although somewhat arbitrarily, choose a value of 1.2 wt.%, which corresponds roughly to the concentration of Al in dehydrated CI chondrite, and is close to the concentration of Al in most anhydrous chondrite groups (Lodders and Fegley 1998). We assume that the initial value of 26Al/27Al in the disk was uniformly 5 × 10−5, the so-called canonical value in CAIs (Jacobsen et al. 2008; MacPherson et al. 2010). A uniform distribution of canonical 26Al in the disk is indicated by the identical 26Mg/24Mg in the Earth, the Moon, Mars, and bulk chondrites (Thrane et al. 2006); by the correlation between the initial 26Mg/24Mg and the 26Al/26Mg ages of a suite of chondrules studied by Villeneuve et al. (2009); and by time intervals between specific events measured using the 26Al/26Mg chronometer being corroborated by other chronometers (e.g., Connelly et al. 2008). We are aware that Larsen et al. (2011) reported significant variation in 26Mg/24Mg in objects with solar 27Al/24Mg, and proposed that the initial 26Al/27Al in parts of the inner solar system where planetesimals accreted and where chondrules formed may have been substantially lower than the canonical level where CAIs were made. However, Wasserburg et al. (2011) found a wide variation in the initial 26Mg/24Mg of different CAIs with identical canonical 26Al/27Al, which leaves an open verdict for the case made by Larsen et al. (2011). Finally, we take the decay energy per atom of 26Al as 3.1 MeV (Castillo-Rogez et al. 2009). A plausible 10% uncertainty in both the initial 26Al/27Al, and in the wt. % Al, leaves our estimated initial radioactive energy at 6.6 ± 1 kJ g−1.

In addition to 26Al, the short-lived isotope 60Fe may have contributed to radioactive heating. However, the initial concentration of 60Fe and whether it was uniformly distributed in the disk remain unknown (Telus et al. 2012). The ratio at t = 0 of 60Fe/56Fe (1.5 × 10−6) assumed by Sanders and Taylor (2005) now seems far too high. Telus et al. (2011) suggest that it was between 3 and 5 × 10−7, making the contribution of 60Fe to heating <0.5 kJ g−1. Moreover, as 60Fe’s half-life of 2.6 Myr (Rugel et al. 2009) is more than three times longer than that of 26Al (0.72 Myr), its contribution to overall heating during the critical first 2 Myr would have been trivial, and we therefore ignore it in this paper.

Figure 1a shows the temporal decline in energy stored as 26Al in each gram of dry primitive dust, starting from the initial 6.6 kJ g−1, through almost seven half-lives during the first 5 Myr. To put this decline in perspective, 6.6 kJ g−1 is about four times larger than the 1.6 kJ g−1 needed to fully melt the insulated interior of a planetesimal at a temperature of 1850 K. The estimate of 1.6 kJ g−1 assumes starting from cold (250 K), with specific heat capacity, Cp = 837 J kg−1 K−1, and latent heat of fusion = 2.56 × 105 J kg−1 (Hevey and Sanders 2006). Thus, planetesimals that accreted during the first two half-lives of 26Al, or roughly during the first 1.5 Myr, would have had the potential to become completely molten in their fully insulated interiors.

Figure 1.

 a) Exponential decline with time of the potential thermal energy stored as 26Al in a gram of “dry” primitive dust. b) Time at which solidus (1425 K) and liquidus (1850 K) temperatures are reached in the fully insulated interior (deeper than approximately 5 km at t = 1 Myr to deeper than approximately 20 km at t approximately 5 Myr—see Figs. 2 and 4) of a planetesimal as a function of the time of its cold (250 K) instantaneous accretion and assuming no melt migration during heating. Arrows A, B, C, and D illustrate the timing of initial and total melting following cold accretion at t = 0, 0.75, 1.4, and approximately 2 Myr, respectively (see text for explanation). Accretion-time intervals labeled 1, 2, 3, and 4 relate to the fields shown in Fig. 4. The lower edge of the gray zone is the 1850 K liquidus calculated using latent heat and specific heat capacity values from Ghosh and McSween (1999), which are greater than those adopted here.

Figure 1b shows the time it would have taken to reach the onset of melting (the solidus temperature, approximately 1425 K) and also the completion of melting (the liquidus temperature, approximately 1850 K) of the fully insulated interior (i.e., with zero heat loss) as a function of the time of cold planetesimal accretion, assuming no migration of the 26Al heat source. With accretion at t = 0 (arrow “A” in Fig. 1b), heating would have been rapid and the liquidus would have been reached by about t = 0.3 Myr. This is in good agreement with the timing of earliest planetesimal melting and core formation, shown by 182W-deficit dating of iron meteorites (Burkhardt et al. 2008, 2012; Kruijer et al. 2011) and, although errors in the dating are large, such early melting clearly endorses the assumption that 26Al was the heat source.

With accretion at = 0.75 Myr (arrow “B”), the initial heating rate would have been half that for arrow “A,” but rapid enough for total internal melting to have been achieved by t = 1.5 Myr.

As a third example, with accretion at t approximately 1.5 Myr (arrow “C”), the insulated interior of a planetesimal would have carried just enough 26Al to reach the liquidus, but melting would not have been completed until after t approximately 5 Myr. This example may explain the paucity of chondrules that date from before t approximately 1.5 Myr (Kita and Ushikubo 2012). Assuming that chondrules (regardless of their formation mechanism) were produced in large numbers before t approximately 1.5 Myr, the scarcity of those old chondrules in meteorites must reflect their poor survival rate. We imagine that such chondrules accreted to planetesimals before t approximately 1.5 Myr and became buried in their insulated interiors where they would later have been melted down and destroyed. If this explanation is correct, then it implies that these pre-1.5 Myr chondrules, once made, did not linger in space, but accreted to planetesimals promptly and were thence destined for a magmatic grave.

Finally, with accretion after about t approximately 2 Myr (arrow “D”), the level of 26Al would have been too low to have heated the planetesimal’s interior to the solidus, so no melting at all would have taken place. The timing is consistent with the evidence that chondrites (which of course did not melt) accreted after about t approximately 2.5 Myr (by when chondrule production was in decline), and again corroborates the view that 26Al was the main heat source within planetesimals.

In summary, the timing of core formation before = 1 Myr, the scarcity of chondrules made before t = 1.5 Myr, and the accretion of chondrites after t = 2.5 Myr, combine to uphold our conviction that planetesimal heating by 26Al was a key factor in the evolution of planetesimals in the infant solar system.

The Structure of Molten Planetesimals

To visualize the changing internal structure of a molten planetesimal, we use the results of Hevey and Sanders (2006) who presented simulations of the heating, melting, and cooling of initially cold, porous planetesimals that accreted instantaneously. Their model is based on a radiogenic heat budget of 6.4 kJ g−1 of dust at t = 0, which is very close to the value of 6.6 kJ g−1 we estimate here. We note that they used an incorrect 26Al decay energy (4 MeV). That value wrongly includes approximately 1 MeV of energy that is not deposited as heat, but is lost in neutrinos. However, the error was fortuitously compensated by a lower concentration of Al (0.9 wt.%) compared with the 1.2 wt.% we use here.

As an example of their results, Fig. 2a illustrates the changing temperature profile within a planetesimal that accreted cold (250 K) at t = 0 and had a radius of 50 km (after early sintering and shrinkage). After a little over 0.3 Myr of heating, with the 26Al heat source evenly distributed, the interior deeper than approximately 5 km would have become uniformly hot and 50% molten (approximately 1725 K). At this stage, the interior is assumed to have lost rigidity and become cohesionless magmatic slurry undergoing turbulent thermal convection. With continued intense heating beyond 0.3 Myr, the magma is assumed to have remained at sub-liquidus temperatures, but to have increased in volume as the overlying rigid carapace was melted upward from its base and its thickness reduced from approximately 5 km at = 0.3 Myr to just 0.5 km by t = 0.5 Myr (Fig. 2b). By that time, the rate of conductive heat loss through the residual crust would have reached a maximum, equaling the rate of internal heat production, and the crust’s thickness would have been at a minimum. Thereafter, with heat production lower than heat loss, no further melting would have occurred, and the crust would have thickened, slowly at first but ever more rapidly, over the next 2 or 3 Myr and beyond. A cartoon of the state of the planetesimal at = 2 Myr is shown in Fig. 3.

