Chronometry of Meteorites and the Formation of

the Earth and Moon Artist’s impression of a planetary collision. IMAGE COURTESY OF NASA Thorsten Kleine1 and John F. Rudge2

1811-5209/11/0007-0041$2.50 DOI: 10.2113/gselements.7.1.41

he planets of the Solar System grew by collisions, starting with the parameters that controlled plane- aggregation of tiny dust particles within the solar nebula and culmi- tary evolution, and on the conse- quences of collisional growth on Tnating in giant collisions between large planetary bodies. These giant the composition and differentia- impacts occasionally caused the formation of satellites such as the Earth’s tion of the planets. Moon. Our understanding of planet formation is based on information from DATING THE EARLY various sources, including meteorites – leftovers from the earliest stages of SOLAR SYSTEM planet formation – and samples from the Earth and Moon. By combining Two Kinds of Chronometers results from isotopic dating of these materials with dynamic modelling of Both long-lived and short-lived the solar nebula and planet formation, researchers can reconstruct the accre- radioisotopes are used for dating tion and early evolution of planetary bodies during the fi rst ~100 million the processes that occurred in the early Solar System (TABLE 1). Long- years of Solar System history. lived radioisotopes (e.g. 238,235 206,207 147 143 KEYWORDS: meteorites, early Earth, Moon, core formation, radioisotopes, chronology U– Pb, Sm– Nd) provide absolute ages, but for processes that occurred ~4.5 INTRODUCTION billion years ago only U–Pb ages are suffi ciently precise. The Solar System formed ~4.6 billion years ago by gravita- Short-lived radioisotopes (e.g. 26Al–26Mg, 182Hf–182W) are tional collapse of a localized, dense region of a large inter- unstable nuclides that existed at the beginning of the Solar stellar molecular cloud. Contraction of this cloud led to System but have since decayed away. Their former presence the accretion of a central star surrounded by a rotating can thus only be detected via the isotopic composition of disk of gas and dust, termed the solar nebula. The planets their daughter products. Short-lived nuclides are powerful accreted from this nebula by the aggregation of dust into dating tools (TABLE 1) because the abundance of their larger and larger bodies, a process that occurred in several daughter isotopes changed rapidly over short time inter- distinct stages. Settling of dust to the disk’s midplane was vals, such that the precise measurement of the daughter’s followed by growth into planetesimals ~1 km in size. isotopic composition can provide relative ages with high Collisions among these planetesimals resulted in the rapid temporal resolution. The key assumption is that the short- formation, in less than 1 million years (My), of about a lived radionuclides were homogeneously distributed hundred Moon- to Mars-sized embryos at 1 astronomical throughout the solar nebula. Whether or not this was the unit (AU). The fi nal stage of planet formation occurred on case has been a matter of much debate. The broad consis- a more protracted timescale of several tens of millions of tency of the chronology obtained from different short-lived years and was characterized by large collisions among the nuclides (Nyquist et al. 2009) and the observation that the embryos, one of which is thought to be responsible for the initial Mg isotope composition of chondrules is consistent formation of the Moon (e.g. Chambers 2004). with the 26Al–26Mg evolution of the solar nebula (Villeneuve et al. 2009) provide evidence for a homogeneous distribu- Collisions during accretion and the decay of short-lived tion of extinct nuclides. However, there is some isotopic radionuclides caused melting of planetary interiors, leading variability among inner Solar System planets, refl ecting to a separation of liquid-metal blobs that sank to the centre variable proportions of various nucleosynthetic compo- of the planet, forming a core (e.g. Stevenson 2008). As a nents (e.g. Trinquier et al. 2009). How this affects the short- consequence, all major bodies of the Solar System and lived nuclides is currently unclear. many smaller objects, such as the parent bodies of iron meteorites, are differentiated. The chondritic meteorites were derived from bodies that never melted, however, and IMPORTANT CHRONOMETERS as such provide invaluable hands-on samples for investi- TABLE 1 gating the evolution of the solar nebula prior to the onset OF THE EARLY SOLAR SYSTEM of planet formation. Parent Daughter Half-life (My) Application Current thinking on the accretion and early evolution of 26Al 26Mg 0.73 Relative ages of CAIs and chondrules inner Solar System planets is largely based on isotopic Relative ages of meteorites and their 53Mn 53Cr 3.7 analyses of extraterrestrial matter. These provide key infor- components mation on the timescale of planetary accretion, on the Core formation in planetesimals and 182Hf 182W 8.9 terrestrial planets 1 Institut für Planetologie, Westfälische Wilhelms-Universität Münster 146Sm 142Nd 103 Mantle differentiation in terrestrial planets Wilhelm-Klemm-Strasse 10, 48149 Münster, Germany E-mail: [email protected] Absolute age of CAIs; core formation 235U 207Pb 703 in the Earth 2 Institute of Theoretical Geophysics, Bullard Laboratories University of Cambridge, Madingley Road, Cambridge CB3 0EZ, UK 238U 206Pb 4468 E-mail: [email protected]

