Rapid condensation of the first Solar System solids.

Chondritic meteorites are composed of primitive components formed during the evolution of the Solar protoplanetary disk. The oldest of these components formed by condensation, yet little is known about their formation mechanism because of secondary heating processes that erased their primordial signature. Amoeboid Olivine Aggregates (AOAs) have never been melted and underwent minimal thermal annealing, implying they might have retained the conditions under which they condensed. We performed a multiisotope (O, Si, Mg) characterization of AOAs to constrain the conditions under which they condensed and the information they bear on the structure and evolution of the Solar protoplanetary disk. High-precision silicon isotopic measurements of 7 AOAs from weakly metamorphosed carbonaceous chondrites show large, mass-dependent, light Si isotope enrichments (-9‰ < δ30Si < -1‰). Based on physical modeling of condensation within the protoplanetary disk, we attribute these isotopic compositions to the rapid condensation of AOAs over timescales of days to weeks. The same AOAs show slightly positive δ25Mg that suggest that Mg isotopic homogenization occurred during thermal annealing without affecting Si isotopes. Such short condensation times for AOAs are inconsistent with disk transport timescales, indicating that AOAs, and likely other high-temperature condensates, formed during brief localized high-temperature events.

The proto-Sun and its surrounding protoplanetary disk formed 4.57 Ga ago from the gravitational collapse of the dense core of a molecular cloud. Chondrites are primitive meteorites left over from the evolution of the Solar protoplanetary disk. They comprise chondrules, fine-grained matrix, and the oldest known solids of the Solar System (1), refractory inclusions in the form of Ca- and Al-rich inclusions (CAIs, millimeter- to centimeter-sized high-temperature complex assemblages of refractory oxides and silicates; ref. 2) and amoeboid olivine aggregates (AOAs, fine-grained aggregates of olivine grains associated with variable proportions of CAI-like materials; ref. 3). The mineralogy of refractory inclusions resembles that of the earliest condensates predicted to form from a cooling gas of solar composition, with AOAs appearing to be somewhat lower-temperature objects than CAIs. This has long upheld the notion that the first solids of the Solar System formed by sequential gas–solid condensation of vaporized presolar matter (4–6), although many have undergone subsequent melting events, such as those which produced the ever-mysterious chondrules (e.g., ref. 7).

However, the timescales of condensation and/or partial evaporation during formation of AOAs and non-igneous CAIs remain largely underconstrained. Mere thermodynamics offer little insight into the “prehistory” of the solids and igneous cooling rates estimates refer to melting/crystallization events (8), not original gas–solid condensation. Al–Mg dating provides only an upper bound of a few 10s or 100s of millennia for the condensation epoch (9–12), which says little on the formation timescales of individual condensates. From an astrophysical viewpoint, evolution of isotherms should be fairly slow (on 104 to 106-y timescales; ref. 13) and random motions between significant condensation fronts should take similarly long (14), so it is often assumed that condensates formed over prolonged timescales (4, 15, 16). In contrast, Sugiura et al. (17) inferred from the absence of low-Ca pyroxene on most AOAs cooling rates >0.02 K/h at the end of olivine condensation and Komatsu et al. (18) derived 50 K/h from the presence of silica in one unusual AOA.

Stable isotopes may provide insights into the kinetics of gas–solid processes (19–21). Lighter isotopes are more gas-mobile and will thus preferentially impinge a solid grain, giving it an isotopically light composition provided that back-evaporation has not erased the initial fractionation. However, CAIs commonly show mass-dependent, heavy Si, Mg, O signatures (22) diagnostic of partial evaporation of refractory liquids resulting from the melting of solar condensates (23)—whose formation conditions are hereby blurred.

