Homogeneous Distribution of 26Al in the Solar System from the Mg Isotopic Composition of Chondrules

See allHide authors and affiliations

Science  21 Aug 2009:
Vol. 325, Issue 5943, pp. 985-988
DOI: 10.1126/science.1173907


The timing of the formation of the first solids in the solar system remains poorly constrained. Micrometer-scale, high-precision magnesium (Mg) isotopic analyses demonstrate that Earth, refractory inclusions, and chondrules from primitive meteorites formed from a reservoir in which short-lived aluminum-26 (26Al) and Mg isotopes were homogeneously distributed at ±10%. This level of homogeneity validates the use of 26Al as a precise chronometer for early solar system events. High-precision chondrule 26Al isochrons show that several distinct chondrule melting events took place from ~1.2 million years (My) to ~4 My after the first solids condensed from the solar nebula, with peaks between ~1.5 and ~3 My, and that chondrule precursors formed as early as 0.87-0.16+0.19 My after.

Models of the evolution of the early solar system rely on knowledge of the precise time scales for the physical and chemical processes that occurred in the early accretion disk and led to the formation of calcium-aluminum–rich inclusions (CAIs) and chondrules, which are the building blocks of primitive meteorites (chondrites). Short-lived 26Al [half-life (T1/2) = 0.73 million years (My)] is possibly the most accurate chronometer for the first few million years of solar system history, provided that it was homogeneously distributed in the disk. Previous studies of various meteoritic components showed that 26Al was widespread in the disk, but its level of homogeneity has never been quantified precisely. From numerous mineral 26Al isochrons in CAIs, a canonical value of ~5 × 10−5 was inferred for the initial 26Al/27Al ratio [hereafter (26Al/27Al)0] when CAIs crystallized (1). This value is slightly lower than 5.23 (±0.13) × 10−5, which was determined from the bulk 26Al isochron of Allende CAIs (2) [used in our work instead of 5.85 (±0.05) × 10−5 (3)], which indicates that the crystallization of CAIs followed closely after (50,000 years at most) the condensation of their precursors. The bulk 26Al isochron precisely defines an initial Mg isotopic composition (δ26Mg*0) for CAI precursors of −0.040 ± 0.029 per mil () (2). If 26Al and Mg isotopes were homogeneously distributed in the inner solar system, then CAIs δ26Mg*0 can be taken as the initial Mg isotopic composition of the solar system when 26Al/27Al was 5.23 (±0.13) × 10−5. Any object that formed later (when 26Al/27Al < 5.23 × 10−5) must be characterized by a more radiogenic δ26Mg*0 value. The increase of δ26Mg*0 is a function of the Al/Mg ratio of the reservoir in which the object, or its precursors, resided before the last melting event. Al-rich and ferromagnesian chondrules from unequilibrated ordinary chondrites (UOCs) and carbonaceous chondrites (CCs) show mineral 26Al isochrons with (26Al/27Al)0 < 2 × 10−5 (412); that is, much lower than that of CAIs. Because the δ26Mg*0 of chondrules has never been precisely determined, the data from chondrules cannot be compared to those of CAIs nor to the theoretical Mg isotope growth curve calculated for the solar nebula for a homogeneous distribution of 26Al and Mg isotopes.

We developed high-precision analysis of Al and Mg isotopes using the Centre de Recherches Petrographiques et Geochimiques (CRPG)–CNRS Cameca ion microprobe (ims 1270) to determine precisely (even for low Al/Mg ratios) both the slope [from which comes access to (26Al/27Al)0] and the intercept (δ26Mg*0) of mineral 26Al isochrons in chondrules. We performed multicollection analyses of 24Mg, 25Mg, 26Mg, and 27Al using four Faraday cups (figs. S1 and S2) (13) in the major silicate phases (olivine, low-Ca pyroxene, and mesostasis) of 15 chondrules [three type I (FeO-poor), 11 type II (FeO-rich), and one Al-rich] from the LL3.0 UOC Semarkona. A typical external reproducibility of ±0.005‰ (2σ) can be obtained for δ26Mg* in magnesium silicate standards (fig. S2) (13). In chondrules, internal errors (2σ) for δ26Mg* were ±0.015‰ to ±1.98‰, depending on Mg content (table S1).