ELEMENTS, VOL. 7, PP. 41–46 41 FEBRUARY 2011 The Hf–W System 26Al–26Mg chronometry indicates that most chondrules in ordinary chondrites formed ~2 million years (My) after The decay of 182Hf to 182W is particularly useful for studying CAIs. Chondrules in carbonaceous chondrites appear to planetary accretion and core formation (e.g. Kleine et al. have formed over a longer time span, some as late as 4–5 2009). During core formation, tungsten (W) preferentially My after CAIs (e.g. Krot et al. 2009). Further, there appears partitions into the metal while hafnium (Hf) remains in to be peaks in the distribution of Al–Mg ages for individual the silicate mantle. Over time the mantle accumulates chondrules from a given chondrite group, suggesting that radiogenic 182W through the decay of 182Hf, while the W chondrules from a single parent body formed in distinct isotope composition of the core remains constant. Provided events over a period of several million years (Villeneuve that core formation took place while 182Hf was extant, the et al. 2009). However, chondrules should have been rapidly mantle will evolve a 182W excess compared to the undif- remixed in a turbulent solar nebula, such that the distinct ferentiated planet, whereas the core maintains the W physical and chemical properties of a given chondrule isotope composition acquired at the time of core formation population could only be preserved by rapid accretion and thus develops a 182W defi cit. When calculating core following their formation (Alexander 2005). Reconciling formation ages, it is assumed that the Hf–W isotopic evolu- dynamic models of the solar nebula with the chronology tion of the bulk, undifferentiated planet was similar to that of chondrule formation will be an important step towards of undifferentiated chondritic meteorites. The basis for this fully understanding the accretion of chondrite parent bodies. assumption is that Hf and W are refractory elements that were not strongly fractionated by processes within the solar Remnants of the First Protoplanets nebula. As such, they should occur in constant relative abundances throughout the Solar System. We will see later, Although chondrites contain some of the oldest material however, that for larger bodies such as the Earth this in the Solar System, they are not derived from the fi rst assumption may not be strictly valid. planetary objects. The chronology of chondrules suggests that most chondrite parent bodies accreted more than ~2 My after CAI formation (Krot et al. 2009). In contrast, FROM DUST TO PLANETESIMALS Hf–W chronology of magmatic iron meteorites – samples Making Chondrites from Nebular Gas and Dust from the metallic cores of protoplanets – indicates that accretion and differentiation of their parent bodies must Chondrites contain Ca–Al-rich inclusions (CAIs), whose have occurred within ~1 My of CAI formation (FIG. 1). Iron composition is similar to that expected for the fi rst refrac- meteorites were thus derived from protoplanets that were tory nebular condensates and whose age is the oldest already differentiated before chondrules formed and chon- known among materials formed within the Solar System. drite parent bodies accreted (Kleine et al. 2009). Consequently, CAIs are commonly used to defi ne the age of the Solar System, and the timing of all other events is While iron meteorite parent bodies were traditionally given in reference to the time of CAI formation (FIG. 1). viewed as samples from the metal cores of small bodies While CAIs have been precisely dated using the U–Pb (diameter ~20–200 km), it is now thought that they were system (Bouvier and Wadhwa 2010), their exact age derived from much larger bodies that were disrupted by remains somewhat uncertain due to U isotope variations hit-and-run collisions (Yang et al. 2007). Iron meteorites among CAIs (Brennecka et al. 2010). Nevertheless, a may thus be remnants of some of the earliest planetary combined high-precision U and Pb isotope study of an embryos, which perhaps had formed within the terrestrial Allende CAI provides an absolute age of 4567.2 ± 0.5 planet region and were later scattered in the asteroid belt million years before present (Ma) (Amelin et al. 2010), an (Bottke et al. 2006). age that is consistent with those obtained from short-lived 26 chronometers (Kleine et al. 2009; Krot et al. 2009; Nyquist Al and the Evolution of Planetesimals et al. 2009) and provides the current best estimate for the The meteorite chronometry summarized above indicates age of the Solar System. that planetesimal accretion started at about the same time as CAIs formed and continued over at least 4–5 My. The CAIs probably formed close to the young Sun and were following few million years were dominated by a variety subsequently transported outwards to several astronomical of parent-body processes, including the extrusion of lava units (AU) where the chondrites accreted (Ciesla 2010). fl ows on some asteroidal bodies, peak metamorphism in This outward transport was most effective for those CAIs the ordinary chondrites, and aqueous alteration on some that formed within the fi rst ~105 years, such that these carbonaceous chondrite parent bodies. Although these CAIs were preferentially preserved. This is consistent with processes took place in very different thermal regimes, they 26Al–26Mg chronometry evidence that CAIs in chondritic all occurred ~4–10 My after CAI formation (FIG. 1). meteorites formed over a period of only 20,000 years (Jacobsen et al. 2008). These coeval but different thermal histories of meteorite parent bodies probably refl ect their different initial 26Al Chondrules are the major components of most chondrites abundances. There is an inverse correlation between peak and are thought to have formed by rapid heating, melting temperatures reached inside planetesimals and their accre- and cooling of dust aggregates within the solar nebula. tion age (FIG. 2). Since the variation in accretion times is