Yet the relatively fine-grained and sometimes porous AOAs have never been melted and have undergone only moderate thermal annealing during the evolution of the Solar protoplanetary disk (3, 24), implying they might represent fairly pristine condensates. Moreover, because AOAs record the condensation of olivine, which represented 1/3 of CI (Ivuna-like) chondritic matter (25)—second only to ice in terms of sheer mass—they are more representative of the protoplanetary disk than CAIs. However, in situ stable isotopic characterizations of AOAs have been performed only for oxygen, an element notoriously subject to poorly understood mass-independent fractionations within the protoplanetary disk, potentially blurring any mass-dependent condensation signal (26, 27). Here, we report the oxygen, silicon, and magnesium isotopic composition of Mg-rich olivine grains in AOAs from a set of carbonaceous chondrites characterized by minimal thermal metamorphism. We use our results to quantify the conditions under which primordial dust formed in the Solar System and discuss the implications on the structure and evolution of the Solar protoplanetary disk.

Material and Methods

We surveyed all AOAs in a section each of Kaba (thin section N4075 from the Natural History Museum, Vienna, Austria), Northwest Africa (NWA) 5958 (thick section NWA 5958–1 from the Muséum national d’Histoire naturelle, Paris, France) and Miller Range (MIL) 07342 (thick section MIL 07342,9 from the NASA Antarctic Search for Meteorites program). Kaba is an oxidized Bali-like CV chondrite (28, 29). MIL 07342 is a CO chondrite. Although opaque assemblages are altered and evidence of incipient metasomatism appears on chondrule mesostasis, olivine compositions in type I chondrules (e.g., Fo94.3–99.5) show no diffusional exchange; the chondrite is deemed of type 3.0–3.2 by the Antarctic Meteorite Petrographic Description database. NWA 5958 is a C2-ung CM-like chondrite, whose type II chondrule olivine Cr content as well as opaque petrography indicate minimal thermal metamorphism (<300 °C), yet chondrule mesostases have undergone extensive aqueous alteration, which did not affect olivine (30, 31).

Scanning electron microscope observations of AOAs were performed at CRPG-CNRS (Nancy, France) using a JEOL JSM-6510 with 3-nA primary beam at 15 kV. Quantitative compositional analyses of olivine in AOAs were performed using a CAMECA SX-100 electron microprobe [at CAMPARIS (Sorbonne Université, Paris, France)]. A 20-nA focused beam accelerated to 15-kV potential difference was used for spot analyses of olivine with 20-s analysis times. The PAP software was used for matrix corrections.

We analyzed the silicon isotopic compositions of olivines by secondary ion mass spectrometry (SIMS) using the multicollector CAMECA IMS 1270 E7 at CRPG. Olivines were sputtered with an ∼5-nA primary Cs+ beam set in Gaussian mode and accelerated at 10 kV (32). Secondary negative (28–30) Si− ions were accelerated at 10 kV and analyzed in multicollection mode on 3 off-axis Faraday cups (L′2, C, and H1, respectively). Charge accumulation on the sample surface was neutralized with an electron gun. The mass-resolving power was set at M/∆M = 5,000 (slit 2) to completely resolve interferences on masses, refs. 28–30. The yields and backgrounds of the Faraday cups were calibrated at the beginning of each analytical session. Automatic centering of the transfer deflectors and mass was implemented in the analysis routine. A 10 × 10-μm2 raster was applied to the primary beam to ensure flat-bottomed pits. We used 4 terrestrial standard materials (San Carlos olivine, synthetic forsterite, quartz, diopside) to define the terrestrial fractionation line. As there is no matrix effect for Si isotopes for olivine with Mg# (= 100 × Mg/[Mg+Fe]) above 70 (32), we used the San Carlos olivine to correct for instrumental mass fractionation. Measurements typically consisted of a 90-s presputtering, automatic mass and beam centering, and 50 cycles of 4-s integrations separated by 1-s waiting times. Thus, each measurement took ∼7 min. Under these conditions, internal precision on δ29Si and δ30Si was ±0.05 through 0.30‰ and ±0.10 through 0.60‰ (2σ SE), respectively, depending on the sample, and the external reproducibility on δ29Si and δ30Si for the San Carlos olivine was ±0.10‰ and ±0.15‰, respectively (2σ SE). We also analyzed the oxygen and magnesium isotopic compositions of olivines in the same AOAs following previously established protocols (see for technical details).

