182Hf-182W age dating of a 26Al-poor inclusion and implications for the origin of short-lived radioisotopes in the early Solar System

182Hf–182W age dating of a 26Al-poor inclusionand implications for the origin of short-livedradioisotopes in the early Solar SystemJesper C. Holsta, Mia B. Olsena, Chad Patona, Kazuhide Nagashimab, Martin Schillera, Daniel Wielandta, Kirsten K. Larsena,James N. Connellya, Jes K. Jørgensenc, Alexander N. Krota,b, Åke Nordlundc, and Martin Bizzarroa,1

aCentre for Star and Planet Formation and Natural History Museum of Denmark, University of Copenhagen, DK-1350 Copenhagen, Denmark; bHawaiiInstitute of Geophysics and Planetology, University of Hawaii at Manoa, Honolulu, HI 96822; and cCentre for Star and Planet Formation and Niels BohrInstitute, University of Copenhagen, DK-2100 Copenhagen, Denmark

Edited by Neta A. Bahcall, Princeton University, Princeton, NJ, and approved April 18, 2013 (received for review January 8, 2013)

Refractory inclusions [calcium–aluminum-rich inclusions, (CAIs)]represent the oldest Solar System solids and provide informationregarding the formation of the Sun and its protoplanetary disk.CAIs contain evidence of now extinct short-lived radioisotopes(e.g., 26Al, 41Ca, and 182Hf) synthesized in one or multiple starsand added to the protosolar molecular cloud before or during itscollapse. Understanding how and when short-lived radioisotopeswere added to the Solar System is necessary to assess their validityas chronometers and constrain the birthplace of the Sun. Whereasmost CAIs formed with the canonical abundance of 26Al corre-sponding to 26Al/27Al of !5 ! 10"5, rare CAIs with fractionationand unidentified nuclear isotope effects (FUN CAIs) record nucleo-synthetic isotopic heterogeneity and 26Al/27Al of <5 ! 10"6, pos-sibly reflecting their formation before canonical CAIs. Thus, FUNCAIs may provide a unique window into the earliest Solar System,including the origin of short-lived radioisotopes. However, theirchronology is unknown. Using the 182Hf–182W chronometer, weshow that a FUN CAI recording a condensation origin from a solargas formed coevally with canonical CAIs, but with 26Al/27Al of !3 !10"6. The decoupling between 182Hf and 26Al requires distinct stel-lar origins: steady-state galactic stellar nucleosynthesis for 182Hfand late-stage contamination of the protosolar molecular cloudby a massive star(s) for 26Al. Admixing of stellar-derived 26Alto the protoplanetary disk occurred during the epoch of CAIformation and, therefore, the 26Al–26Mg systematics of CAIs can-not be used to define their formation interval. In contrast, ourresults support 182Hf homogeneity and chronological significanceof the 182Hf–182W clock.

meteorite inclusions | short-lived radionuclides | Solar System formation

Meteorites and their components contain evidence for thepresence of now extinct short-lived (t1/2 < 10 Ma) radio-

nuclides (e.g., 41Ca, 26Al, 60Fe, 53Mn, and 182Hf) during theearliest stages of the Solar System’s evolution. These radio-isotopes are believed to have an external, stellar origin, and wereeither inherited from the ambient interstellar medium or injec-ted into the protosolar molecular cloud before or contempora-neously with its collapse (1). Understanding how and when theseradioisotopes were added to the nascent Solar System can con-strain the astrophysical environment where our Sun formed and,therefore, test models of Solar System formation. The oldestSolar System solids preserved in chondritic meteorites are cal-cium–aluminum-rich inclusions (CAIs), which define an absoluteage of 4,567.30 ± 0.16 Ma (2). These millimeter-to-centimeterobjects are believed to have formed as fine-grained condensatesfrom a 16O-rich gas of approximately solar composition in a re-gion with high ambient temperature (>1,300 K) and low totalpressures (!10–4 bar). This environment existed in the innermostpart of the protoplanetary disk during the early stage of its evo-lution characterized by high mass accretion rates (!10"5 M☉ y–1)to the proto-Sun (3). Formation of CAIs near the proto-Sun

is also indicated by the presence in these objects of the short-lived radioisotope 10Be formed by solar energetic particle irra-diation (4). Some of the CAIs subsequently experienced meltingand evaporation to form distinct coarser igneous inclusions,such as the compact Type A and Type B CAIs commonly ob-served in CV meteorites (carbonaceous chondrite of the Viga-rano type) (5).The majority of CAIs in unmetamorphosed chondrites contain

high abundance of radiogenic 26Mg (26Mg*), the decay product of26Al (t1/2 ! 0.7 Ma), corresponding to an inferred initial 26Al/27Alratio of !(4.5–5.5) ! 10–5 (6). Recent high-precision 26Al–26Mgsystematics of bulk CV CAIs define the so-called canonical26Al/27Al ratio of (5.252 ± 0.019) ! 10–5 (7). The uncertainty ofthe canonical 26Al/27Al ratio corresponds to !4,000 y, implyinga brief episode of condensation and melt evaporation thatresulted in Al/Mg fractionation event(s). However, it is uncertainwhether the canonical 26Al/27Al ratio reflects that of the bulkSolar System or, alternatively, only a local snapshot of theevolving inhomogeneous protoplanetary disk.A rare subset of refractory grains [platy hibonite crystals

(PLACs) and blue aggregates (BAGs), ref. 8] and inclusions(6, 9) have low initial 26Al/27Al ratios (<5 ! 10–6). Of particularinterest are the coarse-grained igneous inclusions with fraction-ation and unidentified nuclear effects (FUN CAIs, ref. 10),which, in addition to their low initial abundance of 26Al, arecharacterized by large mass-dependent fractionation effects andnucleosynthetic anomalies in several elements. These observa-tions are interpreted to reflect formation of FUN CAIs bythermal processing of presolar dust aggregates before the in-jection of 26Al and its homogenization in the protoplanetary disk(11). If this interpretation is correct, FUN CAIs can provideinsights into the timing of admixing of 26Al to the forming pro-toplanetary disk and, in turn, the origin of short-lived radio-isotopes in the early Solar System. However, a late formation of26Al-poor CAIs after decay of 26Al cannot be excluded. Indeed,CAIs are known to have experienced multistage thermal pro-cessing in the protoplanetary disk and/or on their chondriteparent bodies (5) that could have erased their radiogenic 26Mg.Thus, obtaining a robust age estimate for a FUN inclusion isa critical step toward a better understanding of the significanceof 26Al-poor inclusions.

Author contributions: J.C.H. and M.B. designed research; J.C.H., M.B.O., C.P., K.N., M.S.,D.W., K.K.L., J.N.C., and M.B. performed research; J.C.H., M.B.O., C.P., K.N., M.S., D.W.,J.N.C., J.K.J., A.N.K., Å.N., andM.B. analyzed data; and J.C.H., A.N.K., andM.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300383110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1300383110 PNAS Early Edition | 1 of 5

EART

H,ATM

OSP

HERIC,

ANDPL

ANET

ARY

SCIENCE

S

Results and DiscussionMost known FUN CAIs were discovered >30 y ago and werelargely consumed during destructive isotopic measurements. Toidentify additional FUN CAIs suitable for age dating, we con-ducted a systematic search of coarse-grained refractory inclu-sions in the Allende CV carbonaceous chondrite, given that CVchondrites contain the highest proportion of igneous CAIsamong the distinct chondrite groups. Of !220 inclusions in-vestigated, only one FUN CAI was identified on the basis of itsbulk magnesium–isotope composition. This FUN inclusion,named STP-1, is a coarse-grained igneous Type B2 CAI com-posed of melilite, spinel, Al,Ti-diopside, and anorthite. STP-1contains only minor amounts of secondary minerals (nepheline,sodalite, grossular, and monticellite), indicating that it largelyavoided secondary alteration processes. Similar to most pre-viously identified FUN CAIs (10, 12), STP-1 shows mass-dependent enrichment in the heaviest isotopes of magnesium, aswell as deficits in the mass-independent components of 26Mg(26Mg*) and 54Cr of !300 and !3,500 ppm, respectively (Table1). Trace element analysis demonstrates that STP-1 is charac-terized by a Group II rare-earth element (REE) pattern (Fig. 1),indicative of condensation from a gas depleted in the most re-fractory REEs (13). On a three-isotope oxygen diagram, compo-sitions of spinel, anorthite, hibonite, and most Al,Ti-diopsidegrains plot along a mass-dependent fractionation line witha slope of 0.52 and Δ17O value of –24 ± 1‰. Oxygen–isotopecompositions of melilite and some Al,Ti-diopside grains deviatefrom this line and show Δ17O values ranging from "17 to –4‰and from –24 to –17‰, respectively (Fig. 2). The igneous textureand the fractionated magnesium and oxygen–isotope compositionfavoring the heaviest isotopes imply that STP-1 experiencedmelt evaporation at low total pressure (14) following conden-sation of its precursor material. To define the initial abundanceof 26Al at the time of crystallization of STP-1, we have in-vestigated the 26Al–26Mg systematics of its primary minerals bysecondary ionization mass spectrometry. Multiple analyses ofspinel, Al,Ti-diopside, melilite, hibonite, and anorthite crystalsdefine an internal isochron corresponding to an initial26Al/27Al ratio of (2.94 ± 0.21) ! 10–6 (Fig. 3A), that is, muchlower than the canonical value of !5 ! 10–5.Absolute age dating of CAIs by the Pb–Pb method requires

knowledge of the uranium–isotope composition of individualinclusions (2). However, the uranium concentration in STP-1 isdepleted by a factor of !100 compared with canonical CAIs,possibly due to loss during melt evaporation under oxidizingconditions. As such, STP-1 in particular, and FUN inclusions ingeneral, may not be suited for uranium-corrected absolute Pb–Pbdating using current state-of-the-art mass spectrometry techni-ques. To define the formation age of STP-1, we have insteadinvestigated its 182Hf–182W systematics by the internal isochronapproach. With a half-life of !9 Ma, the 182Hf-to-182W decayscheme is one of the most widely used chronometers to un-derstand the timing of solid formation in the early Solar System(15). In addition, currently available data support the proposal