The 14 ferromagnesian chondrules show well-resolved mineral isochrons (two examples are shown in Fig. 1, with the remainder shown in fig. S4), with (26Al/27Al)0 ranging from 1.619 (±0.167) × 10−5 to 3.023 (±1.240) × 10−6 and δ26Mg*0 values ranging from −0.0185 (±0.0140)‰ to 0.0047 (±0.0098)‰ for Sem-Ch138, Sem-Ch83, and Sem-Ch121, respectively (table S1). Because most Mg-rich olivines in type I chondrules might not have crystallized from the chondrule melt (14, 15), it is theoretically not correct to calculate an isochron between olivine and glass in Sem-Ch2. The (26Al/27Al)0 for Sem-Ch2 can, however, be determined precisely (fig. S4). The Al-rich chondrule (Sem-Ch4) shows constant 26Mg excesses [δ26Mg* = 0.098 (±0.016)‰, fig. S4] that are not correlated with Al/Mg ratio but would indicate that the last melting event occurred at least 4 My after CAIs; that is, when (26Al/27Al)0 < 10−6.

Fig. 1

(A and B) Backscattered electron image of two Semarkona type II chondrules: porphyritic olivine-pyroxene chondrule Sem-Ch114 and porphyritic pyroxene chondrule Sem-Ch138 (white circles show ion probe spots). The major mineral phases are olivine (ol), low-Ca pyroxene (opx) surrounded by glassy mesostasis (mes), and a few Fe-Ni metal grains (met). (C and D) Isochron diagrams obtained from 14 and 18 ion probe analyses, respectively (open circles are for low-Ca pyroxene, open triangles for olivine, and the open square for mesostasis, plotted with 2σ error bars) (fig. S4, table S1) (13).

Because of their high precision, the present chondrule Al-Mg data can be compared to the growth curve calculated for Mg isotopes in the solar nebula, assuming a simple closed-system evolution (Fig. 2). If 26Al and Mg isotopes were homogeneously distributed in the nebula, any chondrule formed at a given time t, from precursors extracted from the nebula at that time, must have (26Al/27Al)0 and δ26Mg*0 identical to those of the nebula growth curve at time t. The present ferromagnesian chondrules do plot, within errors, on the solar nebula growth curve, on which Earth also plots (Fig. 2), implying a homogenous distribution of 26Al and Mg isotopes in the inner solar nebula at the time of formation of CAIs. The clustering of chondrule data along the solar nebula growth curve implies a relative heterogeneity of ~±10% (that is, ±0.5 × 10−5 for 26Al/27Al ratios and ±0.004‰ for δ26Mg*, fig. S5). However, this is an upper bound, because part of the chondrule variability may be due to the following: (i) different closed-system evolution paths for up to ~1 My with bulk 27Al/24Mg slightly different from chondritic ratio and (ii) the possible presence within chondrule precursors of CAI-type material. One-million-year-old CAIs would have a bulk δ26Mg* of ~0.42‰ (for a bulk 27Al/24Mg of 2), implying that their fraction in chondrule precursors could not be higher than ~1% (that is, 0.004/0.42). Finally, it can be seen from Fig. 2 that the bulk 26Al isochron for CAIs with (26Al/27Al)0 = 5.85 (±0.05) × 10−5 and δ26Mg*0 = −0.0317 (±0.0038)‰ (3) is not consistent with a homogeneous distribution of 26Al, because that conclusion would imply a positive terrestrial δ26Mg* value of ~0.015‰ (16), and is not consistent with most of our chondrule data.

Fig. 2

Solar system growth curve (solid line) of Mg isotopes anchored by the (26Al/27Al)0 and δ26Mg*0 of bulk CAIs [red square, 5.23 (±0.13) × 10−5 and –0.040 (±0.029)‰] (2) and calculated for a chondritic 27Al/24Mg ratio of 0.101 (30). The green field corresponds to the growth curve that could be calculated from the bulk CAIs isochron of 5.85 (±0.05) × 10−5 and –0.0317 (±0.0038)‰ (green square) (3). All ferromagnesian chondrules (colored diamonds) plot on the solar nebula growth curve and define an error envelope (in red) at ±0.5 × 10−5 for 26Al/27Al ratios and ±0.004‰ for δ26Mg* (red arrows). Type I Sem-Ch2 is not shown, because its δ26Mg*0 cannot be determined precisely (fig. S4). The steep solid line (and its blue error envelope) show how a model age can be calculated for the extraction from the nebula of the precursors of the Al-rich chondrule (blue diamond) using its bulk 27Al/24Mg ratio of 0.7. Two sigma error bars are shown.