Results

AOAs in NWA 5958 and MIL 07342 have compact textures characterized by continuous olivine (Fig. 1 and ) (33). Ca–Al-rich vermicular patches are altered, although diopside persists, in AOAs in NWA 5958. AOAs in Kaba show smaller olivine zones, diopside, and Ca–Al-rich areas (). Olivine in all AOAs investigated herein are Mg-rich with Mg# in the range 98.9 to 99.9 (). AOA olivine grains have O isotopic compositions with δ18O values ranging from −48.6 to −39.8‰, δ17O values ranging from −47.7 to −41.5‰, and Δ17O values ranging from −23.9 to −19.2‰ (). The silicon isotopic compositions of AOA olivines show large, mass-dependent isotopic variations with δ29Si values ranging from −5.1 to −1.0‰ and δ30Si values ranging from −9.7 to −0.2‰ (Figs. 1 and 2 and ). We did not observe any correlation between the Si compositions and the analytical location within the AOAs (Fig. 1 and ). The magnesium isotopic compositions of olivines in AOAs show slightly positive δ25Mg values ranging from 0.09 to 1.81‰ (Fig. 3 and ). We did not observe any correlation between the average Si and Mg isotopic compositions of the AOAs, nor with their elemental compositions.

Fig. 1.

A representative AOA and the Si isotopic compositions of its olivines. (A) Back-scattered electron image of AOA N1-8 from the CM-related chondrite NWA 5958 (30, 31). SIMS measurements are indicated by colored ellipses and the Si isotopic compositions associated with each analysis point. The color of the ellipses indicates the variability of the Si isotopic compositions by the range of the δ30Si value: blue, −3‰ < δ30Si < 0‰, orange, −6‰ < δ30Si < −3‰, and red, δ30Si < −6‰. (B) δ29Si – δ30Si diagram showing the mass-dependent variation observed in N1-8; MDFL, mass-dependent fractionation line.

Fig. 2.

Silicon 3-isotope plot showing the large range of mass-dependent isotopically light values measured in all analyzed AOAs in NWA 5958 (red), MIL 07342 (black), and Kaba (blue).

Fig. 3.

δ25Mg values of AOAs analyzed in NWA 5958, MIL 07342, and Kaba.

Duration and Contexts of AOA Condensation.

The mass-dependent character of the Si isotopic variations precludes their explanation by nucleosynthetic anomalies in the AOA-forming region. Provided bulk carbonaceous chondrites are a suitable proxy for the composition of the reservoir from which AOAs formed, the light isotope enrichment is also at variance with the positive 2‰/u fractionation expected from olivine-gas equilibrium (34) and inferred by ref. 15 to account for the Si isotopic variations of bulk chondrite. Partial evaporation would likewise produce heavy compositions in olivine, unless it occurred subordinately on isotopically lighter objects. In contrast, light isotopic enrichment would be expected from condensation, as the light isotopomers of gaseous SiO would collide with the growing solids at greater rates proportional to their thermal velocity (e.g., ref. 20)where k is the Boltzmann constant, T is the temperature, and m is molecular mass. We note that the greatest observed Si isotopic fractionations, around −5‰/u (Fig. 2 and ), are not far from the maximum kinetic fractionations of −11‰/u, assuming that the condensation coefficient γ has no mass dependence (). This indicates back-evaporation has had little opportunity to curtail the isotopic fractionation of the condensate. This is possible if the reservoir temperature evolved through the olivine condensation interval over a time shorter than the “SiO condensation timescale,” that is, the theoretical (e-folding) time of SiO condensation (on the available solids) in the absence of evaporation, given by (20) ():where n is the number density of condensate particles, assumed to be identical spheres of density ρ = 3 × 103 kg/m3 and radius a. If indeed the reservoir temperature evolved more rapidly than t, high supersaturation would develop and make back-evaporation subdominant. In Fig. 4, we show the distribution of δ30Si expected for different cooling timescales λ−1 (λ being defined as the decay constant of the saturation density of Mg, related by stoichiometry to SiO; see ). It is seen that only λt > 1, that is, λ−1 values shorter than t, can reproduce the range of the data. Thus, AOAs must have formed over a timescale of max(t λ−1)=t.

Fig. 4.