Figure 2.

 a) Temperature profiles at selected times inside a planetesimal with a 50 km radius and zero porosity that accreted at t = 0 and a temperature of 250 K and was heated by 26Al decay. Broken lines are profiles during heating (until 0.5 Myr) and continuous lines are profiles during cooling. Convection began soon after t = 0.3 Myr, and by t = 0.5 Myr the molten, convecting interior had expanded to within about 0.5 km of the surface (after Hevey and Sanders 2006). b) Depth of solid rock and crust (gray) for the same body as a function of time.

Figure 3.

 Cartoon showing the internal structure of the 50 km radius planetesimal exemplified in Fig. 2 at t = 2 Myr. The core is assumed to be fully formed, but it is possible that small droplets of metal may have been held in suspension by turbulent convection (symbolized by curved arrows) in the magma ocean. The base of the crust is arbitrarily taken as the level at which the interior is 50% molten; the lower crust, with less than 50% melting, is deemed to be rigid. The 2 m of dust shown on the surface is predicted by instantaneous accretion; in reality, continuous accretion probably led to a considerable thickness of cool, loose, dusty debris, particularly after about t = 2 Myr when 26Al heating was very weak.

The model predicts that during maximum heat loss, only about 2 m of porous, unconsolidated, and extremely insulating dust separated solid, sintered rock at 700 K from the surface at 250 K. However, this prediction assumes that all accretion was completed instantaneously at t = 0. In reality, at least some accretion would have continued after the initial aggregation of material. Beyond about t approximately 2 Myr, with the 26Al heat source fading, any such late accretion would not have melted, but accumulated as a coating of loose, or weakly consolidated, dusty debris, which could have attained a considerable thickness.

What if the radius had been much smaller than the 50 km chosen in Fig. 2? Hevey (2001) showed that a body with a 20 km radius would have melted substantially by t = 0.3 Myr, and its crust would have thinned down to a minimum of 1.5 km by t = 0.5 Myr, but its high surface-to-volume ratio would then have led to rapid cooling, and the body would have been largely solid by t = 3 Myr. A body 10 km in radius would scarcely have melted at all. In a comparable thermal model, Moskovitz and Gaidos (2011) predicted similar melting behavior.

What if the radius had been larger? Going up in size, if the radius had been doubled, and was 100 km instead 50 km, the ratio of internal heat production to surface area would also have doubled, giving twice the heat flow and halving the thickness of insulating crust to a mere 250 m during the period of peak heat loss between t = 0.5 and t = 1.5 Myr. In this case, the crust may perhaps easily have foundered, exposing incandescent magma at the surface. With a still larger radius, the insulating carapace would have become even thinner and even more susceptible to foundering.

Figure 4 shows the effects of a planetesimal’s radius on its predicted melting behavior combined with the effects of the timing of its accretion discussed above (Fig. 1b). Four fields, numbered (1) to (4), correspond to the four accretion time intervals shown in Fig. 1b. Planetesimals starting in field (1) would have become extensively molten before t = 1.5 Myr, with very thin crusts as depicted in Figs. 2 and 3. Such planetesimals, we suggest below, would have been potential sources of chondrules by impact splashing. Planetesimals starting in field (2) would also have undergone extensive internal melting, but beneath a thicker insulating crust than for field (1), and with a longer heating period, generally becoming molten from t approximately 1.5 Myr up to t approximately 5 Myr depending on the time of accretion. Planetesimals starting in field (3) would have melted only partially, and they probably would have remained rigid. It is possible that the melt fraction (basalt magma) would have migrated upward, and that these planetesimals included the parent asteroids for primitive achondrites like the lodranites and ureilites. Planetesimals from field (4) would never have melted, although they may have become heated and metamorphosed. They would have become chondrite parent bodies. Figure 4 can be regarded as a refinement of the related, but oversimplified and rather misleading, two-field diagram presented by Hevey and Sanders (2006, fig. 6) in which planetesimals were shown either to have melted or not melted. It also bears similarities to fig. 5 of Moskovitz and Gaidos (2011), although the latter has later accretion times for given outcomes because it assumes 4 MeV, and not 3.1 MeV, as the decay energy per atom of 26Al.

Figure 4.

 Plot showing the eventual outcome of heating by 26Al decay in planetesimals as a function of radius and time of instantaneous cold accretion. The boundaries that delineate the four different fields are interpolated from Hevey and Sanders (2006, fig. 6) and Fig. 1b. Field (1) delimits planetesimals that will become substantially molten beneath a thin (e.g., <1 km) insulating crust before t approximately 1.5 Myr. Such planetesimals, we argue, will be prime candidates for bursting into chondrule spray if disrupted by impact between t approximately 1.5 and t approximately 2.5 Myr.

The thermal model of Hevey and Sanders (2006) assumes that the internal 26Al heat source remains evenly distributed at all stages during heating and cooling. However, some authors question this assumption, arguing that basalt migrates rapidly upward as soon as it is generated (e.g., Moskovitz and Gaidos 2011). As nearly all aluminum enters the basaltic melt fraction, the removal of such melt would also remove the heat source, forestall further internal heating, and invalidate the pattern of melting shown in Fig. 2. Wilson and Goodrich (2012) even suggested that basalt migration away from the zone of partial melting was so rapid that “high degrees of mantle melting never occurred in any asteroids.”

While we acknowledge that basalt was removed in the case of the ureilite parent body, which we believe to have accreted late, in field 3 of Fig. 4, we cannot accept that basalt migrated from its source in all cases of planetesimal melting. Global magma oceans and very high melt fractions almost certainly did develop within some young planetesimals. They were inferred by Taylor et al. (1993) who argued that the IVB iron meteorites crystallized from liquid iron that contained <1 wt.% sulfur (Goldstein et al. 2009) at >1770 K, a temperature at which primitive silicate material would have been well over 50% molten. Taylor et al. (1993) also noted that, if pallasite olivine represents unmelted residue, then the silicate melt fraction must have been between 70% and 90%. Keil et al. (1989) argued that the unusual texture of the Shallowater aubrite indicates a molten enstatite magma ocean at 1850 K. So while the Hevey and Sanders model is necessarily simplified, the evidence for magma oceans with high melt fractions suggests that the model is not wildly wrong. The issue of precisely how global magma oceans were created is a matter for future investigation; for now, we merely speculate that the mechanism may possibly have been linked to gradual accretion with the newly added material at an early stage (e.g., t < 1.5 Myr) continually “dissolving” in any highly radioactive rising basalt magma (see Kleine et al. 2012), or it was perhaps linked to the onset of convection before significant melting had occurred, facilitated by a possible substantial reduction in bulk viscosity (Schölling and Breuer 2009). Regardless of the details of the melting mechanism, we suspect that substantially molten interiors were the norm rather than the exception in planetesimals that accreted within field (1) of Fig. 4.

The Prevalence of Early Molten Planetesimals

We suggest that early accretion leading to extensive melting and core formation before t approximately 1.5 Myr was widespread in the protoplanetary disk. Between 100 and 150 separate parent bodies appear to be represented by the meteorites in the world’s collections, one body for each meteorite group plus many more represented by ungrouped meteorites (Meibom and Clark 1999; Burbine et al. 2002). Only about one in five of these inferred bodies is chondritic. The rest of them, some four of every five, melted and became differentiated bodies. If the low ε182W measured in magmatic iron meteorites is representative of the timing of melting in general, then it is conceivable, indeed likely, that most of the material in the inner solar system existed as molten, or partially molten, planetesimals by the close of the first 1.5 Myr.