that the 182Hf nuclide was uniformly distributed in the earlySolar System with an initial 182Hf/180Hf ratio of (9.85 ± 0.40) !10"5 (16–18). First, the 182Hf–182W isochron of canonical CAIsintersects the carbonaceous chondritic composition, suggestingthat the precursor material of primitive asteroids that presumablyaccreted in the outer Solar System had similar initial 182Hf con-tent as CAIs (16). Second, there is excellent agreement betweenuranium-corrected Pb–Pb and Hf–W ages of rapidly cooled mag-matic meteorites (19). Lastly, CAIs with variable tungsten nucle-osynthetic anomalies define the same initial 182Hf abundance (17),indicating that the source of 182Hf is decoupled from that re-sponsible for nucleosynthetic heterogeneity in refractory elementssuch as tungsten. The Al,Ti-diopside, anorthite and melilite frac-tions separated from STP-1 define a statistically significant 182Hf–182W isochron corresponding to an initial 182Hf/180Hf ratio of(9.60 ± 1.10) ! 10–5 and intercept of –113 ± 27 ppm, when the186W/183W is used to correct for instrumental mass fractionation(Fig. 3B). Using a different isotope pair for internal normaliza-tion yields an identical initial 182Hf/180Hf within uncertainty buta different intercept of –26 ± 26 ppm (see the legend of Fig. 3).These two intercept values are distinct from the inferred SolarSystem initial μ182W value of – 351 ± 10 ppm (18), indicating thepresence of tungsten nucleosynthetic heterogeneity in STP-1. Incontrast, the internal 182Hf–182W isochron of STP-1 correspondsto a 182Hf/180Hf ratio that is identical within analytical uncertaintyto the Solar System initial 182Hf/180Hf ratio of (9.85 ± 0.40) ! 10"5inferred from canonical CAIs (18). Accepting the inferred initial182Hf/180Hf ratio of STP-1 and associated uncertainty at facevalue, we calculate an age difference of 0:33"1:47+1:67 Ma betweenformation of STP-1 and canonical CAIs, which is not consistentwith the time interval of 3:02"0:07+0:08 Ma inferred from the 26Al–26Mgsystem. Thus, we conclude that the formation age of STP-1 and, byextension, that of 26Al-poor FUN CAIs, is coeval with canonicalCAIs within the uncertainty of our measurements. These resultsare consistent with the proposal that 182Hf was homogeneouslydistributed in the solar protoplanetary disk at the time of forma-tion of the Solar System’s first solids (16–18).The contrasting initial abundances of 26Al recorded by CAIs

with identical initial 182Hf/180Hf ratios indicate that 26Al washeterogeneously distributed in the protoplanetary disk duringthe epoch of CAI formation. Therefore, the 26Al–26Mg system-atics of CAIs cannot be used to define the duration of the CAI-forming event(s). Likewise, the apparently restricted range ofinferred initial 26Al/27Al ratios defined by bulk analyses of ca-nonical CAIs from CV carbonaceous chondrites does not nec-essarily imply a homogeneous distribution of 26Al throughout thesolar protoplanetary disk during and after the epoch of CAIformation. Assessing the degree of 26Al homogeneity in the diskrequires careful comparison between the 26Al–26Mg and ura-nium-corrected Pb–Pb ages of objects with simple thermal his-tories. We note that the age difference between the formation ofcanonical CAIs and rapidly cooled angrite meteorites inferredfrom the assumption-free uranium-corrected Pb–Pb dating methodis not consistent with that suggested by the 26Al–26Mg system,

Table 1. 182Hf–182W systematics of mineral fractions and bulk Mg and Cr isotope compositions of the STP-1 FUN CAI

Sample 27Al/24Mg μ25Mg μ26Mg* μ53Cr μ54Cr 180Hf/183W μ182W (6/3) μ184W (6/3) μ182W (6/4) μ183W (6/4)

Bulk 3.18 ± 0.06 9,331 ± 20 "303 ± 10 260 ± 9 "3,579 ± 15Melilite 1.978 ± 0.119 "18 ± 27 "28 ± 15 69 ± 22 41 ± 23Anorthite 2.675 ± 0.161 35 ± 27 "27 ± 16 115 ± 35 43 ± 24Diopside 14.22 ± 0.85 622 ± 52 "31 ± 24 676 ± 58 46 ± 37

Isotope ratios expressed in the μ-notation, which reflect 106 (ppm) deviations from the terrestrial reference standard. Uncertaintiesrepresent the external reproducibility or internal precision, whichever is larger. Mg and Cr isotope data were acquired following techni-ques outlined in Bizzarro et al. (33) and Trinquier et al. (34), respectively. (6/3), internally normalized to 186W/183W; (6/4), internallynormalized to 186W/184W.

2 of 5 | www.pnas.org/cgi/doi/10.1073/pnas.1300383110 Holst et al.

supporting widespread 26Al heterogeneity in the solar pro-toplanetary disk (2, 7).The decoupling between the initial abundances of 26Al and

182Hf in early-formed refractory inclusions requires distinctstellar sources to account for the presence of these short-livednuclides in the early Solar System. In addition, the Hf–W datareported here for the STP-1 FUN CAI are most easily un-derstood in the context of a homogeneous distribution of 182Hfat the birth of the Solar System. If correct, this implies that thecarrier of this short-lived radioisotope was well mixed within theSolar System’s parental molecular cloud. Heavy r-process iso-topes such as 182Hf are thought to be synthesized during theexplosions of core-collapse supernovae of less than 11 M☉ (20).Because the lifetime of these “lower mass” massive stars is sig-nificantly longer than the typical lifetime of giant molecularclouds (21), they are not expected to contribute appreciableamounts of freshly synthesized radionuclides into star-formingregions. Therefore, in agreement with models of the chemicalevolution of the galaxy (20, 22), we infer that the initial182Hf/180Hf ratio of !1 ! 10–4 recorded by FUN and canonicalCAIs reflects long-term, steady-state galactic stellar nucleosyn-thesis before the formation of the protosolar molecular cloud. Agalactic origin for 182Hf in the early Solar System is consistentwith the view that this radionuclide was homogeneously distrib-uted in the protoplanetary disk at the time of formation of theSolar System’s first solids as inferred from earlier work and the182Hf–182W data presented here for the STP-1 FUN inclusion.In contrast with 182Hf, the Solar System’s initial inventory of

26Al is approximately 10 times higher than the background levelsof the galaxy inferred from γ-ray astronomy (23) and/or modelsof the galactic chemical evolution (20, 22), requiring late-stageaddition of stellar debris to the Solar System’s parental molec-ular cloud. A possibility is that the observed variable 26Alabundances during the epoch of CAI formation reflect hetero-geneity in the CAI precursor material (24). Such heterogeneitycould result from selective thermal processing of presolar car-riers, including the carrier(s) of 26Al, thereby generating reser-voirs enriched or depleted in presolar components (25). However,26Al-poor objects such as FUN CAIs, PLACs, and BAGs showlarge-scale nucleosynthetic heterogeneity in the stable 48Ca and50Ti nuclides, including both enrichments and depletions (8, 10,12, 26, 27), implying that the heterogeneity preserved in theseobjects is unrelated to 26Al. Moreover, the mineralogy of STP-1,coupled with its group II REE pattern and 16O-rich composition,suggest that the precursor material of this inclusion formed bycondensation from a gas of solar composition depleted in the mostrefractory REEs, similar to the majority of fine-grained CAIs.Thus, although it is possible that the variable nucleosynthetic

anomalies present in 26Al-poor objects reflect selective thermalprocessing of their precursors, we conclude that the initial 26Al/27Alof !3 ! 10–6 recorded by STP-1 represents the 26Al abundance inthe CAI-forming region when this inclusion crystallized.Solids ultimately thermally processed in the CAI-forming re-

gion are believed to represent molecular cloud material accretingto the proto-Sun from the infalling envelope via the protoplan-etary disk (28). In contrast with 182Hf, our results suggest that thecarrier of 26Al was heterogeneously distributed in the protosolarmolecular cloud and, by extension, during infall of envelopematerial to the protoplanetary disk. However, at the time offormation of the earliest Solar System solids, the innermostprotoplanetary disk is thought to have been physically well mixed(29), implying that the observed 26Al heterogeneity during for-mation of FUN and canonical CAIs may be temporal rather thanspatial. Thus, the different levels of 26Al in these inclusions couldreflect admixing of stellar derived carrier(s) of 26Al to the pro-toplanetary disk during the epoch of CAI formation. Progressiveadmixing of 26Al to the disk can be understood in the frameworkof the inside-out collapse model of prestellar cores, where theinnermost portion of the core collapses first, followed by thesuccessive outer layers (30). This interpretation requires the in-nermost part of the protostellar molecular cloud to have beendepleted in 26Al compared with the remaining cloud, and thatthe formation of FUN CAIs predates canonical CAIs, the latterbeing allowed by the 182Hf–182W age uncertainty of the STP-1FUN CAI.A galactic origin for the initial abundance of 182Hf recorded by

the FUN and canonical CAIs requires !18 Ma of free decaybetween last nucleosynthetic event that produced the heavy

YbTmErHoDyTbGdEuSmNdPrCe uLaL

REE

/CI

0.1

1.0

10.0

Fig. 1. Chondrite-normalized bulk REE abundances in the Allende STP-1FUN CAI. This CAI is characterized by a Group II REE pattern and a smallnegative Ce anomaly.

Fig. 2. Three-isotope oxygen diagram of oxygen–isotope compositions ofindividual minerals in the Allende STP-1 FUN CAI. Similarly to the majorityof FUN CAIs (31), the oxygen–isotope compositions of anorthite, spinel,hibonite, and most Al,Ti-diopside grains in STP-1 plot along a mass-dependent fractionation line defining an initial Δ17O value of ! "24‰,that is, similar to the oxygen–isotope composition of canonical CAIs andthat of the Sun (31, 32). Oxygen–isotope compositions of melilite andsome of the Al,Ti-diopside grains plot along a line with a slope of !1,suggesting subsequent isotope exchange with a 16O-depleted gaseousreservoir. In contrast with most FUN CAIs from CV chondrites character-ized by 16O-poor compositions of melilite and anorthite, melilite in STP-1 showsa range of Δ17O values, whereas anorthite is uniformly 16O-rich. These obser-vations indicate that STP-1 is more pristine than all previously known FUN CAIs.