Probability density of the δ30Si values of the studied AOAs (blue) compared to a closed-system condensation model. The model curves correspond to different speeds of evolution of the system, measured by the dimensionless parameter λt = −t dlnn/dt (where n is the saturation concentration of Mg; ), with values of λt = 0.1 (green), 1 (blue), 10 (black), and ∞ (gray). The range of Si isotopic fractionations observed herein can only be reproduced when λt > 1, although different thermal histories are likely convolved in the measured distribution.

So what is the magnitude of t? For a steady α disk of mass accretion rate , using the pressure expression (), we havewhere α is the turbulence parameter (35), σ is the Stefan–Boltzmann constant, m = 2.34 u is the mean nebular gas molecular mass, ε is the solid/gas mass ratio, and is the solar mass. Quite insensitively to the values of the disk parameters, our calculations thus reveal a short duration on the order of a day (∼105 s) for the condensation of AOAs, or weeks if smaller values of γ apply (e.g., ref. 36 and Fig. 4). This is consistent with the cooling rate lower limit of 0.02 K/h derived by (17) from the quasisystematic absence of low-Ca-pyroxene in AOAs, as well as the 50 K/h cooling rate suggested for the silica-bearing AOA of ref. 18. This may also account for the disequilibrium gas–solid condensation effects described for AOAs in Allan Hills 77037 (24).

Such durations are, however, much shorter than the 104 to 106-y timescales of secular evolution (and cooling) of the disk as a result of progressive accretion into the Sun (13). They are also much shorter than radial transport timescales, which ref. 37 estimated on the order of 103 y (). Even vertical transport toward cooler surface layers, away from the disk midplane, would take much longer than the orbital period (38), even if winds are taken into account (). However, the smooth temperature profiles generally invoked in the inner disk are only average; heating would have been dictated by the dissipation of turbulence and likely prone to local heterogeneities, for instance in current sheets in magnetorotational instability-driven turbulence (39–41). Outside the “nominal” condensation front of forsterite, transient temperature excursions may have led to the olivine vaporization and recondensation. Perhaps late mixing with isotopically unfractionated SiO from the surroundings prevented the positive isotopic compositions predicted at the end of closed-system condensation (Fig. 4) but not observed in the data, although we cannot exclude an effect of the finite size of the ion probe beam (). In our model (illustrated schematically in Fig. 5), AOAs would have formed at greater heliocentric distances than a mere consideration of the average isotherms would suggest. Those which we now see in chondrites would have eventually escaped the evaporation/recondensation cycle of their matter during outward migration toward less-heated regions, e.g., as a result of turbulent diffusion or disk expansion (13, 42, 43). The “frozen-in” AOAs would then have eventually reached the cold regions of chondrite accretion.

Fig. 5.

Schematic representation of AOA formation. Localized heating beyond the nominal olivine condensation front may have evaporated solids, leaving refractory residues. Upon cooling, such CAI-like materials served as seeds for olivine condensation slightly beyond the nominal olivine condensation front. The rapid condensation inferred herein (day to weeks) would have produced large negative mass-dependent silicon isotopic fractionations. Agglomeration of AOAs occurred over a longer timescale on the order of a year. Subsequent thermal annealing produced the AOAs as we observe them today, homogenizing their Mg isotopic compositions without affecting their silicon isotopic compositions.

Similar processes would likely hold for higher-temperature condensates, i.e., CAIs, at shorter heliocentric distances. There may thus be a continuum between the condensation processes and subsequent reheating events that led to their remelting and crystallization, with cooling rates on the of order 0.1 to 10 K/h (8)—not to mention chondrule-forming processes as well, with cooling rates estimates spanning 1 to 3,000 K/h (ref. 44 and references therein)—the formation of Wark–Lovering rims (45–47) and/or partial evaporation resulting in heavy isotopic enrichments (19). The isolation of the ultrarefractory component presumably lost from the parent reservoir of numerous group II CAIs (48) may have likewise required similarly short timescales. Indeed, the light isotopic enrichment of (refractory) heavy rare-earth elements in such CAIs (49) allows for only limited recondensation during the partial evaporation of the now-lost ultrarefractory complement.

Condensate Agglomeration.