2OGHVWOXQDUURFNV $JHRI(DUWK 0RRQ EHVWHVWLPDWH (DUOLHVWSRVVLEOHDJHRI0RRQ FIGURE 1 (DUOLHVWSRVVLEOHDJHRI(DUWK Timescales (loga- rithmic) for the forma- 3URWR(DUWKVXEVWDQWLDOO\IRUPHG a  tion and evolution of inner Solar System planets as determined $TXHRXVDOWHUDWLRQ &,FKRQGULWHV from isotopic dating. 3HDNPHWDPRUSKLVP RUGLQDU\FKRQGULWHV Dashed lines indicate uncertain timescales; %DVDOWLFPDJPDWLVP &UXVWDOPHWDPRUSKLVP HXFULWHV arrows indicate earliest &KRQGUXOHIRUPDWLRQ possible time of formation. ,URQPHWHRULWHSDUHQWERGLHVDFFUHWHGDQGGLIIHUHQWLDWHG

    7LPH 0\DIWHU6RODU6\VWHPIRUPDWLRQ

ELEMENTS 42 FEBRUARY 2011 26 of the order of several Al half-lives, the meteorite parent 7HPSHUDWXUHFXUYHVIRU 26 DOXPLQLXPKHDWLQJ bodies contained different initial Al abundances. In the  iron meteorite parent bodies, abundant 26Al facilitated ,URQV ,URQPHWHRULWHV NP JOREDOPHOWLQJDQGGLIIHUHQWLDWLRQ melting and core formation, while in the chondrite parent NP bodies too little 26Al remained to cause melting. The  different thermal histories of the ordinary and carbona-  $FDSXOFRLWHVORGUDQLWHV $/   ceous chondrites may also refl ect variable initial 26Al, but the $/ SDUWLDOPHOWLQJ presence of water in the latter may also have played a role.  2UGLQDU\FKRQGULWHV 2&  2& WKHUPDOPHWDPRUSKLVP LQFHQWUH ƒ& &DUERQDFHRXVFKRQGULWHV ACCRETION OF THE EARTH 3HDNWHPSHUDWXUH &2 &2 &5   PLOGWKHUPDOPHWDPRUSKLVP Link between Accretion and Core Formation DTXHRXVDOWHUDWLRQ In the current view of planet formation, Earth accreted &5 most of its mass by collisions with Moon- to Mars-sized  embryos. These collisions were highly energetic and raised  the temperature of Earth’s interior by thousands of degrees, $FFUHWLRQWLPH 0\DIWHU&$,IRUPDWLRQ causing widespread melting and the formation of magma oceans. In these oceans, metal easily separated from molten Accretion age versus peak temperatures reached inside FIGURE 2 silicate and segregated towards the centre. Thus, accretion meteorite parent bodies. Solid lines show tempera- and core formation were intimately linked and, as metal tures reached in the centre of planetesimals 10 and 100 km in segregation is thought to happen much faster than accre- radius due to heating by 26Al decay, assuming a homogeneous 26 tion, the rate of Earth’s accretion can be determined by distribution of Al throughout the Solar System, with an initial dating the formation of Earth’s core. Such age constraints abundance equal to that in CAIs. can be obtained from the Hf–W and U–Pb systems. give a mean-life of accretion of ~11 My, while the end of Dating the Earth’s Core accretion as given by the two-stage model is ~30 My. Using Samples from Earth’s crust and mantle have uniform W the same model, the U–Pb data give timescales that are a 182 isotope compositions and exhibit small excesses in W factor of 2 to 5 times longer. This disparity in calculated compared to undifferentiated meteorites. The most accretion timescales has been a matter of much debate, 182 straightforward interpretation of this W excess is that a and FIGURE 3 illustrates two ways to match the Hf–W and 182 fraction of Earth’s core formed during the lifetime of Hf U–Pb timescales. (Jacobsen 2005; Kleine et al. 2009). Consistent core formation timescales are obtained when The Pb isotope composition of the bulk silicate Earth more general accretion models than the exponential model provides evidence for a major U/Pb fractionation associated are used (FIG. 3). For instance, a rapid accretion of 63% of with Pb loss from Earth’s mantle 50–150 My after CAI Earth’s mass in ~1.5 My followed by a more protracted late formation (Wood and Halliday 2010). It has been a matter accretion matches both the Hf–W and U–Pb observations of much debate as to whether Pb was lost to the core or to in an equilibrium model (Rudge et al. 2010). It is a common space. More extreme is the