Holst et al. PNAS Early Edition | 3 of 5

EART

H,ATM

OSP

HERIC,

ANDPL

ANET

ARY

SCIENCE

S

r-process nuclides and formation of the Solar System (22). Thistime interval, however, is not compatible with the initial26Al/27Al ratio of !3 ! 10"6 defined by the STP-1 FUN CAI,accepting that this value represents that of the 26Al-poor regionof the protosolar molecular cloud before the last addition offreshly synthesized 26Al present in canonical CAIs. If correct,this interpretation suggests that the protosolar molecular cloudwas part of a giant molecular cloud complex (GMC) chemicallyenriched in freshly synthesized matter by earlier generation(s)of massive stars to account for the level of 26Al present inFUN CAIs. Whether our Sun formed during the early or lateevolutionary stages of the GMC requires knowledge of the60Fe/26Al value of canonical and FUN CAIs, given that mostof the 26Al in the early life of the GMC will be synthesized and

ejected by the stellar winds of massive stars where 60Fe isnot produced.

Materials and MethodsWe conducted a systematic search for new FUN CAIs by investigating theMg–isotope composition of numerous igneous CAI-like objects in cut sections ofan !3 kg fragment of the Allende CV carbonaceous chondrite. All igneousCAI-like inclusions of appropriate size were sampled with a computer-assis-ted microdrilling device fitted with 300-μm–diameter diamond-coatedmicrodrills. The sampled material was digested using hydrofluoric (HF)–HNO3 acid mixtures and, after complete dissolution, a 5% aliquot of thesample was taken for Al/Mg ratio determination to 5% accuracy usinga ThermoFisher X-Series II inductively coupled plasma source mass spec-trometer (ICPMS) at the Centre for Star and Planet Formation in Copen-hagen. The magnesium from samples with Al/Mg ratios typical of CAIs waspurified by ion-exchange chromatography and its isotopic compositionanalyzed using a ThermoFisher Neptune multiple collector inductivelycoupled plasma source mass spectrometer (MC-ICPMS) at the Centre forStar and Planet Formation in Copenhagen, following protocols outlined inBizzarro et al. (33). One inclusion was typified by a resolvable deficit in26Mg* of !300 ppm as well as a stable Mg–isotope composition enriched inthe heavy isotopes by !1%/amu. This inclusion, named STP-1, was classi-fied as a FUN CAI and selected for further analysis. Present on the surfacesof two 3-mm–thick sections, the STP-1 FUN CAI is a spherical inclusion of!10 mm in diameter. Once the inclusion was liberated from the Allendemeteorite, polished sections were made from the extracted material forpetrographic characterization, mineral chemistry and in situ 26Al–26Mgand O–isotope work.

Elemental maps of sections and electron microprobe analyses of individualminerals were performed with the University of Hawaii (UH) field-emissionelectron JEOL JXA-8500F operated at 15-kV accelerating voltage, 15-nA beamcurrent, and fully focused beam using five wavelength spectrometers. TheSTP-1 inclusion is a coarse-grained igneous CAI composed of pure anorthite,gehlenitic melilite (Åk6"28), and igneously zoned Al,Ti-diopside (Al2O3 = 17.7"28.5 wt %, TiO2 = 0.03"8.7 wt %), all poikilitically enclosing euhedralcompositionally pure spinel grains. Lath-shaped hibonite grains and spinel–hibonite intergrowths occur in the outermost portion of the inclusion. Thehibonite grains have low contents of MgO (0.2"1.7 wt %) and TiO2 (0.09"3.2wt %). No multilayered Wark–Lovering rim sequence is observed aroundSTP-1. The oxygen isotope composition and Al–Mg systematics of primaryminerals in STP-1 was investigated using the UH Cameca ims-1280 ion mi-croprobe based on techniques described in SI Materials and Methods.

Following removal from the Allende slab and cleaning, bulk fragments ofSTP-1 were preserved for bulk isotope and elemental analyses. The remainingmaterial was gently crushed in an agate mortar under distilled ethanol andminerals were handpicked under binocular microscopes in both plain andback lighting. REE abundances were determined on the X-Series II ICPMS froma bulk aliquot of STP-1, using sample-standard bracketing techniques. Thechromium isotope composition of a separate bulk aliquot was determinedbased on previously published techniques (34) using a ThermoFisher Tritonthermal ionization mass spectrometer at the Centre for Star and PlanetFormation. Following handpicking, the anorthite, melilite, and Al,Ti-dio-spide fractions were rinsed in distilled ethanol followed by 0.02 M HNO3 inan ultrasonic bath for 15 min. After complete dissolution in mixtures ofconcentrated HF:HNO3:H2O2, a 15% aliquot of each sample was extractedand spiked with a mixed 180Hf/186W tracer for elemental abundance deter-minations. Tungsten was purified by ion-exchange chromatography ina two-step procedure inspired from Fritz et al. (37) and Strelow et al. (38).Samples were dissolved in 0.25 M HNO3 + 0.1 M HF + 0.1% H2O2 and loadedon a column containing 2–4 mL of AG50W-X8 resin, and W was eluted alongwith high-field strength elements with 1–2 column volumes (c.v.) of 0.25 MHNO3 + 0.1 M HF + 0.1% H2O2 followed by 2.5 c.v. of 0.1 M HF. After drydown, samples were dissolved in 1 M HF and loaded on a column containing1 mL AG1-X4 resin. Residual elements were eluted sequentially with 5 c.v. of1 M HF, 5 c.v. of 2 M HCl + 0.1% H2O2 and 2 c.v. of 6 M HCl + 0.01 M HF,whereas the W was recovered with 4 c.v. of 6 M HCl + 1 M HF.

Following tungsten purification, isotope data were acquired in staticmode using the ThermoFisher NeptuneMC-ICPMS at the Centre for Star andPlanet Formation. Samples were converted to nitrate form, dissolved ina 2% HNO3 solution containing traces of HF, and introduced into theplasma source by means of an Aridus II desolvating nebulizer. Mass frac-tionation was corrected with the exponential law using the 186W/183W =1.98594 (39). Samples were analyzed only once, and the ratios are repor-ted as relative deviations from the mass-bias corrected NIST 3163 tungstenreference material in the μ-notation (106 deviations). A blank correction

melilite

diopside

hibonite

spinel

anorthite

bulk

A

27Al/24Mg

-0.8-0.6-0.4-0.20.0

0.20.4

2520151050

melilite

diopside

anorthite

800

600

400

200

0

30

20

10

0

B

120010008006004002000

1614121086420

26M

g* (‰

)

26Al/27Al = (2.94 ± 0.21) ! 10–6

!26Mg0 = –0.371 ± 0.010M.S.W.D. = 0.29

182Hf/180Hf = (9.60 ± 1.10) ! 10–5

182W0 = –113 ± 27M.S.W.D. = 0.73

182 W

(ppm

)

Fig. 3. Internal mineral 26Al–26Mg (A) and 182Hf–182W (B) isochron dia-grams for the Allende STP-1 FUN CAI. The well-defined 26Al–26Mg isochronbased on spinel, Al,Ti-diopside, melilite, anorthite, and hibonite shows noevidence for late-stage disturbance, consistent with its pristine mineralogyand the 16O-rich composition of most primary phases. Thus, we infer thatthe 26Al–26Mg isochron defines the initial 26Al/27Al ratio in STP-1 at thetime of its crystallization. M.S.W.D., mean square of weighted deviations.Using the 186W/184W instead of the 186W/183W ratio for internal normaliza-tion returns an initial 182Hf/180Hf value of (9.22 ± 1.1) ! 10"5 and an interceptof –26 ± 26 ppm (M.S.W.D. = 0.31). This 182Hf/180Hf value is identical withinanalytical uncertainty to that obtained using the 186W/183W for internalnormalization, although marginally lower. However, repeated analysis ofa number of distinct aliquots of column-processed BCR-2 (Basalt, ColumbiaRiver) rock standard and the Allende carbonaceous chondrite usinga quantity of W comparable to that present in the mineral fractions of theSTP-1 FUN CAI indicates a superior external reproducibility when the186W/183W is used for internal normalization. Therefore, our preferredapproach is to use the 186W/183W for internal normalization, in agree-ment with earlier studies (17, 18, 35, 36).

4 of 5 | www.pnas.org/cgi/doi/10.1073/pnas.1300383110 Holst et al.

was applied to all samples, and the uncertainty of this correction ispropagated in the final uncertainties of the isotope measurementsreported in Table 1. Full analytical details of procedures used for Hf/W andW isotope measurements, as well as all other data reported in this paper,are presented in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank G. J. Wasserburg for comments on anearlier version of this paper. We also thank Andy Davis and two anonymousreferees for their constructive comments, which improved the quality of thispaper. The Centre for Star and Planet Formation is financed by the DanishNational Research Foundation (Grant DNRF97).

1. Goswami JN (2004) Short-lived radionuclides in the early solar system: The stellarconnection. New Astron Rev 48:125–132.

2. Connelly JN, et al. (2012) The absolute chronology and thermal processing of solids inthe solar protoplanetary disk. Science 338(6107):651–655.

3. D’Alessio P, Calvet N, Woolum DH (2005) Chondrites and the Protoplanetary Disk, edsKrot AN, Scott ERD, Reipurth B (Astrophysical Society of the Pacific, San Francisco), Vol341, pp 353–372.

4. McKeegan KD, Chaussidon M, Robert F (2000) Incorporation of short-lived 10Be ina calcium-aluminum-rich inclusion from the allende meteorite. Science 289(5483):1334–1337.

5. MacPherson GJ (2003) Treatise on Geochemistry 1, ed Davis AM (Elsevier, Amsterdam),pp 201–246.

6. Krot AN, et al. (2009) Origin and chronology of chondritic components: A review.Geochim Cosmochim Acta 73:4963–4997.

7. Larsen KK, et al. (2011) Evidence for magnesium isotope heterogeneity in the solarprotoplanetary disk. Astrophys J 735:L37.

8. Liu M-C, et al. (2009) Isotopic records in CM hibonites: Implications of timescales formixing of isotope reservoirs in the solar nebula. Geochim Cosmochim Acta 73:5051–5079.

9. Makide K, et al. (2011) Heterogeneous distribution of 26Al at the birth of the solarsystem. Astrophys J 733:L31.

10. Wasserburg GJ, Lee T, Papanastassiou DA (1977) Correlated O and Mg isotopicanomalies in Allende inclusions: II. Magnesium. Geophys Res Lett 4:299–302.

11. Sahijpal S, Goswami JN (1998) Refractory phases in primitive meteorites devoid of 26Aland 41Ca: Representative samples of first solar system solids? Astrophys J 509:L137–L140.

12. Papanastassiou DA, Brigham CA (1989) The identification of meteorite inclusions withisotope anomalies. Astrophys J 338:L37–L40.