The 26Al and Mg isotopic homogeneity of the solar system inferred from our data are not in conflict with the existence of supracanonical 26Al/27Al ratios (up to 7 × 10−5) (17) or of large δ26Mg* variations in hibonites (up to 4‰) (18). A variability of ±0.5 × 10−5 for 26Al/27Al ratio simply implies that the fraction of components having supra-canonical ratios within CAIs and chondrules precursors was less than ~7%. Similarly, a variability of ±0.004‰ for δ26Mg* implies that the fraction of hibonite-type material was less than ~0.1%. In such a homogeneous accretion disk, the composition of the Al-rich chondrule can be simply understood as the result of an earlier extraction of its precursors from the nebula followed by a protracted closed-system evolution until the last melting event. A closed-system evolution model based on the bulk 27Al/24Mg ratio of the Al-rich chondrule (determined at 0.70 from chemical mapping by secondary electron microscopy) implies that its precursors were extracted from the nebula ~0.870.16+0.19 My after CAIs (Fig. 2). This could have happened earlier if the precursors did not remain as a closed system for Al and Mg and if, for example, Mg was added to the precursors by condensation, as has been proposed for Al-rich chondrules (19). For the ferromagnesian chondrules, which have 27Al/24Mg ratios close to the chondritic ratio (0.10 to 0.23 for our 13 chondrules), it is not possible to identify an early extraction of their precursors (for example, 1 My earlier) from the present data (with an error of ±0.01 to ±0.02‰ on δ26Mg*0), because their evolution in a closed system before melting would overlap, within errors, with the solar nebula growth curve.

Because 26Al and Mg isotopes are homogeneously distributed in the inner solar system (within approximately ±10%), different values of (26Al/27Al)0 among chondrules imply the existence of individual melting events at different times; for example, ~1.770.47+0.66 My between chondrule Sem-Ch138 and chondrule Sem-Ch83 (table S1). Our data can be used, in conjunction with data from previously published chondrules (26Al/27Al)0 (table S2), to investigate whether several major melting episodes can be identified or whether chondrule formation was a continuous process. Because data precision is variable, to make a meaningful comparison between all the data, we calculated a probability density distribution that sums the Gaussian distributions f(A26l/A27l)0 calculated for each chondrule. The distribution of the 14 ferromagnesian Semarkona chondrules clearly shows five distinct episodes (Fig. 3) at 1.2, 1.6, 2.1, 2.4, and 2.9 My after CAIs [taking (26Al/27Al)0 for CAIs of 5.23 × 10−5] (2). These five episodes are consistent with the (26Al/27Al)0 measured previously in 11 other Semarkona chondrules (Fig. 3 and table S2) (8, 10, 12); the two most prominent peaks in the distribution are at 2.1 and 2.4 My. Four other episodes of chondrule formation at 1.9, 2.2, 2.6, and 4.3 My can be tentatively identified considering all the available data for UOCs and CCs (Fig. 3 and table S2) (47, 9, 11). A Kolmogorov-Smirnov statistical test applied to these distributions confirms that the differences shown in Fig. 3 are statistically significant: The probability for the age distributions to be different between UOC and CC chondrules is 99%. In addition, even with only 25 chondrules, the age distribution of Semarkona chondrules is statistically the same as that of UOC chondrules (82%). Available Pb-Pb ages [from 0.6 (±1.6) My to 5.8 (±1.0) My after CAIs] (2023) are consistent with 26Al ages but not sufficiently precise to identify specific episodes of chondrule formation. The possible existence of a limited number of melting events over a few million years is a fundamental constraint to consider in models of chondrule formation.