The Production of Chondrules

The Case for Making Chondrules from Molten Planetesimals

The above discussion points to a young solar nebula populated by largely molten spheres of primitive magma undergoing turbulent convection at near-liquidus temperatures, each enclosed by a thin outer shell of rigid, thermally conducting crust, and coated by a layer of unconsolidated dust (Fig. 3). Molten metal possibly formed suspended globules that would eventually segregate as cores. As, over time, these planetesimals must have been increasing in size and decreasing in number as a result of mutual collisions and mergers (the necessary early steps along the stochastic road to planet formation), we deduce that many of the collisions would have launched huge plumes of molten droplets (chondrule spray), mixed with loose dust from the planetesimal surfaces, into the disk. Incidentally, while such collision plumes have commonly been described in the literature as a “planetary” setting for chondrule formation, this is misleading. A plume might more appropriately be regarded as a special case of a nebular setting, albeit a rather local and ephemeral one.

This scenario for chondrule formation, which is a clear alternative to the conventional shock-melting of dust-clumps, is the main subject of this article and we will now explore it in detail, amplifying and developing the case for it made recently by Asphaug et al. (2011). These authors noted that most collisions would have been oblique glancing blows in which much of the “overlapping” portion of the smaller body (the projectile) would have been engulfed by the larger body (the target), while the “overshooting” portion would have undergone catastrophic decompression and expanded slowly down-range as a huge fan-shaped plume of droplets and debris.

Asphaug et al. (2011) emphasized that encounter velocities between merging planetesimals at the time of chondrule formation would generally have been close to, or less than, the combined escape velocity of the merging pair, namely tens of meters to a few hundred meters per second (the escape velocity in meters per second is numerically about the same as the planetesimal’s radius in kilometers). At these very low velocities, no shock melting would have occurred; almost the entire enthalpy for melting would already have been generated by the decay of 26Al and stored as magma in the planetesimals. The kinetic energy of impact would have caused mechanical disruption, but would have been negligible compared with 26Al decay as a cause of melting.

The splashing model is, thus, quite different from, and must not be confused with, the production of impact melt spherules by energetic, high-velocity collisions. Such melt droplets include crystal-bearing lunar spherules, evidently made during excavation of the huge impact basins on the Moon (Symes et al. 1998; Ruzicka et al. 2000), and terrestrial impact spherules of which those from the Eltanin site include beautiful coalesced droplets similar to compound chondrules (Kyte et al. 2010). They could also include the chondrules in the CB chondrites, believed to be condensates from a giant impact plume about 4 Myr after CAIs (Rubin et al. 2003; Krot et al. 2005).

We consider in the following paragraphs how the predictions of the “splashing” scenario might be reconciled with what is known or can be inferred about the formation of chondrules. We initially reflect on the following properties of chondrules: chondrule ages; chemical compositions, “peak” temperatures and abundances; chondrule sizes; mutually indented shapes; cooling rates; relict grains; igneous rims; and metal inclusions.

Chondrule Ages Are Mostly 1.5–2.5 Myr After CAIs

A strength of the splashing model is that it can explain why chondrules are mostly between 1.5 and 2.5 Myr younger than CAIs (Connelly et al. 2008; Kleine et al. 2008; Kita and Ushikubo 2012). 26Al-induced planetesimal meltdown means that splash-generated chondrules would have been made from as soon as the first planetesimals had melted, i.e., from about t = 0.3 Myr onward. However, as explained above, we suggest that most chondrules made before t = 1.5 Myr were destroyed because they accreted promptly and became buried inside planetesimals that were still highly radioactive and destined to melt.

Those few chondrules that survive from before t = 1.5 Myr, such as some very old ones in Allende (Connelly et al. 2011), may fortuitously have become embedded in bodies that were too small to retain heat, or embedded in the cool outer layers of larger bodies (LaTourrette and Wasserburg 1998), or perhaps in bodies where low temperatures were buffered by a large amount of ice. As these kinds of low-temperature setting would perhaps also have been necessary for the preservation of CAIs, which date from t = 0, it may be no coincidence that the oldest chondrules appear to survive in meteorites (i.e., CV chondrites) with a large abundance of CAIs.

Why would chondrule production have declined rapidly after about t = 2.5 Myr? Chondrule production by splashing would have continued for as long as molten planetesimals with thin crusts were colliding and merging. By t approximately 2.5 Myr, heat loss by conduction would have far exceeded internal heat generation by 26Al decay, and crusts would probably have been growing thicker, cooler, and mechanically stronger as their underlying magma oceans crystallized (Fig. 2b). In addition, following many collisions and the production of a correspondingly large volume of chondrule-rich debris, thick accumulations of this debris may have built up on the surfaces of remaining intact planetesimals, rendering those planetesimals less susceptible to bursting open on impact. In this light, the duration of chondrule formation between t = 1.5 Myr and t = 2.5 Myr would appear to coincide with a period of transition from a disk populated mainly by molten bodies of various sizes, to a disk with fewer, larger bodies (planetary embryos, perhaps) alongside a suite of newly accreted “second generation” chondritic bodies.

We note that chondrule production did not end abruptly at t = 2.5 Myr; many CR chondrules have ages of around 3 Myr after CAIs (Kita and Ushikubo 2012), based on their very low initial 26Al/27Al ratios, and their Pb-Pb ages.

Chondrules Have “Primitive” Chemistry, They Cooled From Near-Liquidus Temperatures, and They Are Abundant

These three classic features of chondrules are clearly compatible with the splashing model. The essentially unfractionated “primitive” chemistry of most chondrules, including their flat rare-earth element profiles (Jones et al. 2005), is consistent with the high degree of melting we envisage in molten planetesimal interiors prior to their disruption (Fig. 3). The high subliquidus temperatures from which most chondrules cooled (inferred from their igneous textures) is consistent with their derivation from hot convecting magma oceans where the temperature was buffered just below the liquidus by steady radioactive heating off-set by efficient convective and conductive heat loss. The high volume fraction of chondrules, amounting to 80% or more by volume in some ordinary chondrites, is seen as a simple consequence of the dominance of the molten planetesimals that supplied those chondrules.

Chondrule Sizes Are Mostly in the Range 0.1–2 mm Across

The sizes of most chondrules are consistent with droplet sizes expected from the spraying or spattering of larger volumes of magma. They fall in the range observed widely in naturally formed droplets of magma such as those, known as Pele’s tears, produced in basalt lava fountains at Kilauea volcano on Hawai’i. Other examples include the famous orange glass spherules at the Apollo 17 site on the Moon (also presumed to have formed in a lava fountain), and droplets produced by energetic impacts and shock melting, such as the crystal-bearing lunar spherules, and terrestrial impact-melt spherules mentioned above. Also, chondrule-sized droplets were produced experimentally by the spattering of melt in a solar furnace (King 1983). Finally, in their model, Asphaug et al. (2011) showed that chondrule-sized droplets result from equating the total surface energy of droplets in the expanding plume with the energy associated with the catastrophic disruption and decompression of the interior of the molten planetesimal.

The Inferred Time for Cooling to the Solidus Was Typically Several Hours

The successful experimental replication of chondrule textures under controlled rates of cooling shows that near-liquidus droplets probably cooled and solidified over a matter of hours, and not seconds, nor days (Lofgren 1989; Hewins and Radomsky 1990; Radomsky and Hewins 1990). This timescale was also inferred from a study of zoning in metal grains in CR chondrites (Humayun 2012), and it is in broad agreement with the kind of time needed for an impact plume to expand and cool (Asphaug et al. 2011).