13. Boynton WV (1975) Fractionation in the solar nebula: Condensation of yttrium andthe rare earth elements. Geochim Cosmochim Acta 39:569–584.

14. Richter FM, Janney PE, Mendybaev RA, Davis AM, Wadhwa M (2007) Elemental andisotopic fractionation of Type B CAI-like liquids by evaporation. Geochim CosmochimActa 71:5544–5564.

15. Kleine T, et al. (2009) Hf-W chronology of the accretion and early evolution of as-teroids and terrestrial planets. Geochim Cosmochim Acta 73:5150–5188.

16. Kleine T, Mezger K, Palme H, Scherer E, Münker C (2005) Early core formation inasteroids and late accretion of chondrite parent bodies: Evidence from 182Hf-182W inCAIs, metal-rich chondrites and iron meteorites. Geochim Cosmochim Acta 69:5805–5818.

17. Burkhardt C, et al. (2008) Hf-W mineral isochron for Ca,Al-rich inclusions: Age of thesolar system and the timing of core formation in planetesimals. Geochim CosmochimActa 72:6177–6197.

18. Burkhardt C, Kleine T, Dauphas N, Wieler R (2012) Nucleosynthetic tungsten isotopeanomalies in acid leachates of the Murchison meteorite: Implication for hafnium-tungsten chronometry. Astrophys J 753:L6.

19. Kleine T, Hans U, Irving AJ, Bourdon B (2012) Chronology of the angrite parent body andimplications for core formation in protoplanets. Geochim Cosmochim Acta 84:186–203.

20. Huss GR, Meyer BS, Srinivasan G, Goswami JN, Sahijpal S (2009) Stellar sources of theshort-lived radionuclides in the early solar system. Geochim Cosmochim Acta 73:4922–4945.

21. Murray N (2011) Star formation efficiencies and lifetimes of giant molecular clouds inthe Milky Way. Astrophys J 729:133.

22. Wasserburg GJ, Busso M, Gallino R, Nollett KM (2006) Short-lived nuclei in the earlysolar system: Possible AGB sources. Nucl Phys A 777:5–69.

23. Diehl R, et al. (2006) Radioactive 26Al from massive stars in the Galaxy. Nature 439(7072):45–47.

24. Wood JA (1998) Meteoritic evidence for the infall of large interstellar dust aggregatesduring formation of the solar system. Astrophys J 503:L101–L104.

25. Trinquier A, et al. (2009) Origin of nucleosynthetic isotope heterogeneity in the solarprotoplanetary disk. Science 324(5925):374–376.

26. Fahey A, Goswami JN, McKeegan KD, Zinner E (1985) Extreme 50Ti enrichments inprimitive meteorites. Astrophys J 196:L17–L20.

27. Ireland T (1990) Presolar isotopic and chemical signatures in hibonite-bearing re-fractory inclusions from the Murchison carbonaceous chondrite. Geochim CosmochimActa 54:3219–3237.

28. Williams JP, Cieza LA (2011) Protoplanetary disks and their evolution. Annu Rev As-tron Astrophys 49:67–117.

29. Boss AP (2011) Evolution of the solar nebula. IX. Gradients in the spatial heteroge-neity of the short-lived radioisotopes 60Fe and 26Al and the stable oxygen isotopes.Astrophys J 739:61.

30. Myers PC, Evans NJ II, Ohashi N (2000) in Protostars and Planets IV, eds Boss AP,Russell SS (Univ of Arizona Press, Tucson), pp 217–245.

31. Krot AN, et al. (2010) Oxygen isotopic composition of the sun and mean oxygenisotopic composition of the protosolar silicate dust: Evidence from refractory in-clusions. Astrophys J 713:1159–1166.

32. McKeegan KD, et al. (2011) The oxygen isotopic composition of the Sun inferred fromcaptured solar wind. Science 332(6037):1528–1532.

33. Bizzarro M, et al. (2011) High-precision Mg-isotope measurements of terrestrial andextraterrestrial material by HR-MC-ICPMS – implications for the relative and absoluteMg-isotope composition of the bulk silicate earth. J Anal At Spectrom 26:565–577.

34. Trinquier A, Birck J-L, Allègre CJ (2008) High-precision analysis of chromium isotopesin terrestrial and meteorite samples by thermal ionization mass spectrometry. J AnalAt Spectrom 23:1565–1574.

35. Scherstén A, Elliott T, Hawkesworth C, Norman M (2004) Tungsten isotope evidencethat mantle plumes contain no contribution from the Earth’s core. Nature 427(6971):234–237.

36. Schérsten A, Elliott T, Hawkesworth C, Russell S, Masarik J (2006) Hf-W evidence forrapid differentiation of iron meteorites parent bodies. Earth Planet Sci Lett 241:530–542.

37. Fritz JS, Garralda BB, Karraker SK (1961) Cation exchange separation of metal ions byelution with hydrofluoric acid. Anal Chem 33:882–886.

38. Strelow FWE, Weinert CHSW, Eloff C (1972) Distribution coefficients and anion ex-change behavior of elements in oxalic acid–hydrochloric acid mixtures. Anal Chem 44:2352–2356.

39. Völkening J, Köppe M, Heumann KG (1991) Tungsten isotope ratio determinations bynegative thermal ionization mass spectrometry. Int J Mass Spectrom Ion Process 107:361–368.

Holst et al. PNAS Early Edition | 5 of 5

EART

H,ATM

OSP

HERIC,

ANDPL

ANET

ARY

SCIENCE

S

Supporting InformationHolst et al. 10.1073/pnas.1300383110SI Materials and MethodsIdentification of FUN Inclusions. An !3 kg fragment of the AllendeCV meteorite (carbonaceous chondrite of the Vigarano type)was cut into numerous 3-mm–thick sections using a 250-μmdiamond-coated wire saw that was operated dry to expose the freshsurface of the meteorite. High-resolution photographic imageswere produced for each section and these were characterized vi-sually to identify all igneous calcium–aluminum-rich (CAI)-likeinclusions. Igneous CAI-like inclusions greater than 2 mm in di-ameter and present on at least two sections were sampledusing a Micromill sampling device fitted with 300-μm–di-ameter diamond-coated microdrills. The sampled materialwas transferred to Savillex beakers and digested using HF–HNO3 acid mixtures on a hotplate at 130 °C for 48 h. Aftercomplete dissolution, a 5% aliquot of the sample was taken forAl/Mg ratio determination to 5% accuracy using the Thermo-Fisher X-Series II inductively coupled plasma source mass spec-trometer (ICPMS) at the Centre for Star and Planet Formation inCopenhagen, to discard inclusions with low Al/Mg ratios such aschondrules and amoeboid olivine aggregates. All samples with Al/Mg ratio higher than 1.5 were classified as potential CAIs, whichconstitute about 50% of the sampled inclusions. For these sam-ples, we purified the magnesium by ion-exchange chromatogra-phy and analyzed its isotopic composition using a ThermoFisherNeptune multiple collector inductively coupled plasma sourcemass spectrometer following protocols outlined in Bizzarro et al.(1). Out of !220 bona fide CAIs analyzed, only one inclusion wastypified by a resolvable deficit in 26Mg* of !300 ppm as well asa stable Mg–isotope composition enriched in the heavy isotopes by!1%/amu, which is characteristic of many known fractionationand unidentified nuclear effects (FUN) inclusions. Based on thisobservation, this inclusion, named STP-1, was classified as a FUNCAI and selected for further analysis. Present on the surfaces oftwo 3-mm–thick sections, the STP-1 FUN CAI is a spherical in-clusion of !10 mm diameter. It was liberated from the Allendematrix using a variable-speed Dremel fitted with either cone-shapeddiamond-coated cutting tools or dental drill bits. Once the in-clusion was liberated, the easily identified dark matrix was carefullyremoved from all surfaces using the Dremel. A !200-μm–thicksection was made from the extracted material for petrographiccharacterization and in situ 26Al–26Mg and O–isotope work.

X-Ray Elemental Mapping and Electron Probe Microanalysis. STP-1was exposed in three sequential polished sections (1–3), whichwere studied in reflected light using optical microscopy. Part ofthe central portion of the CAI was possibly lost during cutting.Each section was mapped in Mg, Ca, Al, Si, Ti, Na, K, Cl, and FeK! X-rays with a resolution of 5 μm/pixel using the University ofHawaii (UH) field-emission electron microprobe JEOL JXA-8500F operating at 15-kV accelerating voltage, 100-nA beamcurrent and 3-μm beam size, and studied in backscattered elec-trons with 25-nA beam current and fully focused beam. To in-vestigate the distribution of primary and secondary minerals inSTP-1, (i) Mg, Ca, and Al, (ii) Cl, Na, and Mg, and (iii) Ti, Ca,and Al X-ray maps were combined using a red-green-blue colorscheme. These elements and color scheme allow one to distin-guish spinel, hibonite, melilite, Al,Ti-diopside, anorthite, neph-eline, and sodalite (Figs. S1, S3, and S4). Electron microprobeanalyses of individual minerals were performed with the JEOLJXA-8500F operated at 15-kV accelerating voltage, 15-nA beamcurrent, and fully focused beam using five wavelength spec-trometers. For each element, counting times on both peak and

background were 30 s. Minerals with known chemical compo-sitions were used as standards. Matrix effects were correctedusing procedures described in Pouchou and Pichoir (2).

Petrology, Mineralogy, and Mineral Chemistry of the STP-1 FUN CAI.STP-1 is a coarse-grained igneous CAI composed of pure an-orthite, gehlenitic melilite (Åk6"28), and igneously zoned Al,Ti-diopside (Al2O3 = 17.7"28.5 wt %, TiO2 = 0.03"8.7 wt %), allpoikilitically enclosing euhedral compositionally pure spinelgrains (Table S1). Lath-shaped hibonite grains and spinel–hibonite intergrowths occur in the outermost portion of the in-clusion. The hibonite grains have low contents of MgO (0.2"1.7wt %) and TiO2 (0.09"3.2 wt %). The Ti-poor compositionsobserved in some rare pyroxene grains are found in crystals lo-cated at the boundary between melilite and anorthite (dmis-teinbergite) and appears to have crystallized at the eutectic pointbetween these minerals from the last portion of melt. This prob-ably explains the low Ti abundances in these pyroxenes. No mul-tilayered Wark–Lovering rim sequence is observed around STP-1(Figs. S2 and S5). The inclusion experienced only a small degree ofsecondary alteration resulting in formation of nepheline, sodalite,and Fe-bearing Al-rich, Ti-poor pyroxene (FeO, 2.5"6.3 wt %,Al2O3, 5.1"16.2 wt %, TiO2, 0.10"0.27 wt %), and enrichment ofspinel in FeO (up to 19.5 wt %) in its peripheral portion (Table S1,Figs. S1–S5). In addition, melilite crystals are cross-cut by thinveins of grossular, Al-rich, Ti-poor diopside, and Na-bearing pla-gioclase (0.35"0.89 wt % Na2O). Primary coarse anorthite crystalsshow no evidence for replacement by secondary minerals, butdisplay cleavage planes, occasionally filled by grossular (Fig. S5D).

