Widespread magma oceans on asteroidal bodies inthe early Solar SystemRichard C. Greenwood1, Ian A. Franchi1, Albert Jambon2 & Paul C. Buchanan3

Immediately following the formation of the Solar System, smallplanetary bodies accreted1, some of which melted to produceigneous rocks2,3. Over a longer timescale (15–33 Myr), the innerplanets grew by incorporation of these smaller objects4,5 throughcollisions. Processes operating on such asteroids strongly influ-enced the final composition of these planets4, including Earth5.Currently there is little agreement about the nature of asteroidaligneous activity: proposals range from small-scale melting, to neartotal fusion and the formation of deep magma oceans2. Here wereport a study of oxygen isotopes in two basaltic meteorite suites,the HEDs (howardites, eucrites and diogenites, which are thoughtto sample the asteroid 4 Vesta6) and the angrites (from anunidentified asteroidal source). Our results demonstrate thatthese meteorite suites formed during early, global-scale melting($50 per cent) events. We show that magma oceans were presenton all the differentiated Solar System bodies so far sampled.Magma oceans produced compositionally layered planetesimals;the modification of such bodies before incorporation into largerobjects can explain some anomalous planetary features, such asEarth’s high Mg/Si ratio.Oxygen isotope analyses were undertaken by laser-assisted

fluorination (Supplementary Methods). All major HED rock types(Fig. 1) and a representative suite of five angrite samples wereanalysed (full results are given in Supplementary Table 1). Fromthe results it is clear that the HEDs and angrites formed from twoisotopically distinct reservoirs (Fig. 1). The results do not support theD17O value of20.23 previously obtained for Angra dos Reis7, givingit an isotopic composition indistinguishable from the HEDs. (D17Ois defined in Supplementary Methods.) Using two distinct sub-samples, we found that Angra dos Reis has a D17O value of20.080 ^ 0.009 and plots on the angrite fractionation line (AFL)along with the other angrites studied.The d18O variation displayed by the HEDs shows strong miner-

alogical control (d18O is defined in Supplementary Methods), with18O-rich values for eucrites reflecting high plagioclase contents,whereas 18O-poor values for diogenites result from their orthopyr-oxene-rich mineralogy. Polymict breccias have major element com-positions intermediate between diogenites and eucrites, beingphysical mixtures of these two end-members2. This is also reflectedin their oxygen isotope variation, with most samples plottingbetween the diogenites and eucrites (Fig. 1). Two exceptions arePasamonte and Bholghatti, which plot above and below the eucritefractionation line (EFL), respectively (Fig. 1). The deviation ofBholghatti is consistent with the presence of 1.3% carbonaceouschondrite material, similar to the C2 clasts previously identified inthis howardite8. The polymict eucrite Pasamonte has a less negativeD17O value than the EFL, reflecting the presence of extraneousmaterial with a relatively high D17O value. Pasamonte is known tobe contaminated by ejecta material, which siderophile element data

suggest was either H-group or CI chondrite9. If this materialwas exclusively ordinary chondrite-like, the isotopic compositionof Pasamonte is consistent with a mixture of basaltic eucrite plus3–3.5% of this non-indigenous component.Basaltic eucrites have been subdivided into three compositional

groups: (1) main group eucrites, (2) Nuevo Laredo trend eucrites,and (3) Stannern trend eucrites2. We have analysed one sample fromboth the Nuevo Laredo (Lakangaon) and Stannern (Stannern)trends. These are isotopically indistinguishable from the othereucrites studied (Supplementary Table 1), and indicate that theequilibration of oxygen isotopes pre-dates the evolution of specificHED lithologies.Various models have been proposed to account for the genesis of

the HEDs. Early schemes invoked fractional crystallization, with thediogenites representing cumulates and the eucrites forming from theresidual liquid10. Problems in reproducing the bulk composition ofeucrites by this process led to the suggestion that they formed by lowpressure partial melting (15–20%)11. Total melting of the parent bodyhas also been proposed, with subsequent core segregation andcrystallization of the molten mantle to produce the diogenite-eucritesequence12. The merits of these models can be assessed by examiningthe conditions that led to oxygen isotopic homogenization in theHED parent body.No group of chondritic meteorites has an oxygen isotope compo-