Chondrules in Some Meteorites Have Mutually Indented Shapes

Hutchison and Bevan (1983) noted that some chondrules in Tieschitz (H3) are apparently molded against neighboring chondrules, indicating that they were still hot and plastic, if not liquid, when they came into contact. Holmén and Wood (1986) reported similar textures in other chondrites. This remarkable feature was also described by others (Sanders and Hill 1994; Hutchison 1996; Zanda 2004), but it has received only limited attention, perhaps because of arguments that the deformed shapes could have resulted from the compaction of chondrules into voids by shock (Sneyd et al. 1988; Scott et al. 1992).

Recently, Metzler (2011, 2012) described spectacular examples of unshocked, mutually molded chondrules in large clusters, which he called “cluster chondrites.” They occur as clasts up to 10 cm across in ordinary chondrite breccias. He noted “this rock type consists of a mixture of deformed and undeformed chondrules and is characterized by low abundances of inter-chondrule matrix, low abundances of distinct chondrule fragments, and restricted variations of chondrule sizes.”Hewins et al. (2012) have since reported similar clustering of chondrules in Semarkona.

Cluster chondrites are consistent with the splashing model. They suggest that molten chondrules in dense swarms, as would be expected in an impact plume, aggregated either into “sticky clusters” at least 10 cm across, like giant compound chondrules made from many thousands of individuals, or they accreted rapidly to the surface of the target body, forming a blanket of molded chondrule rock in the manner envisaged by Asphaug et al. (2011). In the latter case, if the target body were to have become the projectile in a later collision, then fragments of this welded chondrule blanket would have been launched and may eventually have come to reside in the kind of chondritic breccia described by Metzler. The accretion of hot chondrules directly onto a planetesimal surface is suggested by the parallel orientation of flattened molded chondrules seen, for example, in the Bovedy (L3) chondrite (Sanders and Hill 1994).

The good size sorting observed by Metzler is significant. Size sorting has been observed widely in chondrites, but its origin remains unclear. In the case of cluster chondrites, it appears that the sorting happened locally, and very rapidly, during the brief interval between the formation and accretion of a batch of chondrules.

Many Chondrules Contain So-Called Relict Grains (Xenocrysts)

Xenocrysts (foreign crystals) within a chondrule are recognized because they are chemically out of equilibrium with other crystals in the same chondrule and may also have anomalous oxygen isotopic compositions (Jones et al. 2005; Ushikubo et al. 2012). They were first reported by Nagahara (1981) and Rambaldi (1981). Nagahara noted that these peculiar crystals imply that the chondrule host did not condense from a cooling vapor, so must have been melted, and thus xenocrysts became dubbed “relict grains” and were widely assumed to provide evidence for precursor dust-clumps.

Connolly and Hewins (1995) showed experimentally that chondrule liquids tend to wet and swallow up dust grains that impinge on them. We suggest, therefore, that xenocrysts were trapped and engulfed by melt droplets in flight, having been launched into the impact plume from the loose surficial regolith. In this context, perhaps smaller dust particles in the plume may similarly have been engulfed, and so become the seed crystals deemed necessary for the development of porphyritic textures in chondrules, as such seed crystals may not have been ubiquitous in the convecting magma prior to collision.

Some Chondrules Appear to Have Been Melted More Than Once

Textural evidence in chondrules for “multiple re-heating” events, such as igneous rims, has been widely reported (e.g., Rubin and Krot 1996). In the context of collision plumes, much of the re-launched loose debris will have been derived from older chondrules. Thus, chondrules which bear evidence of having been through a few distinct heating and melting phases, may simply have been caught up in collision plumes more than once. Their coarse-grained rims may either be the enveloping mantles of compound chondrules, or simply be the result of heating of former dusty rims while suspended within the incandescent plume. Also, it is possible that, with turbulent motion in the expanding plume, an individual chondrule may have moved from a hot to a cooler region and back again. Incidentally, we note that crystal-bearing lunar spherules, which originated in impact plumes, in some cases also show chondrule-like features that suggest re-heating (Ruzicka et al. 2000).

Some Chondrules Contain Blebs of Metal

Small blebs of metal inside chondrules (e.g., Wasson and Rubin 2010) may have origins that are consistent with the splashing model. Tiny globules of metal may have been kept in suspension in the magma ocean by turbulent convection (Fig. 3). Some of these globules of metal may have rained down late into the magma ocean as the overlying crust, including any late-accreted material, was melted from below and thinned down (Fig. 2). Another possible explanation for metal inside chondrules is that miniscule droplets of metal spray may simply have become engulfed, rather like xenocrysts, within silicate droplets in the expanding impact plume.

Problems with Making Chondrules from Clumps of Dust

We now consider a number of additional features of chondrules that appear to pose difficulties for the conventional view that chondrules began as clumps of dust. We discuss sodium in chondrule olivine, oxidized iron in chondrules, chondrule diversity, abnormally large crystals in chondrules, giant chondrules, the scarcity of dust-clumps, and the cause of melting. In all cases, we show that the features can be explained in the context of collision and splashing.

Sodium Has Surprisingly High Concentrations in Olivine Crystals Within Chondrules

Alexander et al. (2008) reported high concentrations of Na in chondrule olivine which, together with an absence of isotopically mass fractionated isotopes of alkalis, point to high gas pressures and very closely spaced chondrules (high chondrule densities) in enormous clouds measuring hundreds to several thousand kilometers across. They inferred that these large chondrule clouds were generated by shock-melting of equally large clouds of precursor dust-clumps that would quickly have collapsed under their own gravity and become new planetesimals.

We question the feasibility of their scenario. Conventional dust-heating models do not explain how nebular processes could have concentrated the dust-clumps to the required extent, nor how nebular shock heating, or any other external heat source, could have heated and melted such a huge mass of dust (equivalent to an entire planetesimal) in what was effectively a single event. In addition, the energy would need to have been delivered at just the right moment—after the cloud had become gravitationally unstable, but before its collapse was completed. Finally, it is difficult to see how cold primitive matrix dust could have been mixed in with the hot chondrules prior to accretion.

By contrast, a molten planetesimal collision would have created an enormous (planetesimal-scale) transient dense cloud of droplets. The droplets would have been immersed in their own Na-saturated vapor and therefore capable of retaining high Na concentrations during olivine crystallization, at least during the early stages of plume growth. Dispersal of droplets after solidification in the cooling, expanding cloud would have allowed mixing of chondrules from different impacts and addition of primordial dust containing presolar grains. The dust could have come directly from the nebula or it could have been derived as recycled dust from the poorly consolidated regolith of the disrupted planetesimals.

In a parallel study to that of Alexander et al. (2008), Hewins et al. (2012) reported similarly high levels of Na in chondrule olivine, and also reported Na in melt inclusions in the olivine, in the mesostasis, and in bulk chondrules, all in Semarkona (LL 3.0). They inferred partial evaporative loss of Na from the molten chondrules, followed by its later re-condensation and its incorporation in chondrule mesostases. Unlike Alexander et al. (2008), they were not committed to the idea of conventional shock melting, and suggested as an alternative that chondrules may have formed in “debris clouds after protoplanetary collisions,” but they did not elaborate on how the melt droplets formed.

Chondrules in enstatite chondrites are not only Na-rich but contain evidence for sulfidation of silicates and metal-sulfide nodules (e.g., Rubin 1983). Lehner et al. (2011) and Petaev et al. (2012) propose that ordinary ferromagnesian chondrules reacted with an S-rich and H-poor gas above 1000°C. Such conditions, we suggest, would have been more easily created within an impact-generated plume rather than within the conventional solar nebula.