Bulk Trace Elements Determination (Rare-Earth Element and Uranium).Rare-earth element (REE) abundances were determined on theThermo X-Series II ICPMS from a separate !5.5-mg bulk aliquotof STP-1, of which 0.5% of the total solution, dissolved in 400 μL2% HNO3, was used for the analysis. The sample was bracketedby analyses of a synthetic REE standard solution with a concen-tration of 1 ppb, and the data were reduced in Iolite (3) using theTraceElements data reduction scheme with the “semi-quantita-tive” setting. Based on measurements under similar conditions ofthe BCR-2 (Basalt, Columbia River) and BHVO-2 (Basalt, Ha-waiian Volcanic Observatory) rock standards, we estimate theaccuracy of our REE results to be 23% (2 SD), apart for Eu, Gd,Tb, Dy, and Ho for which we estimate the relative accuracy to be45% (2 SD) because of the low count rates obtained for theseelements. The REE data of a bulk rock aliquot of STP-1, reportedin absolute concentration as well as normalized to the CI chon-drite data of Palme and Jones (4), are presented in Table S2.The uranium content of an object of known age can be cal-

culated from the amount of radiogenic 206Pb present today,which, in turn, is determined from the total amount of Pb, its Pbisotopic composition, and the initial Pb isotopic composition atthe time of formation. In a companion study, we analyzed the Pbisotopic compositions of a number of fractions of the STP-1FUN CAI spiked with an equal atom 202Pb–205Pb tracer ofknown concentration following a stepwise cleaning and dissolu-tion procedure of a 23.0-mg fragment of this object. Eight of the14 fractions analyzed defined a linear array in 204Pb/206Pb vs.207Pb/206Pb space, with the remaining 6 fractions plotting slightlybelow the line. The line regresses through the isotopic compo-sition of the Solar System as estimated by Tatsumoto et al. (5).Points falling below the line are attributed to a small amountof terrestrial Pb contamination in these fractions that was not

Holst et al. www.pnas.org/cgi/content/short/1300383110 1 of 12

removed during the precleaning steps. Subtracting sufficientterrestrial Pb from these 6 fractions to transpose them onto thelinear array in 204Pb/206Pb vs. 207Pb/206Pb space defined by the 8fractions results in Pb isotopic compositions for all 14 fractionsthat represent binary mixtures of initial Pb and radiogenic Pb.An estimate of 5.36 pg of radiogenic 206Pb in the fragment an-alyzed is calculated by arithmetically combining the Pb in all 14fractions and subtracting the initial Pb component (based on the204Pb/206Pb ratio of the Solar System initial). This corresponds toan average concentration of 0.38 ppb of U in STP-1. This con-centration is 91–140 times lower than the U contents of threerecently analyzed canonical CAIs from the chondrite Efremovka(6). Given the limited amount of material available for STP-1,this concentration is well below the minimum required fora sufficiently precise U isotopic measurement to calculatea meaningful absolute Pb–Pb age.

Analytical Protocols for in Situ Oxygen Isotope Measurements. Oxy-gen–isotope compositions of primary minerals in STP-1 weremeasured in situ by secondary ionization mass spectrometry(SIMS) with theUHCameca ims-1280 ionmicroprobe. An!1 nACs+ primary ion beam was focused to a diameter of !7"10 μmandrastered over 7 ! 7 mm2 area on the sample for data collection.Secondary ions of 16O–, 17O–, and 18O– were measured simulta-neously in multicollection mode with the magnetic field controlledby an NMR probe. 16O– and 18O– were measured by multicollectorFaraday cups (FCs) with low mass-resolving power (MRP !2,000),whereas 17O– was measured using the axial monocollector electronmultiplier (EM) with MRP !5,600, sufficient to separate the in-terfering 16OH" signal. To correct for instrumental mass-frac-tionation effects, Burma spinel, Miyakejima anorthite, terrestrialdiopside, and San Carlos olivine were used as standards. Reporteduncertainties include both the internal precision of an individualanalysis and the external reproducibility for standard measure-ments during a given analytical session. Regions sputtered duringO–isotope measurements were photographed before and aftermeasurements. Oxygen–isotope compositions of minerals analyzedare listed in Table S3 and shown in Fig. 2 of the main paper. Be-cause only melilite and several analyses of Al,Ti-diopside in STP-1deviate from the mass-dependent fractionation line defined byspinel, hibonite, anorthite, and most analyses of Al,Ti-diopside, allSIMS spots in melilite and Al,Ti-diopside are shown in Figs. S6–S8.

Analytical Protocols for in Situ 26Al–26Mg Measurements. Magne-sium- and aluminum–isotope compositions of the primary mineralsin STP-1 were measured in situ with the UH Cameca ims-1280 ionmicroprobe using primary 16O" ion beam. Two analytical proce-dures were used to measure 26Al"26Mg systematics. Minerals withhigh Al/Mg ratios (anorthite and hibonite) were analyzed by mag-netic field switching using EM and FC detectors for magnesiumisotopes and 27Al+, respectively. 27Al+ was measured simulta-neously with 25Mg+. Primary beam currents of 300 and 60 pA wereused for anorthite and hibonite, respectively. A spot size was!5"10μm.TheMRPs forMg–isotopes and 27Al+were!3,800 and!2,300,respectively, sufficient to separate interfering ions. Automatedcentering of the secondary beam in the field aperture of the massspectrometer, high-voltage offset control, and mass-peak center-ing were applied before each measurement. Minerals with low Al/Mg ratios (spinel, melilite, and Al,Ti-diopside) were measuredwith multicollection mode using four FC detectors. The primarybeam currents were set to !5, 8, and 12 nA for spinel, Al,Ti-di-opside, andmelilite, respectively. A spot size was!20"30 μm. Themagnetic field was controlled by the NMR probe. The MRP wasset to!2,300. Automated centering of secondary beam in the fieldaperture and high-voltage offset control were applied before eachmeasurement. Instrumental mass fractionation was corrected bystandard-sample bracketing by comparing each measurementwith the isotope ratios measured in the appropriate terrestrial

standards including Burma spinel, Madagascar hibonite, Miya-kejima anorthite, synthetic melilite glass, and synthetic Al,Ti-diopside glass. Excess or deficit of radiogenic 26Mg (δ26Mg*) wascalculated using an exponential law with a mass fractionationexponent of 0.511. The reported uncertainties include both theinternal precision of an individual analysis and the external re-producibility for standard measurements during a given analyticalsession. The relative sensitivity factors for aluminum and mag-nesium were determined from the 27Al+/24Mg+ ratios measuredby SIMS and the Al/Mg ratios measured previously by electronmicroprobe for each standard mineral. The 27Al/24Mg and Mgisotope data are reported in Table S4 in the δ-notation, whichreflect permil deviations from the terrestrial composition.

Analytical Protocols for Tungsten Isotope Measurements. Followingremoval from the Allende slab and cleaning, the bulk STP-1inclusion was gently crushed in an agate mortar under distilledethanol, and minerals were handpicked under binocular micro-scopes in both plain and back lighting. Minerals were identifiedbased on their optical properties, mainly surface relief and color.The crushing and handpicking were done repeatedly, at each turnincreasing the mineral purity of the separates. After each crushingcycle, a Nd hand magnet was passed over the separates but in nocase were any magnetic grains detected. Fines were filtered out ofeach mineral separate using a nylon sieve paper with 53-μmmesh.The crushed and handpicked mineral fractions were then suc-cessively rinsed in distilled ethanol followed by 0.02 M HNO3 inan ultrasonic bath for 15 min. At this point, the samples weredried and weighed in 1.5-mL centrifuge tubes of known weight.The anorthite, melilite, and Al,Ti-diopside fractions weighed45.1, 21.2, and 10.8 mg, respectively. The samples were dissolvedin a 5:4:1 mixture of concentrated HF:HNO3:H2O2 for 2 d ona hotplate at 150 °C. After drying at no more than 100 °C, theywere oxidized in 4:1 HNO3:H2O2 to remove organic materialand Os. Keeping the temperature low during sample drying wasfound to be critical to obtain high W yields, as a fraction of theW forms a volatile fluoride complex during sample digestion.The oxidation step was followed by total dissolution in 6 M HCl +0.06 M HF and samples were centrifuged to confirm full disso-lution. At this step, a 15% aliquot of each sample was extractedand spiked with a mixed 180Hf/186W tracer for elemental abun-dance determinations. After drying down, samples were con-verted to nitrate form and dissolved in 0.25 M HNO3 + 0.1 MHF + 0.1% H2O2. This solution was fluxed on a hotplate for 1 dat 100 °C to enable thorough oxidation of Cr. Subsequent puri-fication of W was achieved on cation and anion exchange resinsusing a recipe modified from Fritz et al. (7) and Strelow et al. (8).The solution was loaded on a 2–4-mL AG50W-X8, 200–400mesh column and W along with Al, Hf, Ti, Zr, and Mo waseluted with 1–2 column volumes (c.v.) of 0.25 M HNO3 + 0.1 MHF + 0.1% H2O2 followed by 2.5 c.v. of 0.1 M HF. Adsorbedmatrix elements (e.g., Mg, Cr, Ni) were eluted in 6 M HCl forfuture analyses. A clean-up column with 1-mL AG1-X4, 200–400mesh was used to purify W and the samples were loaded in 1 MHF. Al was eluted with 5 c.v. of 1 M HF and Ti, Hf, and Zr with 5c.v. 2 M HCl + 0.1%H2O2. Residual Hf and Zr were eluted with 2c.v. of 6 M HCl + 0.01 M HF. Finally, W was collected with 4 c.v.of 6 M HCl + 1 M HF. In addition, Mo was unloaded in 1 M HCl.The second column step was repeated twice to obtain sufficientlypure W separates. After drying, the W cuts were oxidized six timesin 4:1 HNO3:H2O2 and dried at 100 °C to remove any remainingOs and to break down organic material from the resin that mayotherwise produce interferences onW isotope measurements. Thetotal W procedural blank of our method and its associated un-certainty were determined from a number of replicate blankanalyses during the course of this study. Based on these meas-urements, we estimate the total procedural blank to be 79 ± 40 pg.A blank correction was applied to all samples, which was negli-