sition matching that of the HEDs. Consequently, proposed modelsfor the bulk composition of the HED parent body generally invokemixtures of more primitive components, such as 70% H chondrite:30% CM chondrite13, or 70% L chondrite: 30% CV chondrite12.Formation from pre-existing bodies is consistent with modelsindicating that early-formed planetesimals experienced significantcollisional evolution14. The structure and grain size of the HEDparent body immediately after formation is unknown, but may haveresembled a ‘rubble pile’15 comprising kilometre-sized blocks mixedwith finer material. Formation from a mix of distinct primitivematerials, such as, carbonaceous and ordinary chondrites, meansthat it must originally have been heterogeneous with respect tooxygen isotopes. However, the data presented here shows that thesource of the HEDs had a uniform D17O value of 20.239 ^ 0.007and hence must have undergone a homogenization event after initialaccretion.It is well established that the HED suite is extremely old, being not

much younger than the age of the Solar System2. Thus, the eucriteAsuka 881394 yields an age of 4,563.2 ^ 0.6Myr (ref. 3), only 4Myryounger than that obtained from calcium and aluminium-richinclusions (CAIs) in the meteorite Efremovka16 (CAI ages are widelyaccepted as dating Solar System formation). Mn-Cr dating indicatesthat the HED parent body formed and differentiated only2.4 ^ 0.9Myr after CAI formation, thus coinciding with the finalphase of chondrule formation1. Hf-W dating also suggests that the

LETTERS

1PSSRI, Open University, Walton Hall, Milton Keynes MK7 6AA, UK. 2Laboratoire MAGIE, Universite Pierre et Marie Curie, CNRS UMR 7047 case 110, 4 place Jussieu,75252 Paris Cedex 05, France. 3Department of Geology, Rhodes University, PO Box 94, Grahamstown 6140, South Africa.

Vol 435|16 June 2005|doi:10.1038/nature03612

916© 2005 Nature Publishing Group

HED parent body formed early, with core separation at 3 ^ 1.4Myrafter CAI formation, followed by mantle differentiation0.9 ^ 0.3Myr later4. These relationships suggest that parent bodyformation and differentiation occurred on a very short timescale, andconstrain any model for the oxygen isotope evolution of the HEDsuite.These time constraints exclude the possibility that oxygen isotopic

equilibration could have occurred at subsolidus temperatures. Even ifthe parent body had formed at the same time as CAIs, then beenheated instantaneously to 1,100 8C and remained at that temperaturefor 2.4Myr, the diffusion distance (x) in olivine17,18 would only havebeen of the order of 0.4mm. (Here x ¼

pDt, whereD is the solid state

diffusion coefficient and t is time.) This indicates that at the onset ofpartial melting (1,170 8C), the source region would have beenheterogeneous with respect to oxygen isotopes. Partial melting ofan unequilibrated source would have led to the formation of eucritesand diogenites with distinct isotopic compositions. Modellingsuggests that differences in D17O of over 0.2‰ between eucritesand diogenites would be produced by such a process (SupplementaryInformation). The fact that diogenites and eucrites both fall on thesame mass fractionation line poses a significant difficulty for models

invoking low degrees of partial melting to explain the genesis of theHED suite11.The geochemical evidence that the HED suite evolved after core

formation12 provides an important constraint when attemptingto define the conditions under which the source region becomeisotopically homogeneous. It has been demonstrated that coreseparation requires a minimum of 50% melting of the silicatefraction19. At this level of partial melting, the mantle of the HEDparent body would have consisted of convecting melt with entrainedolivines and minor pyroxene. Although the melt will be well-mixedowing to convection, the olivine crystals may still be unequilibrated.Complete homogenization of the source would involve diffusiveexchange between the olivines and enclosing melt. The scale ofhomogenization during 50% partial melting (1,450 8C) of a potentialHED source composition12 can be estimated from the diffusion time(t ¼ x2/D) for a 1-cm-diameter olivine crystal (x being the radiusand D the diffusion coefficient17). This olivine size was chosen tomatch that found in pallasite meteorites2. At 1,450 8C, the timescalesinvolved are of the order of 400,000 yr. In comparison, if melting tookplace at 1,350 8C (35% melting), oxygen isotope homogenizationwould require at least 2.6Myr, which is too long on the basis ofevidence fromHf-Wisotopes4. The oxygen isotope data for theHEDsare therefore consistent with a model involving a minimum of50% silicate partial melting, as required for effective separation ofa Fe-Ni-S liquid to form the core. This scale of meltingwould correspond to a crystal-free melt layer of approximately50 km depth.Angrite samples plot on a well-defined mass fractionation line