Chondrules Contain Iron as FeO

Type II chondrules contain significant levels of FeO. Even type I chondrules are rarely completely free of it. FeO poses a serious problem for the production of chondrules from dust-clumps in the nebula because the stabilization of FeO in chondrule melts requires ambient gas that is several orders of magnitude more oxidizing than the standard hydrogen-rich nebula. In an ongoing effort to resolve this conundrum, Fedkin and Grossman (2006, 2010), Fedkin et al. (2012), and Grossman et al. (2012) reaffirmed the view of Alexander et al. (2008) that chondrules were created in close proximity to each other (to retard evaporation of volatile elements like Na), but also concluded that chondrules were enveloped by oxidizing, H2O-bearing, gas to stabilize FeO. They tentatively suggested that such a setting might have existed in the aftermath of collisions involving icy planetesimals. This setting for chondrule formation is difficult to reconcile with the shock melting of dust-clumps because the timing of the heating event would need to have coincided precisely with the brief, ephemeral existence of the H2O-enriched collision plume.

On the other hand, the production of type II chondrules by collision and splashing of molten planetesimals poses no problems, provided of course that the planetesimals were made of FeO-bearing magma. We return to this issue in the final section of the article.

Chondrules Are Chemically Diverse

While chondrules have primitive chemistry in a broad sense, they are not identical, and they vary particularly in their contents of olivine and pyroxene. Their chemical diversity presents problems for the dust-clump hypothesis. A millimeter-sized chondrule precursor clump might typically have contained between 106 and 109 dust grains between 1 and 10 microns across. In a well-mixed disk, therefore, all dust-clumps might be expected to have sampled much the same statistically representative selection of available grains and so shared the same primitive chemical composition. Thus, olivine-rich and pyroxene-rich chondrules are very unlikely to have formed from clumps of micrometer-sized dust grains because nebular processes cannot conceivably have sorted the dust into millimeter-sized monomineralic aggregates.

We suggest, in the context of the splashing model, that magmatic processes in molten planetesimals prior to collision could have generated liquids ranging from olivine-rich to pyroxene-rich. With high degrees of melting, olivine would have been the sole liquidus phase. Olivine has an atomic Si/Mg ratio of 0.5, which is less than solar (0.9), so its removal by crystal settling from a primitive unfractionated magma ocean would have driven the residual melt toward pyroxenitic compositions for which the Si/Mg ratio is 1.

Mostefaoui et al. (2002) observed that olivine-rich chondrules (lower Si/Mg) were made, in general, earlier than pyroxene-rich chondrules (higher Si/Mg). Tachibana et al. (2003) and Kita et al. (2005) confirmed this correlation between Si/Mg and age with more precise measurements and an extended data set. Tachibana et al. (2003) attempted to explain the correlation in terms of conventional flash-melting of dust balls in an open system over a number of cycles. With each cycle, differential evaporation of Si relative to less-volatile Mg meant that successive generations of dust-grain precursors became progressively enriched in condensates with higher Si/Mg.

While we agree that evaporation and condensation processes in the nebula may have been important, we cannot see how, in physical terms, the chondrules (with low Si/Mg) were removed from the system following each cycle. We prefer to explain the correlation in terms of olivine crystal fractionation. We suggest that over time the Si/Mg ratio of the magma oceans in many planetesimals increased “in step” at roughly the same rate, and so did the Si/Mg of the chondrules made from those magma oceans by collision and splashing.

In support of this explanation, we draw attention to evidence in differentiated meteorites that magmatic fractionation driven by crystal settling probably did occur during the first few million years. Baker et al. (2012) reported a deficit in δ26Mg in main group pallasites that corresponds to separation of crystals from liquid at around 1 Myr after CAIs. A similar result was obtained from the Eagle Station pallasite by Villeneuve et al. (2011). On the HED, parent body crystal fractionation, albeit of pyroxene rather than olivine, led to igneous differentiation over 1 or 2 Myr, reflected in a steady increase in δ26Mg going from the most primitive diogenites to cumulate eucrites (Schiller et al. 2011).

Some Chondrules Contain Abnormally Large and Oscillatory-Zoned Crystals

Some single olivine crystals in chondrules are unusually large and may occupy >90 vol.% of the chondrule. They seem unlikely to have crystallized from chondrule-sized melt droplets and more probably originated in much larger volumes of melt. Large rounded olivine grains that occupy almost all the chondrule (e.g., fig. 1e in Jones et al. 2000) are especially difficult to form by conventional mechanisms involving dust-clumps. However, rounded crystals may form in convecting magma under conditions that periodically require dissolution as well as growth. In some terrestrial settings, phenocrysts develop rounded and embayed shapes, apparently for this reason.

Olivine grains are also found as large (<1 mm) isolated single crystals in chondrite matrices and their origin has not been satisfactorily explained. Steele (1989) and Weinbruch et al. (2000) argued that they are nebular condensates, but Jones et al. (2000) found traces of low-Ca pyroxene and mesostasis attached to large rounded forsterite crystals, and inferred that they formed in chondrules. We suggest that the large isolated crystals in the matrix, like those in chondrules, were in suspension in asteroidal magma oceans at the time of collision and disruption.

Oscillatory chemical zoning is present in some large crystals of olivine and pyroxene in both chondrules and chondrite matrices (Steele 1995; Jones and Carey 2006; McCanta et al. 2009; Blinova et al. 2011). Oscillatory zoning in some elements such as phosphorus (McCanta et al. 2009) may result from disequilibrium crystallization, but oscillatory zoning in Fe and Mg (Blinova et al. 2011) is hard to explain by crystal growth in a droplet, and may perhaps be attributed to variations in the environment of a growing crystal that was suspended in convecting magma (Shore and Fowler 1996).

Macrochondrules Exist

Some chondrules, called macrochondrules or mega-chondrules, are more than one centimeter across, yet have compositions and textures like those of normal millimeter-sized chondrules (Binns 1967; Prinz et al. 1988; Hill 1993; Bridges and Hutchison 1997; Ruzicka et al. 1998). Macrochondrules would be difficult to form by a nebular flash-melting process because of the need to transfer heat rapidly to the center of what would presumably have been a large (e.g., golf-ball sized), porous, and thermally insulating dust-clump without glazing or vaporizing the outside first. Also, as the rise in temperature would generally be proportional to the ratio of the surface area to the mass of a dust-clump, then a flash-melting event that delivers just the right amount of energy to make millimeter-sized chondrules would barely affect the temperature of a centimeter-sized clump. We prefer to interpret macrochondrules simply as large blobs of magma that failed to be shaken and dispersed into normal-sized chondrule droplets in the collision plume.

Dust-Clumps Are Rarely Observed in Chondrites

If chondrules had formed from precursor dust-clumps, we might expect to find small lumps of such material in the interchondrule matrix. Although there is evidence for them in some meteorites (Rubin 2011, pp. 553 and 554), millimeter-sized aggregates that might represent plausible precursor dust-clumps are really rather rare (e.g., Scott and Krot 2007). Moreover, matrix is typically more FeO-rich than chondrules, so it is an inappropriate starting material.

Some authors have reported dust-rich chondrules which they interpret as incipiently melted products of nebular flash-heating (Nettles et al. 2001; Ruzicka et al. 2012a). Certainly, these objects display clear evidence of low levels of intergrain melt, but we suggest that they are not necessarily products of dust-clump melting. We can envisage three possible ways that they might have been generated in the collision and splashing scenario. First, they may be fragments broken from the incipiently melted lower rigid crust (Fig. 3) of the disrupted projectile. Second, they may be some sort of welded accretionary lapilli that grew in the impact plume of droplets and dust; this is how we would interpret chondrule Beg-6 reported by Ruzicka et al. (2012a). Third, they may be melt droplets that became choked in flight with dust grains, which is perhaps just a special case of the second example. In both the second and third examples, the principal source of dust would have been the loose accumulation of regolith covering the colliding planetesimals.