Fig. 3

(26Al/27Al)0 and corresponding relative ages after the formation of CAIs (2) determined for the 15 Semarkona chondrules (yellow squares for type I chondrules, yellow circles for type II chondrules, and yellow triangle for Al-rich chondrule) compared with previous analyses of chondrules in the least-equilibrated ordinary and carbonaceous chondrites. Curves show probability density functions (n = number of data points in the distribution) of (26Al/27Al)0 in our Semarkona chondrules (yellow), all Semarkona chondrules (green), CC chondrules (blue), and UOC chondrules without our data (red). All data are available in table S2. Vertical solid lines show peaks that are present in Semarkona chondrules and often in UOC chondrules. Dotted lines show peaks that are present only in UOC chondrules and CC chondrules. The upper window corresponds to absolute Pb-Pb ages calculated from CV Allende chondrules (23) (dark and light blue squares), CR Acfer chondrules (20) (red and yellow squares), and CB Gujba (21) chondrules (dark and light green), depending on assumed age for CAIs (either 4567.2 ± 0.6 My or 4568.5 ± 0.5 My) (20, 22). One sigma error bar is shown.

The large scatter in the age distribution of chondrules (Fig. 3)—with ~10% of all chondrules formed between 0 and 1.5 My after CAIs, ~40% formed between 1.5 and 2.1 My, ~40% formed between 2.1 and 2.8 My, and ~10% formed beyond 2.8 My after CAIs—can be interpreted in two opposite ways: (i) the peak intensities reflect different magnitudes of chondrule formation at specific times and thus the major episodes of chondrule formation (for UOCs and CCs) would have taken place ~1.5 to ~3 My after CAIs; or (ii) the variable peak intensities reflect the poor efficiency of chondrule preservation in the accretion disk before the accretion of UOC and CC parent bodies. In the later scenario, chondrites would be enriched in chondrules formed shortly before accretion (which was constrained for H4 UOC at 26Al/27Al ≈ 2 × 10−7 from the metamorphic cooling of the parent body) (24). Although there is no evidence to favor one of these two extreme hypotheses over the other, it is obvious from the Al-rich chondrule that its precursors were extracted from the nebula ~0.870.16+0.19 My after CAIs and that they remained isolated in the nebula for up to ~3 My before chondrule formation. The presence of chondrules of different origins and different ages in the same few cubic centimeters of Semarkona is consistent with astrophysical models of the disk. Radial mixing by turbulence can efficiently distribute solids within 10 astronomical units in several tens of thousands of years, as long as these solids are small (millimeter or centimeter size) and are coupled to the gas (25), but the rate of destruction of these solids by accretion either by the Sun or by forming planetesimals is unknown. The observed age distribution (Fig. 3) does not exclude an early intense period (~1 My after CAIs or even before) of chondrule formation (or extraction of chondrule precursors), as was suggested by high 26Al/27Al inferred from bulk analyses of chondrules (2628).

Our results imply the following: (i) that 26Al was efficiently homogenized (within ~±10%) in the inner solar system, (ii) that no substantial 26Al (in excess of 26Al/27Al = 5 × 10−5) was produced in the disk after time “zero” as defined by the bulk isochron of CAIs, and (iii) that the 26Al-26Mg systematic has a chronological signification. Nebular models do predict an efficient homogenization of 26Al at ~±10% in the case of an external seeding of the accretion disk with 26Al injected by a nearby supernova (29). In the case of the production of 26Al by irradiation, it is not clear that such a level of homogeneity can be reached, unless most of the irradiated material is evaporated before time zero.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Tables S1 to S2


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We gratefully thank C. Rollion-Bard, D. Mangin, and M. Champenois for their implication and expertise in ion probe techniques; J. Ravaux and A. Kohler for their help with SEM analysis; and J. Marin and P. Burnard for their attentive rereading. Constructive comments by three anonymous reviewers and fruitful discussions with F. Robert and M. Gounelle were highly appreciated and greatly improved this manuscript. This work was funded by L’Agence Nationale de la Recherche grant ANR-08-BLAN-0260-02 T-Tauri, Chem, and European Research Council grant ERC 226846 Cosmochemical Exploration of the first two Million Years of the Solar System (CEMYSS). This is CRPG publication number 2001.
View Abstract

Navigate This Article