(Fig. 1), indicating that their parent body also underwent an early,global-scale oxygen isotope homogenization event. The angrites areequally as old as the HEDS, with D’Orbigny giving an age of4,562.9 ^ 0.6Myr (ref. 20), that is, 4.3Myr after CAI formation.Angrites are compositionally distinct from the HEDs, being criticallysilica undersaturated, whereas HEDs are hypersthene normative2.Experimental work indicates that both HED-like and angrite-likemelts can be produced from similar chondritic precursors by varyingthe oxygen fugacity (f O2

) during melting21. Angrite-like liquids formunder oxidizing conditions (with f O2

ranging from IW þ 1 toIW þ 2, where IW indicates the iron–wustite buffer) and eucriticliquids form under more reducing conditions (f O2

¼ IW–1). Intheory, both types of melts could have formed on the same asteroidif source conditions were locally variable. However, the oxygenisotope evidence presented here demonstrates that angrites andHEDs are the products of melting on two distinct parent bodies. Inaddition, both parent bodies had distinct post-formational histories,

Figure 1 | Oxygen isotope variation diagram for HEDs and angrites.Basaltic eucrites are fine- to medium-grained, commonly brecciated rockscomposed predominantly of pigeonite and plagioclase (,70–92% anorthite;,An70–92)

2. Cumulate eucrites are coarse-grained gabbros that aremineralogically similar to basaltic eucrites, but with more calcic plagioclase(An91–95). Diogenites are coarse-grained cumulates typically comprising84–100% orthopyroxene. Polymict breccias (including howardites) consistof lithic and mineral clasts predominantly derived from various types ofeucrites and diogenites2. Angrites are unique among Solar System basalts inbeing the most alkali-depleted and silica-undersaturated2. Angrites havevery distinct mineralogies, comprising varying amounts of Al-Ti-richdiopside, anorthite, Ca-rich olivine, kirschsteinite and a variety of minorphases. In contrast to HEDs, angrites show little evidence of brecciation andoften contain vesicles. On aD17O versus d18O diagram, samples formed froma homogeneous reservoir that subsequently fractionated by mass-dependent processes plot along horizontal lines. Silicate minerals on Earthhave isotopic compositions consistent with mass-dependent fractionationfrom a single reservoir28, and define the terrestrial fractionation line (TFL).All five analysed angrites define a second horizontal line, the angritefractionation line (AFL), with a mean D17O value of 20.072 ^ 0.007 (1j).HED samples display greater D17O variation than either the angrites orterrestrial silicates (Supplementary Table 1). However, HED polymictbreccias are known to contain extraneous material, such as carbonaceouschondrite clasts8. If polymict breccias are excluded, the remaining HEDsamples (n ¼ 16) show limited D17O variation and define a single eucritefractionation line (EFL) with a mean D17O value of 20.239 ^ 0.007. OurD17O value for theHED suite differs slightly from the previously determinedvalue of 20.219 (ref. 29). This small discrepancy may reflect calibrationdifferences and the fact that the previous study included meteorite finds,which have the potential to be terrestrially contaminated. The value for theEFL determined in the present study is based only on meteorite falls. P,Pasamonte; B, Bholghatti matrix sample.

Figure 2 | Mass fractionation lines for Mars, Earth, Moon, Vesta and theangrite parent body. MFL, Mars fractionation line30; TFL, terrestrialfractionation line; AFL, angrite fractionation line; EFL, eucrite fractionationline (Vesta). Data for the MFL from ref. 30. M, average mantle compositionof Mars30; E, average terrestrial upper-mantle composition24.