A Plausible Heat Source for Melting Dust-Clumps Remains Elusive

The identification of a heat source capable of melting nebular dust-clumps on the required scale has proved to be somewhat intractable. The currently favored nebular shock-melting model (Desch and Connolly 2002; Boss and Durisen 2005; Morris and Desch 2010) claims to provide an appropriate thermal history for the inferred chondrule cooling rates, and Morris et al. (2012) similarly suggest that bow shocks around planetary embryos may provide a plausible thermal regime for making chondrules. Nevertheless, as discussed here, shock melting is difficult to reconcile with such observations as chondrule diversity, macrochondrules, and the retention of Na and FeO in chondrules. Also, shock melting should lead to significant isotopic fractionation effects, but these are not observed. For example, Fedkin et al. (2012) calculated that at enormously inflated PH2O (×550, to maintain iron as FeO rather than metal) and with huge dust enrichment (×600, to retard evaporation), all iron would evaporate during shock-wave heating, that olivine would grow in the unevaporated Mg-rich melt droplets before re-condensation, and large isotopic fractionation effects between olivine and glass would be preserved. As such isotopic fractionation is not seen, they concluded that chondrules were unlikely to have formed in nebular shocks.

Turning briefly to other postulated sources of rapid nebular heating, the once fashionable X-wind model (Shu et al. 1997, 2001) in which proximity to the protosun is linked to melting, seems beset with insurmountable problems and has now been largely discredited (Desch et al. 2010). Electrical discharge heating (Love et al. 1995), electromagnetic radiative heating (Eisenhour et al. 1994), gamma-ray bursts (McBreen and Hanlon 1999), and heating in protostellar jets (Liffman and Brown 1996), along with other heat sources mentioned in a review by Rubin (2000), all have serious shortcomings and are not considered viable.

The splashing model, in total contrast, has a clearly understood and quantifiable heat source in 26Al, and obviates the need to find a mechanism for the rapid heating of dust in the nebula.

Chondrules and the Evolution of Planetesimals

We now explore in the context of the splashing model, the nature of the planetesimals that preceded the period of chondrule formation and the properties of the chondrite parent bodies that postdated chondrule formation.

The Nature of the Postulated “Precursor” Planetesimals to Chondrules

If the molten planetesimal splashing hypothesis is correct, then type I (FeO-poor) and type II (FeO-rich) chondrules suggest, respectively, that the precursor planetesimals themselves were compositionally bimodal, being either volatile-depleted and reduced with most of their iron in metal, or volatile-bearing with much of their iron as FeO.

What Was the Planetesimal Source for Type I (FeO-Poor) Chondrules?

We tentatively suggest that type I chondrules came from molten parent bodies which were the same as, or similar to, the parent bodies of certain iron meteorites. Many iron meteorites are extremely depleted in moderately volatile siderophile elements, and their parent planetesimals had generally melted within the first million years (Burkhardt et al. 2008). As Bland and Ciesla (2010) noted, they may have acquired their volatile-depleted chemical signature by accreting early from partially condensed matter in the still-hot inner region of the infant disk, perhaps close to 1 AU (Bottke et al. 2006) where the ambient gas would have been hydrogen-rich and reducing.

In this light, we wonder whether the giant hit-and-run collision postulated by Yang et al. (2007) to account for the IVA iron meteorite parent body may have created an enormous cloud of type I chondrules. That collision, it has been argued, left in its path a string of secondary planetesimals, one of which (the IVA body) was purportedly a sphere of molten metal perhaps 300 km in diameter covered by a veneer of silicate rock. One particular IVA meteorite, Muonionalusta, has a Pb-Pb age of 4565.3 ± 0.1 Myr (Blichert-Toft et al. 2010), suggesting that the metal was solid by that time and that the postulated giant collision happened somewhat earlier, consistent with the age of chondrules. We note that the published age for Muonionalusta may be about 1 Myr too old, as it is based on an assumed 238U/235U ratio of 137.88 that is probably too high (Connelly et al. 2011). Nevertheless, the uncertainty in its age does not affect our contention that the postulated giant IVA collision could have been synchronous with chondrule formation.

The idea that volatile depletion is a signature of very early planetesimal accretion from a hot partially condensed nebula is supported not only by certain iron meteorites but also by the basaltic achondrites. The HED meteorites and the angrites are both highly depleted in volatile elements, and recent revision of their Sr isotope systematics (Hans et al. 2011; Kleine et al. 2012; Moynier et al. 2012) suggests that their parent bodies accreted, like the iron meteorite parent bodies, close to the time of formation of CAIs. An earlier proposal by Halliday and Porcelli (2001) that volatile depletion in the angrites and HEDs might have resulted from giant impacts 2 or 3 Myr after CAIs now seems less likely.

On the same theme, Scott and Sanders (2009) appealed to a reservoir of volatile-depleted bodies dating from the time of CAI formation as the refractory Mn-poor end-member of a mixing line with Mn-rich CI-like dust to explain the whole-rock Mn-Cr isochron for carbonaceous chondrites.

What Was the Planetesimal Source of Type II (FeO-rich) Chondrules?

We suggest that, in contrast to the reduced and volatile-depleted planetesimals that gave rise to type I chondrules, type II chondrules came from planetesimals that were initially made from aggregates of dust and water-ice. On becoming heated by 26Al, the ice would have melted to water, which would then have reacted, while still at a low temperature, with anhydrous silicate grains and metal to produce minerals such as serpentine and magnetite. On further heating, this oxidized suite of minerals would have remained oxidized as it became dehydrated and finally melted.

The accretion of water-ice would have required a cold nebular setting. This setting may have developed later than the time of accretion of the volatile-poor type I planetesimals, after the initially hot nebula had cooled sufficiently for water-ice to condense, or perhaps it existed farther from the Sun, where it had always been suitably cold. In the latter case, subsequent orbital migration presumably brought oxidized and reduced planetesimals close together, because type I and type II chondrules are found side by side in many meteorites.

We note that there are limits to the amount of water-ice that could have been incorporated into a planetesimal if the planetesimal were later to have melted. The latent heat absorbed by vaporizing ice is a massive 2.6 kJ g−1, and the specific heat of steam is 2 J g−1 K−1. Thus, with only 6.6 kJ g−1 available in the dust, a planetesimal accreting at t = 0 from equal masses of ice and dust could not have melted in time to make chondrules by splashing, and perhaps it never would have melted. However, somewhat lower mass fractions of ice could have led to different levels of oxidation of iron as well as permitting melting in time to make chondrules.

Planetesimals would have grown over a period of time, both from gradual accumulation of nebular dust, and from mergers with other planetesimals. Thus, they would have been composite bodies that incorporated both reduced and oxidized iron prior to their eventual meltdown. This might explain why the distinction between type I and type II chondrules based on their Fe/(FeO + MgO) values, particularly in ordinary chondrites, is not always sharp.

Independent evidence suggesting that water-ice was a principal cause of oxidized iron in meteorites comes from a consideration of Δ17O in water-ice. Water-ice in the disk is thought to have been isotopically very heavy. Values of δ17O and δ18O close to +180‰, i.e., on a slope 1 line passing through Earth, were measured in tiny grains of a nanoscale-mixed magnetite-sulfide phase scattered in the matrix of the pristine carbonaceous chondrite Acfer 094 (Sakamoto et al. 2007). These grains show just how enriched in the heavy isotopes of oxygen water-ice may conceivably have been. If water-ice supplied the oxygen that stabilized iron as FeO in meteorites, as we propose, then the value of Δ17O in a meteorite might be expected to correlate with its degree of oxidation.