It is instructive to compare the condensation timescale calculated above to the grain–grain collision timescale t––that is, the mean time between 2 collisions for a given grain, given by ref. 50:where Δv is the typical particle–particle velocity. Comparing the 2 timescales, we obtainSince Δv is normalized to an upper bound (the critical fragmentation or bouncing velocity; ref. 51), t is resolvably longer than t, implying that aggregation of individual condensates, if colliding below the critical velocity for agglomeration [which might have been greater for higher temperatures (52)], generally occurred after, not during, condensation (Fig. 5). Indeed, unmelted fine-grained CAIs are composed of mineralogically zoned sub–100-µm nodules with more refractory minerals concentrated in their center, consistent with an origin as individual condensates, which encompassed the whole refractory condensation sequence before coagulation (53). Still, because t should be much shorter than radial or vertical transport timescales (14), agglomeration should have occurred largely between genetically related bodies. This accounts for the generally fairly uniform character of fine-grained CAIs at mesoscale, notwithstanding common macroscale layering (53, 54). Same holds for AOAs. One can envision that the high-temperature events that ultimately led to their final condensation likely first evaporated all but the most refractory (CAI-like) phases of any dust grains present. Olivine recondensation would have then mantled these dust grains; perhaps some olivine also condensed homogeneously (such as the 16O-rich ones found in chondrite matrices, refs. 55 and 56). These different olivine-bearing bodies would have then aggregated together. While, as argued above, aggregated bodies might have genetic relationships to one another, they may represent different thermal events, if t was longer than transport timescales or evaporation/recondensation cycles, and thus possibly variable thermal histories. Thus, individual AOAs may group bodies with similar or various ranges in Si isotopic fractionation. Depending on the recondensation events sampled by individual AOAs, different AOAs may vary in their ranges of δ30Si.

Secondary Heating of Amoeboid Olivine Aggregates.

Since δ25Mg corresponds to similar isotopomer mass ratios as δ30Si (in terms of Mg and SiO, respectively), one would expect similar fractionations for the former in AOAs (). Mg isotope fractionations as large as −14‰/u have been reported in fine-grained CAIs in Allende (57, 58). However, we observe little variation of δ25Mg in the AOAs analyzed herein (only 0.4‰ SD), although (59) reported δ25Mg down to −3‰ in some porous AOAs in Kaba. The compact textures of our AOAs suggest significant postagglomeration annealing, which may have led to the Mg isotopic homogenization given the much higher diffusivity of Mg relative to Si (60). Indeed, for a Mg diffusion coefficient of ∼10−17 m2/s at 1,400 K, isotopic homogenization over lengths of 50 µm requires only an integrated (possibly discontinuous) time of a decade or so (Fig. 5), whereas O and Si homogenization requires durations of 6 orders of magnitude longer. A timescale of a decade would agree with 1) the <60 y of isotopic exchange with 16O-poor gas for 2 Leoville AOAs (61) and 2) the effective cooling rate of 0.002 K/h estimated from the closure of Mn and Cr diffusion in AOA olivine (17). Ca from the Ca–Al-rich nodules would have diffused as well into forsterite (17, 31), accounting for its apparent Ca enrichment in compact AOAs (3, 17, 31), which our Si isotopic results no longer allow to explain by a condensation slower than for porous AOAs (17).

The closure of the Mg isotopic system in AOAs would postdate the condensation and aggregation period but would not have been as late as parent body metamorphism, because AOAs retain near-canonical Al–Mg mineral isochrons (e.g., refs. 62 and 63). The Mg isotopic composition of the bulk olivine (to which all olivines would have been homogenized) would have been also closer to the reservoir average than that of Si because Mg would be essentially completely condensed during olivine condensation, while almost half of the original SiO would still remain (ref. 6 and ).

In conclusion, our results suggest that the earliest solids of the Solar System condensed, aggregated, and were processed over short timescales, perhaps on the order of days, in a dynamic and thermally nonuniform disk. Nowhere was the primordial condensation sequence quiescent.

Data Availability.

All of the data are available in the (). All raw SIMS data, summary of the data, and output of the model can be accessed here: https://data.mendeley.com/datasets/dnxgdbzmzk/3.