NATURE|Vol 435|16 June 2005 LETTERS

917© 2005 Nature Publishing Group

with the HEDs being extensively modified by later impacts, whereasangrites show little evidence of brecciation.We have suggested above that homogenization of oxygen isotopes

was a consequence of high degrees of partial melting. One means ofassessing this proposal is to look at the oxygen isotope variation inmeteorite groups that experienced lower degrees of partial melting.Acapulcoites and lodranites formed by,1% to slightly greater than20% partial melting of a single parent body22. Their D17O values(20.99‰ to21.49‰)7 show a significantly greater range than seenamong the HEDs or angrites, indicating that primary heterogeneitieswere preserved at low degrees of partial melting. Ureilites representthe residue formed when a 20–30%melt fractionwas removed from aprimitive chondritic precursor23. Ureilites display major oxygenisotope variation, scattering about the slope-1 line defined byprimitive carbonaceous chondrites7, again demonstrating that rela-tively low degrees of melting are insufficient to cause oxygen isotopehomogenization.The terrestrial planets and differentiated asteroids from which we

have samples plot as separate, well-defined mass fractionation lineson a D17O versus d18O diagram (Fig. 2). This demonstrates that eachformed from a unique mix of precursor materials24 and, followingaccretion, experienced at least one major phase of oxygen isotopehomogenization. Global-scale homogenization of oxygen isotopesmust have been a consequence of the same event that segregated thedifferentiated planets and asteroids into an Fe-Ni core25 and silicate-rich mantle and crust. For the larger bodies (Earth, Moon andMars),there is general agreement that core–mantle segregation occurredwhen the planet underwent almost total melting, leading to theformation of a magma ocean25. Although similar models have beenadvanced for asteroids12, there has been far less consensus about theimportance of such magma oceans in the development of thesesmaller, differentiated bodies26. The results of the present studyclearly demonstrate that high levels of partial melting were attainedon asteroid-sized bodies in the early Solar System.Collisions between bodies in the early Solar System did not always

result in accretion—significant erosion of their outer silicate portionsalso occurred14. Such reworking of the outer layers of differentiatedasteroids may have important implications for the bulk compositionof the planets formed by their accretion. Compositional layering is aninherent feature of asteroids that underwent a magma ocean stage, sothat loss of their outermost Fe- and Si-rich crustal layers wouldincrease the Mg/Si ratio of the residual remnant. The anomalouslyhigh Mg/Si ratio of Earth compared to chondrites may thereforeresult from collisional modification of the differentiated planetesi-mals from which it formed, rather than requiring sequestration of Siin the core, or the need to invoke an as-yet unsampled precursormaterial24. The evidence presented here suggests that global-scalemelting of differentiated asteroids was a major process in the earlySolar System. Later modification of these bodies during the planet-building phase may have been a significantly more important processthan has yet been considered. Imaging of the b Pictoris systemsuggests that such collisional reprocessing may be a widespreadfeature of planetary systems27.

Received 3 December 2004; accepted 4 April 2005.

1. Lugmair, G. W. & Shukloyukov, A. Early solar system events and timescales.Meteorit. Planet. Sci. 36, 1017–-1026 (2001).

2. Mittlefehldt, D. W., McCoy, T. J., Goodrich, C. A. & Kracher, A. in PlanetaryMaterials (ed. Papike, J. J.) Ch. 4 (Mineralogical Society of America,Washington DC, 1998).

3. Nyquist, L. E., Reese, Y., Wiesmann, H., Shih, C.-Y. & Takeda, H. Fossil 26Al and

53Mn in the Asuka 881394 eucrite: evidence of the earliest crust on asteroid 4Vesta. Earth Planet. Sci. Lett. 214, 11–-25 (2003).

4. Kleine, T., Mezger, K., Mumker, C., Palme, H. & Bischoff, A. 182Hf-182W isotopesystematics of chondrites, eucrites, and martian meteorites: Chronology ofcore formation and early mantle differentiation in Vesta and Mars. Geochim.Cosmochim. Acta 68, 2935–-2946 (2004).

5. Halliday, A. N. Mixing, volatile loss and compositional change during impact-driven accretion of the Earth. Nature 427, 505–-509 (2004).

6. Drake, M. J. The eucrite/Vesta story. Meteorit. Planet. Sci. 36, 501–-513 (2001).7. Clayton, R. N. & Mayeda, T. K. Oxygen isotope studies of achondrites. Geochim.

Cosmochim. Acta 60, 1999–-2017 (1996).8. Buchanan, P. C., Zolensky, M. E. & Reid, A. M. Carbonaceous chondrite clasts

in the howardites Bholghati and EET87513. Meteoritics 28, 659–-669 (1993).9. Metzler, K., Bobe, K. D., Palme, H., Spettel, B. & Stoffler, D. Thermal and impact

metamorphism on the HED parent asteroid. Planet. Space Sci. 43, 499–-525(1995).