proposal by Albarède (2009) misconception that in such equilibrium models core forma- that the U–Pb systematics of the bulk silicate Earth refl ect tion must have terminated ~30 My after CAI formation the late addition of volatile-rich material. However, recent (Jacobsen 2005; Albarède 2009). About 95% of the Earth experimental data show that Pb is siderophile under the has accreted by 30 My in the exponential model for Hf–W, conditions of Earth’s core formation, such that the elevated but more general accretion models reach the same 95% U/Pb ratio of Earth’s mantle at least in part resulted from accretion point at much later times. For instance, about core formation (Wood and Halliday 2010). 95% of the Earth is accreted at 52 My after CAI formation The isotopic data have traditionally been interpreted in the equilibrium model shown in FIGURE 3C. within the framework of an exponential model for Earth’s Another way to match the Hf–W and U–Pb timescales is accretion (Halliday 2006). While collisions during Earth’s to invoke disequilibrium during core formation (FIG. 3B). accretion were stochastic, Wetherill (1986) predicted that Both systems provide consistent timescales in the expo- these collisions occurred at an exponentially decreasing nential model if only 40% of the cores of accreted objects rate. For this reason the exponential model is often used fi rst re-equilibrated with Earth’s mantle before entering when calculating accretion timescales (FIG. 3). In such Earth’s core (Rudge et al. 2010). models, age information is given as the mean-life of accre- tion, τ, corresponding to the time taken to accrete 63% of Equilibrium or Disequilibrium Core Formation? Earth’s mass (Jacobsen 2005). The mean-life is the inverse Many of the planetary bodies that accreted to the Earth of a rate constant for exponential change. were pre-differentiated into mantle and core. Assessing the The end of accretion is not well defi ned in such a model degree of equilibration during Earth’s core formation thus as the fi nal ~10% of Earth’s mass was added by a single requires an understanding of what had happened to the collision, which was also responsible for the formation of cores of accreted bodies during their descent through the Moon. The end of accretion has been determined using Earth’s mantle. Effi cient equilibration would have occurred a two-stage model, which assumes that the core segregated if the impactor core completely emulsifi ed as small metal from a fully formed Earth in a single instant. However, droplets in the terrestrial magma ocean (FIG. 4A). This is Earth’s core formed continuously during the collisional likely to have happened in the case of small impactor-to- growth of the Earth and not by a single event at a well- target ratios (Rubie et al. 2007). What happened during defi ned point in time. The two-stage model nevertheless large collisions is poorly understood, however. Owing to provides useful age information as it corresponds to 95% the enormous length scales involved, the degree of equili- growth of Earth’s mass in the exponential model (Rudge bration is diffi cult to quantify (Rubie et al. 2007). However, et al. 2010); as such the two-stage model age provides a Dahl and Stevenson (2010) showed that during large colli- reasonable approximation of the time when Earth’s accre- sions the impactor metal cores would not have completely tion terminated in this model. emulsifi ed but would have sunk down into Earth’s core as unequilibrated blobs (FIG. 4D). In this case, some memory Two Ways to Make Sense of Tungsten of the accretion and differentiation history of Earth’s and Lead Isotopes precursor planetary bodies is retained in the isotopic and Using the exponential growth model and assuming that chemical composition of Earth’s mantle. core formation was an equilibrium process, the Hf–W data