Holst et al. www.pnas.org/cgi/content/short/1300383110 2 of 12

gible for the melilite and anorthite fractions (less than 1 ppm).The blank correction applied on the Al,Ti-diopside fraction cor-responds to !30 ppm and, although within the final uncertainty ofthe isotope measurement, is not negligible. Final uncertainties ofthe isotope measurements presented in Table 1 of the main paperinclude the uncertainty of the blank correction.Tungsten isotope data were acquired using the ThermoFisher

Neptune MC-ICPMS located at the Centre for Star and PlanetFormation, Natural History Museum of Denmark, University ofCopenhagen. Following tungsten purification, samples were con-verted to nitrate form, dissolved in a 2% HNO3 solution con-taining traces of HF, and introduced into the plasma source bymeans of an Aridus II desolvating nebulizer (dry plasma). Thetypical sample aspiration rate with this introduction system was!0.05 mL/min. Isotope data were acquired in static mode usingfive Faraday collectors set up as follows: 183W in the axial col-lector and 182W in the low-1 collector on the low mass side of theaxial Faraday, and 184W, 186W, and 188Os in the high-1, high-2,and high-3 collectors on the high mass side of the axial Faraday.The 188Os collector was connected to an amplifier with a 1012-Ωfeedback resistor, whereas the remainder of the collectors wasconnected to amplifiers with 1011-Ω feedback resistors. Theisobaric interferences from 184Os and 186Os on the W mass arraywere corrected by monitoring the 188Os signal intensity, but thiscorrection was in all cases less than 5 ppm and, thus, negligible.The sensitivity of the instrument in low-resolution mode was!1,200 V/ppm. Samples and standards were analyzed witha signal intensity of !300–500 mV on mass 183W and ensuringthat the signal intensity of the sample and standard werematched to within 2%. Each analysis comprised a total of 1,259 sof baseline measurements obtained on-peak (in the same 2%HNO3 solution containing traces of HF used to dissolve samplesand standards) and 839 s of data acquisition (100 scans in-tegrated over 8.39 s). Sample analyses were interspaced withanalyses of the National Institute of Standards and Technology(NIST) 3163 standard as follows: standard1, standard2, sample1,standard3, standard4, sample2... A wash time of 30 min wasapplied after each sample and standard analyses. The typicaltotal amount of tungsten consumed per analysis using this ap-proach was !10 ng. The total amount of W available for analysisfor the melilite, anorthite and Al,Ti-diospide fractions was 7, 6,and 1.5 ng, respectively. Therefore, the length of the data ac-quisition sequence was reduced to !500 s for the melilite andanorthite fractions and !300 s for the Al,Ti-diospide fraction,but not for the bracketing standards.All data reduction was conducted off-line using the freely

available Iolite data reduction software that runs within Igor Pro.The data reduction modules used for W isotope ratio calculationsare freely available and can be obtained from the authorson request. Background intensities were interpolated using asmoothed cubic spline, as were changes in mass bias with time.Iolite’s Smooth spline auto choice was used in all cases, whichdetermines a theoretically optimal degree of smoothing based onvariability in the reference standard throughout an analyticalsession. For each analysis, the mean and SE of the measuredratios were calculated using a 2 SD threshold to reject outliers.Although a 2 SD outlier rejection scheme is frowned upon bysome, we submit that this approach is fully justified in the currentstudy given the limited amounts of W available for study. Sam-ples were dissolved in small acid volumes (<400 μL) to maximizesignal intensities and, thus, were entirely consumed duringanalysis. This leads to increased uptake rates in the last portionof the analyses and, as such, increased signal intensities that maynot match that of the bracketing standard. The use of the 2 SDoutlier rejection scheme provides an effective and objectivemeans of filtering out potentially spurious data imparted by thismismatch. Mass fractionation was corrected using the exponen-tial law and two different approaches: (i) The 182W/183W and

184W/183W ratios were corrected for mass fractionation using the186W/183W = 1.98594 (9) and (ii) the 182W/184W and 183W/184Wratios were corrected for mass fractionation using the 186W/184W =0.92767 (9). Samples were analyzed only once, and the ratios arereported as relative deviations from the mass-bias-corrected NIST3163 tungsten reference material in the μ-notation (106 deviations).The accuracy and external reproducibility (2 SD) of the W–

isotope measurements acquired using our protocol was evalu-ated by repeated analysis of a number of distinct aliquots ofcolumn-processed BCR-2 rock standard and the Allende car-bonaceous chondrite using a quantity of W comparable to thatpresent in mineral fractions of the STP-1 FUN CAI. For thisexperiment, a single batch of the BCR-2 and Allende sampleswas digested according to our sample digestion procedure. Oncein solution, five aliquots estimated to contain 5 ng and five ali-quots estimated to contain 1.5 ng of W were extracted from eachof the BCR-2 and Allende digestions, processed individually forW purification and analyzed in the same fashion as the un-knowns. Results from this experiment are presented in Table S5.The W–isotope composition of the various aliquots of the BCR-2rock standard returned values that are identical to the compo-sition of the NIST 3163 terrestrial W standard within the un-certainties of the measurements. The W–isotope compositions ofthe various aliquots of the Allende carbonaceous chondrite areidentical to that reported by earlier studies (10–13). In bothcases, the 186W/183W corrected data returned a superior externalreproducibility compared with the 186W/184W corrected dataand, therefore, the 186W/183W normalization is our preferrednormalization scheme and was used to correct the STP-1 data.Based on these experiments, we infer that the external re-producibility of the 182W/183W ratio is 15 and 21 ppm whenanalyzing 5 and 1.5 ng of W, respectively.

Analytical Protocols for Hf/W Ratio Determination. The 180Hf/184Wratios were determined using the Thermo X-Series quadrupoleICPMS at the Centre for Star and Planet Formation in Co-penhagen. A 15% aliquot of each mineral separate was removedafter dissolution, spiked with a 180Hf-186W mixed-tracer solution,then dried down and converted to nitrate form. Each aliquot wasthen dissolved in 400-μL 0.5 MHNO3 + 0.02 MHF in preparationfor mass spectrometry. To deconvolve both instrumental mass biasand the spike-to-sample ratio of Hf, three isotopes were measured(178Hf, 179Hf, and 180Hf). The 178Hf/179Hf ratio was used to cor-rect for mass bias using a ratio of 2.00296, and the fraction of spikeon 180Hf was then determined using the equation

F!spike" =!178Hf

.180Hf!sample" " 178Hf

.180Hf!true"

".

!178Hf

.180Hf!spike" # 178Hf

.180Hf!true"

"

where 178Hf/180Hf[sample] is the measured ratio after correction forinstrumental mass bias, 178Hf/180Hf[true] is the natural ratio of0.77765, and 178Hf/180Hf[spike] is the spike ratio of 0.00706. Simi-larly, the fraction of spike on 186W was calculated using the183W/186W ratio after correction for mass bias using 184W/183W =2.14117, together with values of 183W/186W[true] = 0.50354 and183W/186W[spike] = 0.00010. The above results were then combinedto calculate the sample 180Hf/184W ratio using the equation

180Hf.

184W!sample" = 180Hf.

186W!spike"

.F!spike"

180Hf

!F!spike"186W! 186W

.184W!true"

where F[spike]180Hf and F[spike]

186W are the results of the abovecalculations, and 186W/184W[true] is the natural ratio of 0.927670.All calculations were performed on a timeslice-by-timeslice basis

Holst et al. www.pnas.org/cgi/content/short/1300383110 3 of 12

in Iolite (3), using a data reduction scheme that is freely avail-able from the authors on request. To assess the accuracy of theapproach, a synthetic Hf–W solution of known composition (Hf/W = 4.02) was measured after being doped with varying propor-tions of the major elements found in CAI material (Ca, Al, Mg,Fe, Ti, and Zr), up to the concentrations present in the mineralseparates. Based on these tests we estimate the external repro-ducibility of the sample Hf/W ratios to be ± 6% (2 SD).

We note that the concentration of W in bulk and mineralseparates of STP-1 is !150 ppb, i.e., depleted by a factor of !8relative to bulk canonical CAIs. In addition, Hf ranges from 140to 890 ppb, which is !4–8 times depleted compared with fassaiteand melilite-rich separates from canonical CAIs (14). Thus,hafnium and tungsten in STP-1 are roughly equally depletedrelative to canonical CAIs, resulting in comparable Hf/W ratiosof the two types of inclusions.

1. Bizzarro M, et al. (2011) High-precision Mg-isotope measurements of terrestrialand extraterrestrial material by HR-MC-ICPMS – implications for the relative andabsolute Mg-isotope composition of the bulk silicate earth. J Anal At Spectrom26:565–577.

2. Pouchou JL, Pichoir F (1984) Un nouveau modèle de calcul pour la microanalysequantitative par spectrométrie de rayons X - Partie I: Application à l’analysed’échantillons homogènes. La Recherche Aérospatiale 3:167–192.

3. Paton C, Hellstrom JC, Paul BT, Woodhead JD, Hergt JM (2011) Iolite: Freeware for thevisualisation and processing of mass spectrometric data. J Anal At Spectrom 26:2508–2518.

4. Palme H, Jones A (2003) Treatise on Geochemistry, ed Davis AM (Elsevier,Amsterdam), Vol 1, pp 41–61.

5. Tatsumoto M, Knight RJ, Allègre CJ (1973) Time differences in the formation ofmeteorites as determined from the ratio of lead-207 to lead-206. Science 180(4092):1279–1283.

6. Connelly JN, et al. (2012) The absolute chronology and thermal processing of solids inthe solar protoplanetary disk. Science 338(6107):651–655.

7. Fritz JS, Garralda BB, Karraker SK (1961) Cation exchange separation of metal ions byelution with hydrofluoric acid. Anal Chem 33:882–886.

8. Strelow FWE, Weinert CHSW, Eloff C (1972) Distribution coefficients and anionexchange behavior of elements in oxalic acid – hydrochloric acid mixtures. Anal Chem44:2352–2356.