Such a correlation is observed. A progressive increase in oxidation state, expressed as molar FeO/(FeO + MgO) in equilibrated pyroxene or olivine, going from the E chondrites, through the H, L, and LL chondrites to the R chondrites, correlates with increasing Δ17O (Fig. 5). These equilibrated compositions are the average of a range of FeO/(FeO + MgO) and Δ17O values that would have existed in the chondrules and matrix of the original unequilibrated assemblage in each case.

Figure 5.

 Plot of mean molar FeO/(FeO + MgO) in equilibrated silicates of five chondrite groups against mean Δ17O of those groups.

The correlation line of Fig. 5 does not extend to bulk carbonaceous chondrites, which have negative values of Δ17O. Nevertheless, chondrules within individual carbonaceous chondrites have recently been found to show a good correlation between Δ17O and FeO/(FeO + MgO). The correlation has been observed in chondrules in CO chondrites (Tenner et al. 2011), CR chondrites (Schrader et al. 2011; Tenner et al. 2012), and in Acfer 094 (Ushikubo et al. 2012). Again, this correlation appears to corroborate the hypothesis that oxidation was induced by H2O derived from water-ice with high Δ17O.

The correlation between oxidation state and Δ17O is not confined to chondrites and chondrules. In ureilites, FeO/(FeO + MgO) in olivine correlates with Δ17O (Clayton and Mayeda 1996; fig. 6 of Goodrich and Delaney 2000). Explaining its cause has proven difficult; Goodrich and Wilson (2011) inferred that the correlation was somehow inherited from the state of the parent body prior to melt extraction. In line with that view, we speculate here that the ureilite parent body accreted in a heterogeneous way, with local internal variation in its content of water-ice. We envisage that the higher the ice content in a given subvolume of the parent body, the higher would have been the final values of FeO/(FeO + MgO) and of Δ17O in that subvolume, and we suspect that these values would have persisted in the solid residue following melt extraction. The same process appears also to have operated in the acapulcoite/lodranite parent body, where a similar correlation is observed (McCoy et al. 1997; Greenwood et al. 2012).

The Accretion of Chondritic Asteroids

Finally, in the context of the collision and splashing model, we speculate on the origin of a number of well-known features of chondrites that, hitherto, have largely defied a satisfactory explanation, and we discuss their wider implications. These features include: (1) the distinctive characteristics of each chondrite group, (2) the so-called “complementarity” between chondrules and matrix, (3) the depletion of metal in most chondrites, and (4) the huge mismatch between the large number of meteorites that are chondritic, and the small number of parent bodies that are chondritic.

Why Is Each Chondrite Group Distinctive?

Each of the 15 chondrite groups is thought to come from its own separate parent body, characterized by a unique suite of chondrules and other components, and having distinctive petrographic, chemical, and isotopic features (e.g., Rubin 2000, 2010; Jones et al. 2005; Scott and Krot 2007). To account for each group’s properties, Jones (2012) inferred that “multiple reservoirs of chondrite components were present in the protoplanetary disk, and that these were separated spatially, temporally, or both, such that limited mixing occurred between the separate reservoirs” before and during accretion of the respective chondrite parent bodies. Aware that turbulent mixing and radial drift in the disk are likely to have quickly destroyed a reservoir’s identity and led to the widespread homogenization of disk materials, Jones (2012) states that “the problem of maintaining such a separation over an extended time period is currently one of the biggest conundrums associated with our overall picture of the early history of the solar system.”Wood (1988) also acknowledged this problem.

As a possible solution to the puzzle, we imagine that the materials in each reservoir were stored for most of the time, not in the nebula as dispersed dust, chondrules, and other tiny objects that could easily have been mixed radially, but as loose or poorly consolidated outer layers on planetesimals whose orbital radii remained more or less constant. Thus, we envisage each reservoir as a circum-solar annulus whose planetesimal population was the main carrier of its unique chemical and isotopic peculiarities. We imagine, over time, a succession of planetesimal collisions and mergers within an individual annulus, such that each impact plume injected into its local zone a fresh batch of chondrules, along with older recycled debris from the regoliths and crusts, and that the whole mixture was promptly accreted to new, or existing, planetesimals before there was time for it to have been dispersed throughout the disk. In this way, the distinctive chemical, petrographic, and isotopic traits of each “reservoir” would have been kept largely intact and eventually preserved in each individual chondrite parent body (=chondrite group). Also, as each annulus would have remained an essentially closed chemical system, the scrambling of materials within it could not have erased its initial near-solar chemistry despite more than 2.5 Myr of processing.

Undoubtedly, there would have been some exchange of dust and chondrules between neighboring annular reservoirs, and the degree of intermixing may well have increased with the passage of time as planetesimals became fewer and larger, and underwent orbital migration. It is perhaps for this reason that particular kinds of chondrule are not always confined to single groups, but may be present in a number of groups (e.g., H, L, and LL), albeit in different proportions in each.

How many of our imagined chondrule-forming collision events are recorded in an individual chondrite? Statistical peaks in the distribution of chondrule ages (Villeneuve et al. 2009) and the clustering of oxygen isotope compositions in Mg-rich olivine in chondrules (Libourel and Chaussidon 2011) hint that the number may have been in single figures.

Why Do Chondrules and Matrix Display Compositional “Complementarity”?

In some carbonaceous chondrites, particularly in those of the CR group, the chondrules and matrix are chemically quite distinct from each other, yet the two complement each other, such that when mixed together, they have almost perfect solar, i.e., CI-like, element ratios (Hezel and Palme 2010). This remarkable relationship, dubbed “complementarity,” suggests that chondrules and matrix were formed from a single volume of starting material with solar-composition. Hezel and Palme (2010) envisage a local part of the nebula where the chemically distinct refractory chondrules and volatile-rich dust were created from the same batch of primitive dust, and then promptly recombined and added to the chondrite parent body. With this kind of local, self-contained process in mind, Palme et al. (2011) claimed that “complementarity” rules out the production of chondrules from molten planetesimals.

However, we suggest that “complementarity” need not have been the outcome of local processing, and that it can be reconciled with “splashing.” We imagine that each evolving annulus of planetesimals and dust remained an essentially closed chemical system, as outlined above. With each collision and shower of hot chondrule droplets, evaporation and recondensation would have led to a net transfer of the more volatile elements from the chondrules to the surrounding fine dust. If the chondrules and dust remained within their host annulus then, sooner or later, they would have been reunited. In this way, even with many successive phases of collision and re-accretion, a complementary relationship between chondrules and matrix would have been preserved.

Why Are Some Chondrite Groups Strongly Depleted in Siderophile Elements?

Chondrite groups other than the H, EH, and CH/CB groups are depleted to varying degrees in Fe, Ni, and other siderophile elements relative to the CI chondrites (e.g., fig. 2 in Krot et al. 2003). The depletion is particularly marked in the L and LL groups; in the LL group, Mg-normalized molar siderophile element concentrations are at roughly half their CI levels (Krot et al. 2003).

We speculate here that this well-known metal depletion is a simple consequence of planetesimal mergers. In a typical oblique-impact merger, as portrayed by Asphaug et al. (2011), the molten metallic core of the projectile body would have become substantially embedded inside the larger target body, while the ejecta plume would have been dominated by material from the projectile’s silicate mantle (Fig. 6). If, with successive collisions, the metal repeatedly showed this preference for joining the larger body of the merging pair, then the ejecta would have become progressively depleted in metal and enriched in silicate, and so would the chondrite parent bodies constructed from that ejecta. The corollary is that the target bodies, and ultimately the planets, would have a proportionately higher fraction of metal than the solar average. The process we envisage is the same, albeit on a much smaller scale, as that envisaged for the formation of the Moon by giant impact, which led to a large deficit of iron within the Moon and an increase of iron in the Earth’s core.

Figure 6.

 Cartoon showing a cross-section through a colliding pair of planetesimals following an oblique impact during accretion, based on results of Asphaug et al. (2011). The iron core of the impactor is shown largely embedded in the target body, leaving the ejecta plume depleted in iron metal and other siderophile elements.