ELEMENTS 43 FEBRUARY 2011 It has been thought likely that the distribution of sidero- formation but provide no information on when the last phile elements in Earth’s mantle refl ects equilibrium parti- 14% was added. In contrast, the U–Pb isotopic data contain tioning between metal and silicate in a deep terrestrial little information regarding Earth’s early accretion but magma ocean. This view was based on the observation that show that addition of the fi nal 10% of Earth’s mass must the depletions of siderophile elements in Earth’s mantle have begun by 120 My after CAI formation. provide such an excellent match to those expected from These bounds on Earth’s accretion timescale change partitioning experiments (FIG. 4B). However, disequilibrium depending on the degree of disequilibrium during core core formation models provide an equally good match to formation. FIGURE 4F shows an example of a disequilibrium the observed siderophile element abundances in Earth’s model with the degree of equilibrium being close to the mantle (FIG. 4E) (Rudge et al. 2010). Thus, the distribution of minimum value constrained by the Hf–W data. In this siderophile elements in Earth’s mantle provides no additional example, the Hf–W data require that ~50% of Earth’s mass information on the extent of core–mantle equilibrium. had accreted by ~60 My after CAI formation, while the other Tying Models of the Earth’s Core half could have been added any time later. However, combined with the U–Pb constraints, the last 10% of Earth’s mass must to an Accretion Timescale have been added by ~140 My after CAI formation. The discussion up to this point illustrates that there are many different ways to interpret the isotopic record of Dating the Moon and Earth’s Terminal Accretion Earth’s accretion and core formation. In general, it is not The Moon is thought to have formed at the very end of possible to unequivocally choose a particular accretion Earth’s accretion when a Mars-sized body collided with the model and timescale as the most probable scenario (Kleine proto-Earth. Determining the age of the Moon thus et al. 2009; Rudge et al. 2010). Nevertheless, the isotopic provides an independent constraint on how quickly the data allow some constraints to be placed on the fraction Earth formed. Dating the giant Moon-forming impact of Earth’s mass that must have accreted by a certain time. directly is diffi cult, however. The earliest possible time of The Hf–W system mainly constrains Earth’s early accretion, Moon formation is given by a Hf–W two-stage model age while the U–Pb system is mainly sensitive to the late stages of ~36 My after Solar System formation (Touboul et al. of Earth’s accretion (FIGS. 4C AND 4F). For instance, in equi- 2007). The latest possible time of Moon formation is given librium core formation models, the Hf–W data require that by the age of the ferroan anorthosites; they are the oldest at least 85% of the Earth was accreted by ~35 My after CAI

AD 3DUDPHWHUVIRUHTXLOLEULXPPRGHO E 'LVHTXLOLEULXPFRUHIRUPDWLRQB 

 +I²: 