9. Völkening J, Köppe M, Heumann KG (1991) Tungsten isotope ratio determinations bynegative thermal ionization mass spectrometry. Int J Mass Spectrom Ion Process107:361–368.

10. Kleine T, Münker C, Mezger K, Palme H (2002) Rapid accretion and early coreformation on asteroids and the terrestrial planets from Hf-W chronometry. Nature418(6901):952–955.

11. Kleine T, Mezger K, Münker C, Palme H, Bischoff A (2004) 182Hf-182W systematics ofchondrites, eucrites and martian meteorites: Chronology of core formation and earlymantle differentiation in Vesta and Mars. Geochim Cosmochim Acta 68:2935–2946.

12. Scherstén A, Elliott T, Hawkesworth C, Norman M (2004) Tungsten isotope evidence thatmantle plumes contain no contribution from the Earth’s core. Nature 427(6971):234–237.

13. Irisawa K, Yin Q-Z, Hirata T (2009) Discovery of non-radiogenic tungsten isotopicanomalies in the Allende CV3 chondrite. Geochem J 43:395–402.

14. Burkhardt C, et al. (2008) Hf-W mineral isochron for Ca,Al-rich inclusions: Age of thesolar system and the timing of core formation in planetesimals. Geochim CosmochimActa 72:6177–6197.

Cl: :Na Mg

sod

nph

Ti: :Ca Al

px

0.5 mm

Mg: :Ca Al

sp

hib

melpx

a b

c d

Fig. S1. Backscattered electron image (A) and combined X-ray elemental maps in (B) Mg (red), Ca (green), and Al (blue), (C) Ti (red), Ca (green), and Al (blue),and (D) Cl (red), Na (green), and Mg (blue) of the coarse-grained Type B2 FUN CAI STP-1 (section 1) from the CV carbonaceous chondrite Allende. Regionoutlined in B is shown in detail in Fig. S2. The CAI consists of gehlenitic melilite (mel) and igneously zoned Al,Ti-diopside (px) poikilitically enclosing euhedralspinel grains (sp). A bottom part of the CAI contains abundant hibonite (hib) grains (blue in B). Euhedral lath-shaped hibonite grains and spinel–hiboniteintergrowths occur also around the CAI periphery. Melilite grains are cross-cut by grossular–anorthite (grs-an) veins; in the CAI periphery, melilite is partiallyreplaced by nepheline (nph, green in D) and sodalite (sod, yellow in D). A part of the CAI (appearing as a black hole) was drilled out for bulk Al–Mg isotopemeasurement.

Holst et al. www.pnas.org/cgi/content/short/1300383110 4 of 12

sp

sp

sp

hib

hib

andr

and

sec an

an

mel

mel

nphgrs

grs

sod

sod

fa

sp

hib

hib

and

mel grs

sec an

a b

c d

e f

sec px

Fig. S2. Backscattered electron images of the Allende FUN CAI STP-1 section 1 (A–F). Regions outlined in A and E are shown in detail in B and F, respectively.The section mainly consists of gehlenitic melilite (mel) poikilitically enclosing euhedral spinel grains (sp). Lath-shaped hibonite (hib) grains and spinel–hiboniteintergrowths occur in the outermost portion of the inclusion; no Wark–Lovering rim layers are present outside the hibonite-rich zone. Melilite grains are cross-cut by grossular–anorthite veins and in the outermost portion of the CAI are partially replaced by sodalite (sod), secondary Fe-bearing Al-rich pyroxene (sec px),andradite (andr), ferroan olivine (fa), grossular (grs), and secondary Na-bearing plagioclase (sec an).

Holst et al. www.pnas.org/cgi/content/short/1300383110 5 of 12

1 mm

Ti: :Ca AlMg: :Ca Al

sp

an

mel

px

hib

a b

d

Na

c

Si

grs-an veins

Fig. S3. Combined X-ray elemental maps in (A) Mg (red), Ca (green), and Al (blue) and (B) Ti (red), Ca (green), and Al (blue), and elemental maps in Si (C) andNa Ka (D) of the coarse-grained Type B2 FUN CAI STP-1 (section 2) from the CV carbonaceous chondrite Allende. The CAI consists of gehlenitic melilite (mel),anorthite (an), and igneously zoned Al,Ti-diopside (px) all poikilitically enclosing euhedral spinel grains (sp). Euhedral lath-shaped hibonite (hib) grains andspinel–hibonite intergrowths are rare and occur in the outermost portion of the inclusion. Melilite grains are cross-cut by grossular–anorthite (grs-an) veins; inthe CAI periphery, melilite is partially replaced by secondary Na-rich minerals (nepheline and sodalite). Part of the central portion of the CAI was possibly lostduring cutting.

Holst et al. www.pnas.org/cgi/content/short/1300383110 6 of 12

Mg: :Ca Al

Cl: :Na Al

sp

px

hib

pxan

sod

nph

mel

grs-an veins

a b

c d

STP-1 Ti

Si1 mm

Fig. S4. Combined X-ray elemental maps (A and D) in (A) Mg (red), Ca (green), and Al (blue) and (D) Cl (red), Na (green), and Al (blue), and elemental maps inTi (B) and Si Ka (C) of the coarse-grained Type B2 FUN CAI STP-1 (section 3) from the CV carbonaceous chondrite Allende. Region outlined in A is shown in Fig.S5A. The CAI consists of gehlenitic melilite (mel), anorthite (an), and Al,Ti-diopside (px) all poikilitically enclosing euhedral spinel grains (sp). Euhedral lath-shaped hibonite (hib) grains and spinel-hibonite intergrowths occur in the outermost portion of the inclusion. Melilite grains are crosscut by grossular-anorthite (grs-an) veins and partially replaced by nepheline (nph), sodalite (sod), and secondary anorthite. Part of the central portion of the CAI was possiblylost during cutting.

Holst et al. www.pnas.org/cgi/content/short/1300383110 7 of 12

mel

mel

melsp

sp

px

grs

hib

sec px

sec an

sec an

sod

sod

andr

a b

c

mel

sp px

grs grsan

d

sphib

Fig. S5. Backscattered electron images of regions in section 3 (A"C) and 2 (D) of the Allende FUN CAI STP-1. Region outlined in A is shown in detail in B. TheCAI consists of gehlenitic melilite (mel), anorthite (an), and Al,Ti-diopside (px) all poikilitically enclosing euhedral spinel grains (sp). Euhedral lath-shapedhibonite (hib) grains and spinel-hibonite intergrowths are rare and occur in the outermost portion of the inclusion. The CAI lacks Wark-Lovering rim layerscommonly observed around non-FUN coarse-grained CAIs. Melilite grains are cross-cut by grossular-anorthite (grs-an) veins and in the outermost portion of theCAI are partially replaced by sodalite (sod), secondary Fe-bearing Al-rich pyroxene (sec px), andradite (andr), and secondary Na-bearing plagioclase (Table S1).Regions indicated by red and yellow lines in D correspond to spots sputtered during measurements of oxygen- and aluminum-magnesium isotope compo-sitions, respectively, by the UH Cameca ims-1280 ion microprobe.

mel

sp

hib

px

px

sec px

sec an

sec ol

sod

grs+anveins

Mg: :Ca Al BSE

Si Tia b

c d

px#4, O = 21.8‰17 –

mel#3, 17O = 9.9‰–mel#2, O = 17.3‰17 –

px#3, O = 22.5‰17 –

px#5, O = 23.8‰17 –

px#6, O = 23.1‰17 –

mel#4, 17O = 15.9‰–mel#5, 17O = 15.9‰–

mel#1, 17O = 10.9‰–

mel#7, 17O = 11.4‰–

mel#6, 17O = 4.1‰–

Fig. S6. (A) Combined X-ray elemental map in Mg (red), Ca (green), and Al (blue) and (B) backscattered electron image, and elemental maps in Si (C) and Ti Ka

(D) X-rays of a region in section 2. The locations of SIMS spots in melilite (mel) and Al,Ti-diopside (px) and their Δ17O values are indicated in B. Two spots in theouter portion of an Al,Ti-diopside grain (highlighted in yellow) are slightly depleted in 16O compared with two other analyses (highlighted in red) closer to theCAI core. Both 16O-depleted spots are in the Ti-depleted part of the pyroxene. Compositions of melilite are 16O-depleted to different degrees relative to thoseof Al,Ti-diopside. There is no obvious correlation between the degree of 16O depletion and spot location within melilite.

Holst et al. www.pnas.org/cgi/content/short/1300383110 8 of 12

px#10, O = 24.2‰17 –

px#11, O = 23.8‰17 –

px#9, O = 23.8‰17 –

px#13, O = 24.3‰17 –px#8, O = 21.1‰17 –px#12, O = 16.8‰17 – px#14, O = 20.7‰17 –

px#15, O = 21.9‰17 –

px#1, O = 24.0‰17 –

px#2, O = 24.0‰17 –

px#7, O = 24.4‰17 –

a

c

b

d

Fig. S7. Backscattered electron images of regions in section 2 (A–C) and section 1 (D) of STP-1 showing the locations of SIMS spots in Al,Ti-diopside (px) andtheir Δ17O values. Several spots in Al,Ti-diopside in the section 1 (highlighted in yellow in D) are slightly depleted in 16O (Δ17O range from "16.8 to "21.9‰)compared with the other analyses (Δ 17O < "23‰; highlighted in red) closer to the CAI core. The 16O-depleted spots appear to be closer to the peripheralportion of the CAI and grossular–anorthite veins than the 16O-rich spots.

an#6, 17O = 24.8‰–

an#1, 17O = 25.3‰–

an#2, 17O = 24.3‰–

mel#11, 17O = 5.0‰–

mel#9, 17O = 5.5‰–mel#10, 17O = 5.1‰–

an#3, 17O = 24.2‰–

an#4, 17O = 24.6‰–

an#5, 17O = 24.6‰–

mel#8, 17O = 7.0‰–

a

c

b

d

Fig. S8. Backscattered electron images of regions in section 2 (A–D) of STP-1 showing the locations of SIMS spots in melilite (mel) and anorthite (an) and their Δ 17Ovalues. All spots in anorthite have similar 16O-rich compositions (Δ 17O < "24‰), whereas all analyses of melilite are significantly 16O depleted (Δ 17O > "7‰).