10. Mason, B. Meteorites (Wiley, New York, 1962).11. Stolper, E. Experimental petrology of eucrite meteorites. Geochim. Cosmochim.

Acta 41, 587–-611 (1977).12. Righter, K. & Drake, M. J. A magma ocean on Vesta: Core formation and

petrogenesis of eucrites and diogenites. Meteorit. Planet. Sci. 32, 929–-944(1997).

13. Boesenberg, J. S. & Delaney, J. S. A model composition of the basalticachondrite planetoid. Geochim. Cosmochim. Acta 61, 3205–-3225 (1997).

14. Agnor, C. & Asphaug, E. Accretion efficiency during planetary collisions.Astrophys. J. 613, L157–-L160 (2004).

15. Nolan, M. C., Asphaug, E., Greenberg, R. & Melosh, H. J. Impacts on asteroids:Fragmentation, regolith transport, and disruption. Icarus 153, 1–-15 (2001).

16. Amelin, Y., Krot, A. N., Hutcheon, I. D. & Ulyanov, A. A. Lead isotopic ages ofchondrules and calcium-aluminum-rich inclusions. Science 297, 1678–-1683(2002).

17. Reddy, K. P. R., Oh, S. M., Major, L. D. Jr & Cooper, A. R. Oxygen diffusion infosterite. J. Geophys. Res. 85, 322–-326 (1980).

18. Valley, J. W. Stable isotope thermometry at high temperatures. Rev. Mineral.Geochem. 43, 365–-402 (2001).

19. Taylor, G. J. Core formation in asteroids. J. Geophys. Res. 97, 717–-726 (1992).20. Glavin, D. P., Kubny, A., Jagoutz, E. & Lugmair, G. W. Mn-Cr isotope

systematics of the D’Orbigny angrite. Meteorit. Planet. Sci. 39, 693–-700(2004).

21. Jurewicz, A. J. G., Mittlefehldt, D. W. & Jones, J. H. Experimental partialmelting of the Allende (CV) and Murchison (CM) chondrites and the origin ofasteroidal basalt. Geochim. Cosmochim. Acta 57, 2123–-2139 (1993).

22. McCoy, T. J., Keil, K., Muenow, D. W. & Wilson, L. Partial melting and meltmigration in the acapulcoite-lodranite parent body. Geochim. Cosmochim. Acta61, 639–-650 (1997).

23. Guan, Y. & Crozaz, G. Microdistributions and petrogenetic implications of rareearth elements in polymict ureilites. Meteorit. Planet. Sci. 36, 1039–-1056 (2001).

24. Drake, M. J. & Righter, K. Determining the composition of the Earth. Nature416, 39–-61 (2002).

25. Rubie, D. C., Gessman, C. K. & Frost, D. J. Partitioning of oxygen during coreformation on Earth and Mars. Nature 429, 58–-61 (2004).

26. Mittlefehldt, D. W. & Lindstrom, M. M. Geochemistry of eucrites: Genesis ofbasaltic eucrites and Hf and Ta as petrogenetic indicators for altered Antarcticeucrites. Geochim. Cosmochim. Acta 67, 1911–-1935 (2003).

27. Telesco, C. M. et al. Mid-infrared images of b Pictoris and the possible role ofplanetesimal collisions in the central disk. Nature 433, 133–-136 (2005).

28. Miller, M. F., Franchi, I. A., Sexton, A. S. & Pillinger, C. T. High precision D17Omeasurements of oxygen from silicates and other oxides: method andapplications. Rapid Commun. Mass Spectrom. 13, 1211–-1217 (1999).

29. Wiechert, U. H., Halliday, A. N., Palme, H. & Rumble, D. Oxygen isotopeevidence for rapid mixing of the HED meteorite parent body. Earth Planet. Sci.Lett. 221, 373–-382 (2004).

30. Franchi, I. A., Wright, I. P., Sexton, A. S. & Pillinger, C. T. The oxygen-isotopiccomposition of Earth and Mars. Meteorit. Planet. Sci. 34, 657–-661 (1999).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank M. Drake for comments on the manuscript.

Author Information Reprints and permissions information is available atnpg.nature.com/reprintsandpermissions. The authors declare no competingfinancial interests. Correspondence and requests for materials should beaddressed to R.C.G. (r.c.greenwood@open.ac.uk).

LETTERS NATURE|Vol 435|16 June 2005

918© 2005 Nature Publishing Group