 +I²: 8²3E  +I²: 8²3E 8²3E  +I²: 6KDSHSDUDPHWHU   8²3E

)UDFWLRQRIHTXLOLEUDWHGFRUHPDWHULDO      7LPHVFDOHSDUDPHWHUo 0\ 7LPH 0\DIWHU6RODU6\VWHPIRUPDWLRQ CF ([DPSOHDFFUHWLRQFXUYHVD G :LVRWRSHHYROXWLRQ a0\U IXOOHTX   

  a0\U HTX  :   a0\U IXOOHTX  2EVHUYHG%6(

 &+85 %XON(DUWK ²  a0\ a0\ a0\    )UDFWLRQRI(DUWK VPDVVDFFUHWHG W W W      7LPH 0\DIWHU6RODU6\VWHPIRUPDWLRQ 7LPH 0\DIWHU6RODU6\VWHPIRUPDWLRQ

%OXH (TXLOLEULXPH[SRQHQWLDOPRGHOFRQVLVWHQW 5HG (TXLOLEULXPPRGHOFRQVLVWHQW *UHHQ 'LVHTXLOLEULXPH[SRQHQWLDOPRGHO ZLWK+I²:EXWQRWZLWK8²3E ZLWK+I²:DQG8²3E FRQVLVWHQWZLWK+I²:DQG8²3E

Two ways to match the Hf–W and U–Pb timescales for in an exponential model. (C) Examples of accretion curves for the FIGURE 3 Earth’s accretion. (A) Parameters τ and β for the Earth. The numbers on the curves refer to values for τ. The end of Weibull accretion model of the form M(t) = 1 – exp{–(t/τ) β}, where accretion, as given by the time of 95% accretion, is different in M(t) is the fraction of Earth’s mass accreted at time t, τ is the mean- these three models. (D) W isotope evolution of Earth’s mantle for life of accretion, and β is the shape parameter of the Weibull func- the three accretion models shown in (C). BSE = Bulk Silicate Earth, tion. For β = 1 the exponential model is recovered. There is a CHUR = Chondritic Uniform Reservoir, full equ. = full equilibration, region of overlap (red) where Hf–W and U–Pb provide consistent 40% equ. = 40% equilibration. ε182W is the deviation from the W timescales. (B) Effect of disequilibrium during core formation on isotope composition in parts per 10,000; the present-day ε182W the calculated accretion timescales. The region of overlap (green) is of the BSE is 0 by defi nition. for ~40% equilibration, where Hf–W and U–Pb give identical results

ELEMENTS 44 FEBRUARY 2011 (48,/,%5,80&25()250$7,21 ',6(48,/,%5,80&25()250$7,21

AD 0DJPDRFHDQGLIIHUHQWLDWLRQ DG &RUHPHUJLQJ

8QHTXLOLEUDWHG 0ROWHQ LURQEOREV PDQWOH (TXLOLEUDWHG PHWDOGURSOHWV &RPSOHWHO\ PROWHQPDQWOH 6ROLG 0HWDO PDQWOH GLDSLU

&RUH &RUH

BE 6LGHURSKLOHDEXQGDQFHVLQ(DUWK VPDQWOH EH 6LGHURSKLOHDEXQGDQFHVLQ(DUWK VPDQWOH

%XON(DUWKFRPSRVLWLRQ  %XON(DUWKFRPSRVLWLRQ 

 

2EVHUYHG 2EVHUYHG &KRQGULWHQRUPDOL]HG &KRQGULWHQRUPDOL]HG 3UHGLFWHG 3UHGLFWHG FRPSRVLWLRQRI(DUWKҊVPDQWOH FRPSRVLWLRQRI(DUWKҊVPDQWOH   7L *D =Q 6L 7D 1E 0Q &U 9 &X )H 3E &R 3 1L : 7L *D =Q 6L 7D 1E 0Q &U 9 &X )H 3E &R 3 1L :

CF %RXQGVRQ(DUWK VDFFUHWLRQ I %RXQGVRQ(DUWK VDFFUHWLRQF

 0RRQIRUPLQJLPSDFW  0RRQIRUPLQJLPSDFW IRUELGGHQ UHJLRQ +I:  IRUELGGHQ IRUELGGHQ  UHJLRQ +I: UHJLRQ 83E

  IRUELGGHQ

a0\ UHJLRQ 83E

 

a0\   )UDFWLRQRI(DUWKDFFUHWHG )UDFWLRQRI(DUWKҊVPDVVDFFUHWHG        7LPH 0\DIWHU6RODU6\VWHPIRUPDWLRQ 7LPH 0\DIWHU6RODU6\VWHPIRUPDWLRQ