Holst et al. www.pnas.org/cgi/content/short/1300383110 9 of 12

Table S1. Representative electron microprobe analyses of primary and secondary minerals in theAllende FUN CAI STP-1

Mineral SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O Total Ak, mol %

Primary mineralsAnorthite 42.6 n.d. 36.9 n.d. 0.03 n.d. 0.05 20.2 n.d. n.d. 99.8 ––

Al,Ti-diopside 41.0 0.03 24.5 0.17 0.07 n.d. 9.2 25.4 n.d. n.d. 100.4 ––

Al,Ti-diopside 37.7 8.7 22.0 0.10 0.03 n.d. 8.1 24.9 n.d. n.d. 101.6 ––

Hibonite 0.22 0.52 89.0 n.d. 0.37 n.d. 0.27 8.7 n.d. n.d. 99.0 ––

Hibonite 0.15 3.2 84.4 n.d. 0.31 n.d. 1.6 8.4 0.04 n.d. 98.0 ––

Melilite 23.2 0.04 35.3 n.d. 0.10 n.d. 1.1 40.8 n.d. n.d. 100.4 6.3Melilite 28.5 n.d. 27.2 n.d. n.d. n.d. 4.0 40.3 0.30 n.d. 100.2 28.1Spinel 0.29 0.43 70.7 0.36 0.16 n.d. 28.2 0.68 n.d. n.d. 100.9 ––

Spinel 0.30 0.29 65.4 0.12 19.5 0.05 14.8 0.29 0.04 n.d. 100.8 ––

Secondary mineralsAl-rich pyroxene 47.6 0.10 10.8 n.d. 5.1 n.d. 12.3 24.7 0.29 n.d. 100.7 ––

Andradite 36.1 n.d. 0.59 n.d. 26.9 0.06 0.31 32.7 n.d. n.d. 96.6 ––

Ferroan olivine 37.0 0.07 0.62 n.d. 26.0 0.12 34.9 0.46 0.04 n.d. 99.2 ––

Grossular 38.1 n.d. 23.6 n.d. 0.64 0.10 1.2 35.0 n.d. n.d. 98.6 ––

Nepheline 43.7 0.11 36.9 n.d. 1.1 n.d. 0.46 5.2 9.9 1.4 98.9 ––

Anorthite 43.5 n.d. 37.1 n.d. 0.29 0.03 0.29 18.3 0.89 n.d. 100.4 ––

n.d., not detected.

Table S2. REE concentrations of a bulk aliquot of the STP-1 FUNCAI

REE Concentration, ppm 2 SD CI normalized

La 1.795 0.413 7.326Ce 3.274 0.753 5.131Pr 0.641 0.147 6.648Nd 3.212 0.739 6.775Sm 0.903 0.208 5.861Eu 0.255 0.115 4.404Gd 0.604 0.272 2.959Tb 0.093 0.042 2.476Dy 0.457 0.206 1.798Ho 0.038 0.017 0.662Er 0.090 0.021 0.542Tm 0.121 0.028 4.716Yb 0.474 0.109 2.871Lu 0.017 0.004 0.679

CI normalized represents the absolute concentration normalized to the CIchondrite reference values reported by Palme and Jones (4). The quoteduncertainty reflects the accuracy of our measurements and is typically 23%(2 SD) for most REE apart from Eu, Gd, Tb, Dy, and Ho, which have an un-certainty of 45%.

Holst et al. www.pnas.org/cgi/content/short/1300383110 10 of 12

Table S3. Oxygen–isotope compositions of individual minerals inthe Allende STP-1 FUN CAI

Mineral δ17O 2σ δ18O 2σ Δ17O 2σ

Al,Ti-diopside "38.0 0.9 "32.5 0.9 "21.1 1.1Al,Ti-diopside "42.6 1.0 "36.2 1.0 "23.8 1.1Al,Ti-diopside "43.4 1.0 "36.9 1.0 "24.2 1.1Al,Ti-diopside "42.5 1.0 "36.0 1.0 "23.8 1.1Al,Ti-diopside "29.3 1.1 "24.1 1.0 "16.8 1.2Al,Ti-diopside "43.8 1.0 "37.5 1.0 "24.3 1.1Al,Ti-diopside "36.8 1.0 "30.9 1.0 "20.7 1.1Al,Ti-diopside "39.0 1.0 "32.9 1.0 "21.9 1.1Al,Ti-diopside "43.7 0.7 "37.9 0.8 "24.0 0.8Al,Ti-diopside "43.4 0.7 "37.5 0.7 "24.0 0.8Al,Ti-diopside "40.5 0.7 "34.6 0.8 "22.5 0.9Al,Ti-diopside "39.2 0.8 "33.5 0.8 "21.8 0.9Al,Ti-diopside "42.4 0.8 "35.7 0.8 "23.8 0.9Al,Ti-diopside "42.3 0.7 "36.9 0.7 "23.1 0.8Al,Ti-diopside "43.9 0.7 "37.5 0.8 "24.4 0.8Anorthite "43.5 0.8 "35.1 0.8 "25.3 0.9Anorthite "42.5 0.8 "35.0 0.8 "24.3 0.9Anorthite "43.2 0.8 "36.5 0.9 "24.2 0.9Anorthite "43.4 0.9 "36.0 0.9 "24.6 1.0Anorthite "43.0 0.8 "35.4 0.8 "24.6 0.9Anorthite "42.6 0.8 "34.2 0.9 "24.8 1.0Hibonite "42.8 1.0 "37.8 0.7 "23.1 1.0Hibonite "42.3 1.1 "34.7 0.7 "24.3 1.1Hibonite "42.9 1.0 "35.3 0.7 "24.6 1.1Hibonite "43.8 1.0 "36.4 0.8 "24.8 1.1Hibonite "40.6 1.0 "33.9 0.7 "23.0 1.1Hibonite "43.1 1.0 "35.8 0.7 "24.5 1.1Melilite "17.7 0.8 "13.2 0.6 "10.9 0.8Melilite "30.6 0.8 "25.6 0.7 "17.3 0.8Melilite "15.9 0.8 "11.6 0.7 "9.9 0.9Melilite "27.8 0.8 "22.9 0.6 "15.9 0.8Melilite "27.5 0.9 "22.3 0.6 "15.9 0.9Melilite "4.4 0.7 "0.6 0.8 "4.1 0.9Melilite "17.7 0.8 "12.2 0.6 "11.4 0.9Melilite "9.8 0.9 "5.5 0.7 "7.0 0.9Melilite "7.0 0.8 "3.0 0.7 "5.5 0.9Melilite "5.7 0.7 "1.2 0.7 "5.1 0.8Melilite "6.3 0.7 "2.6 0.7 "5.0 0.8Spinel "43.6 1.0 "38.0 0.7 "23.8 1.0Spinel "46.4 1.1 "42.9 0.7 "24.1 1.1Spinel "44.4 1.0 "38.2 0.7 "24.6 1.0Spinel "43.2 1.0 "37.5 0.7 "23.7 1.0Spinel "41.4 1.0 "33.2 0.7 "24.1 1.1Spinel "39.5 1.0 "29.0 0.7 "24.5 1.1Spinel "38.3 1.0 "27.9 0.7 "23.7 1.1Spinel "39.7 1.0 "30.2 0.7 "24.0 1.1Spinel "43.3 1.0 "35.7 0.7 "24.7 1.1

Holst et al. www.pnas.org/cgi/content/short/1300383110 11 of 12

Table S4. 27Al/24Mg ratios and Mg isotope compositions of indi-vidual minerals in the Allende STP-1 FUN CAI

Sample 27Al/24Mg 2σ δ26Mg* 2σ δ25Mg 2σ

Anorthite 744.5 16.0 14.99 2.41 8.48 1.18Anorthite 773.5 16.4 14.24 3.06 9.99 1.46Anorthite 247.3 5.3 5.46 2.36 8.44 1.16Anorthite 643.4 16.0 13.43 3.21 8.07 1.49Anorthite 829.0 17.5 16.85 3.22 10.38 1.52Anorthite 754.0 15.9 15.75 2.91 10.19 1.36Anorthite 768.7 16.3 16.91 3.12 8.75 1.47Hibonite 1147.9 23.7 27.46 8.28 27.87 4.11Hibonite 165.2 3.4 2.93 3.35 13.11 2.37Hibonite 474.9 10.0 10.60 4.55 23.25 2.79Hibonite 129.9 2.7 1.83 2.95 4.44 2.21Spinel 2.5 0.4 "0.31 0.09 10.88 0.15Spinel 2.5 0.4 "0.29 0.11 10.99 0.15Spinel 2.5 0.4 "0.33 0.10 10.95 0.15Spinel 2.5 0.4 "0.33 0.09 10.79 0.15Spinel 2.5 0.3 "0.34 0.08 10.84 0.15Melilite 17.1 0.8 0.026 0.564 12.09 0.27Melilite 12.8 0.7 0.017 0.530 12.24 0.26Melilite 10.6 0.6 "0.182 0.546 12.05 0.27Melilite 6.2 0.6 "0.107 0.507 13.20 0.25Melilite 14.7 1.5 "0.158 0.597 12.43 0.28Melilite 23.4 1.1 0.068 0.619 12.03 0.29Melilite 23.6 1.1 "0.145 0.563 12.25 0.27Fassaite 2.4 0.5 "0.416 0.335 11.14 0.17Fassaite 2.8 0.6 "0.257 0.300 10.76 0.16Fassaite 1.8 0.4 "0.415 0.295 10.85 0.15Fassaite 3.0 0.5 "0.508 0.379 11.23 0.18Fassaite 3.1 0.6 "0.380 0.326 10.81 0.16Bulk 3.180 0.06 "0.303 0.010 9.33 0.02

Table S5. W isotope data of multiple column processed aliquots of Allende and BCR-2 rockpowders

Sample μ182W (6/3) 2 SD μ184W (6/3) 2 SD μ182W (6/4) 2 SD μ183W (6/4) 2 SD N

Allende (5 ng) "227 15 "20 19 "176 19 30 29 5Allende (1.5 ng) "206 20 "13 18 "184 34 20 27 5BCR-2 (5 ng) "3.8 8 1.6 10 "3 19 "2 14 5BCR-2 (1.5 ng) "17 21 8 17 "22 23 "13 26 5

Uncertainties reflect the external reproducibility of the method. N, number of individual column processedaliquots; (6/3), internally normalized to 186W/183W; (6/4), internally normalized to 186W/184W.

Holst et al. www.pnas.org/cgi/content/short/1300383110 12 of 12