Mars, like the Moon, may be depleted in metal as its density is low relative to that of the Earth, even after allowing for differences in the internal pressures of the two planets. This suggests that Mars possibly grew from planetesimals that were predominantly depleted in siderophile elements, like the L and LL chondrites. We speculate that such depletion of metal may be linked to the recent explanation for the small size of Mars proposed by Walsh et al. (2011). These authors postulated that Jupiter and Saturn migrated toward the Sun until their orbits reached a 3:2 mean-motion resonance and then migrated out to near their present locations. This so-called “Grand Tack” would have cleared the disk of most planetesimals as far in as 1.5 AU, leaving in its wake a mere smattering of scattered planetesimals from which Mars could grow. We speculate that the residual, thinned out suite of planetesimals may have been biased toward second-generation chondritic bodies with siderophile deficits.

Why Are Five out of Six Meteorite Falls Chondrites, When Only One in Five Parent Bodies Is Chondritic?

Chondritic meteorites are common, accounting for between 80% and 85% of observed falls, yet as we noted earlier, they appear to have been sourced from only about 20% of the 100–150 inferred meteorite parent bodies. This disparity has led to considerable debate on the composition of asteroids (Burbine et al. 2002). We suspect that the high abundance of chondrites among meteorites is a true reflection of their actual abundance in the asteroid belt, and we postulate that the large number of differentiated parent bodies that have been sampled is no indication of the abundance of these bodies in the asteroid belt today, but is a legacy of the situation that existed in the infant solar system before chondrule formation.

The “splashing” hypothesis contends that as the early molten bodies merged together and grew in size, eventually to produce planets, they released swarms of chondrules and dust, which aggregated to make chondritic asteroids. The latter, second-generation bodies, it appears, came to dominate the tiny mass of material that escaped being subsumed into planets, and they now reside, somewhat battered and brecciated, in the asteroid belt. Of the original molten bodies, perhaps Vesta alone survives intact, while the many others are represented only by fragments of iron, stony iron, and rare achondritic material not from Vesta, as either isolated pieces or assembled into asteroidal rubble piles. We imagine that by t approximately 2.5 Myr, when 26Al had lost its potency and chondrule formation was in rapid decline, most of the original molten planetesimals had already merged into larger bodies and disappeared. As John Wood (2000) so aptly put it, the beginning was “swift and violent.”

Summary and Further Work

We find it remarkable that the chronological evidence for the timing of early planetesimal meltdown (by t approximately 0.3 Myr) and the timing of chondrite accretion (after t approximately 2.5 Myr) coincide so well with the timing of these processes calculated from the inferred 26Al heat source in nebular dust. This coincidence, along with the evidence in our meteorite collections for a very large number of early molten bodies, has led us to envisage the inner solar system during its first 2 Myr as being populated with a great abundance of substantially molten planetesimals. We might name those first approximately 2 Myr, from which so little tangible evidence survives, the solar system’s “meltdown era.” As planetary embryos were probably already forming during this period, the molten planetesimals must have been continuously colliding and merging, becoming fewer in number, and growing larger in size. Within this conceptual framework for the young disk, chondrule production from the “splashing” of molten planetesimals seems an inevitable consequence, with new generations of chondritic planetesimals being spawned from the debris ejected and dispersed during mergers.

We believe that the “splashing” hypothesis can be reconciled with much of what we understand about chondrules, including their ages, chemical compositions, peak temperatures, abundances, sizes, cooling rates, indented shapes, “relict” grains, igneous rims, blebs of metal, retention of Na, presence of FeO, diversity, and large phenocrysts, as well as other issues that constrain chondrule origins such as the formation of macrochondrules, the scarcity of dust-clumps, and the need for a feasible heat source. However, we contend that several of these chondrule properties, most notably the concentration of Na in olivine, challenge the long-standing conventional interpretation of chondrules as shock-melted dust-clumps. We speculate that the bimodal division of chondrules into types I and II reflects a bimodal chemical division of molten planetesimals, such that reduced refractory planetesimals supplied type I chondrules and may also have been the source of many iron meteorites, while oxidized volatile-bearing planetesimals supplied type II chondrules and probably incorporated varying amounts of water-ice, reflected in the correlation between Δ17O values and FeO/(FeO + MgO) in some groups of meteorite. Finally, we propose that at the close of the “meltdown era,” chondritic planetesimals began to appear. We imagine that each chondritic parent planetesimal is a mixture of chondrules and debris launched and re-accreted more than once in a discrete chemically restricted annulus in the disk, to account for its unique traits, its broad solar composition, its level of metal depletion and, in some cases, its “complementarity” between chondrules and matrix.

So where do we go next in the quest to test and develop the scenario presented here?

Conventional ideas about chondrule formation are predicated on the intuitive, but now largely discredited, belief that chondrites are samples of the very first planetesimals to have formed. We submit that this conceptual framework has contributed a great deal to the understanding of disk processes, and that we would be presumptuous to claim that all chondrules were made by “splashing.” After all, some chondrule-like objects in meteorites (CAIs in particular) are widely believed to originate in a high-temperature nebular setting that was not an impact plume. Nevertheless, we believe that meteorite chronology and the likelihood of 26Al heating have brought about a new paradigm, which holds more promise than the conventional view for understanding the evolution of the nascent disk in general, and the origin of chondrules in particular.

To move forward, we need thermal and petrological models of planetesimal evolution that can accommodate collision and accretion history as well as different mechanisms of heat loss from magma oceans, not just during the heating stage but also during cooling. We need more sophisticated models of impact plumes, to try to understand how particles within them would have evolved thermally and spatially under a range of collision conditions including different impact angles, encounter velocities, planetesimal sizes, planetesimal internal structures, and regolith thicknesses. We need further precise chronology for chondrules and differentiated meteorites to strengthen the evidence for meltdown before chondrule formation. In particular, the possibility that the 26Al-26Mg chronometer may be flawed (Larsen et al. 2011), and that the 26Al heat source may have been much weaker than assumed, needs thorough investigation. Also, we need to check our model’s prediction that type II chondrules are younger than the oldest type I chondrules.

On a broader front, we need to better understand the dichotomy of chondrites and why virtually all differentiated meteorites are derived from materials that isotopically resemble ordinary and enstatite chondrites, with so few genetically related to carbonaceous chondrites (Warren 2011). We also need to investigate whether our model can shed light on the scarcity of meteorite breccias carrying both chondritic and differentiated components, whether it can explain why the cooling rates of magmatic iron meteorites are surprisingly high, or why basaltic meteorites did not crystallize at the same time as iron cores segregated (e.g., Kleine et al. 2012), and whether it can tell us anything about the virtual absence of meteorites made from olivine rock (i.e., from planetesimal mantles) that might complement irons, stony-irons, and basaltic achondrites. Finally, all the above avenues of potential future enquiry should intersect the paths being followed in the pursuit of numerical models of planetesimal accretion and orbital evolution (e.g., Bottke et al. 2006; Johansen et al. 2007; Walsh et al. 2011). We are optimistic that all approaches will eventually converge on an agreed, self-consistent picture of the first critical steps of planetary development in the solar system.

Acknowledgements— We are grateful to Stuart Agrell who inspired us both as students, to Bob Hutchison who was steadfast in his conviction that differentiated planetary bodies existed before chondrites, and to John Wood despite his conviction that they did not. We are indebted to Klaus Keil, whom we honored at the Workshop on the Formation of the First Solids in the Solar System, for his friendship and distinguished leadership in meteoritics and cosmochemistry. We acknowledge financial support from Trinity College Dublin (IS) and the NASA Cosmochemistry program (ES). We thank Alan Rubin and Alex Ruzicka for their constructive reviews.

Editorial Handling— Dr. Alexander Krot