(A) Equilibrium core formation: sinking metal droplets (D) Disequilibrium during core formation: some fraction of the FIGURE 4 aggregate to a metal pond at the base of a magma metal core does not emulsify as small metal droplets but descends ocean; from there, the metal descends as diapirs through a mostly through the mantle as large iron blobs, leaving little opportunity to solid lower mantle. (B) Present-day depletions of some siderophile equilibrate with the magma ocean. (E) Same as (B) but assuming elements in Earth’s mantle compared to those predicted based on that only 40% of the newly accreted metal core equilibrated with metal–silicate equilibration in a deep magma ocean. (C) Bounds Earth’s mantle. (F) Same as (C) but for only 40% equilibration. on Earth’s accretion in equilibrium core-formation models. Shaded areas are forbidden. One possible accretion curve is shown. lunar rocks and formed ~4.46 billion years ago (Norman sible, however, that a signature of an early differentiation et al. 2003). within Earth’s mantle remained isolated during ongoing Another way to deduce the earliest possible time of Moon core formation and also survived the giant Moon-forming impact. This enigma is easily resolved if Earth is non- formation is to use the indistinguishable W isotope compo- 142 sitions of the silicate Earth and Moon, which suggest that chondritic, because then the Nd excess of Earth’s mantle the Moon formed more than ~50 My after Solar System does not require mantle differentiation prior to formation formation (Touboul et al. 2007). This age is consistent with of the Moon (Kleine et al. 2009). The Moon and Mars may most Hf–W model ages for the Earth, as well as with the U–Pb also have non-chondritic compositions for refractory elements (Caro et al. 2008). age of the termination of Earth’s core formation (FIG. 1). O’Neill and Palme (2008) proposed that Earth may have A Non-Chondritic Earth? acquired a non-chondritic composition by collisional One of the basic assumptions when using isotopic data to erosion of early-formed crust on precursor planetary model Earth’s differentiation is that refractory elements bodies. Since crust is enriched in incompatible elements occur in precisely chondritic relative abundances (see and has a lower-than-chondritic Sm/Nd ratio, preferential above). However, Earth’s 142Nd – 182W record is more easily removal of crustal material leads to bulk planetary bodies understood if Earth has non-chondritic relative abun- with higher-than-chondritic Sm/Nd. In this case the 142Nd dances of refractory elements. The small 142Nd excess of excess of Earth’s mantle relative to chondrites refl ects the Earth’s mantle relative to chondrites suggests differentia- Nd isotope composition of the bulk Earth and does not tion of the silicate Earth well before 30 My (Boyet and require an early differentiation of its mantle. A non- Carlson 2005), that is, before Earth’s core had fully segre- chondritic composition of the Earth affects the Hf–W gated and before formation of the Moon. It seems implau- chronology of core formation and results in more protracted

ELEMENTS 45 FEBRUARY 2011 timescales compared to those calculated for a chondritic to preferential erosion of early-formed crust from the Earth. For instance, the equilibrium two-stage model age embryos, causing non-chondritic compositions, even for is ~40 My instead of ~30 My after CAI formation (Kleine elements that long were thought to occur in constant rela- et al. 2009). tive proportions throughout the Solar System. Collisions during Earth’s accretion also led to melting and core forma- SUMMARY AND OUTLOOK tion, but exactly how the metal segregated towards Earth’s core remains unclear. It has been thought likely that metal– A lot happened in the fi rst few million years of the Solar silicate equilibration took place between metal droplets System. About 4567 My ago, CAIs formed close to the and the surrounding molten silicate in a magma ocean, young Sun, defi ning the start of the Solar System. It took but there is now evidence that during giant impacts a frac- less than ~1 My to accrete and differentiate the fi rst proto- tion of the impactor’s metal core directly merged with planets, evidence of which are the iron meteorites. These Earth’s core, leaving no opportunity to equilibrate within may have been derived from bodies that were disrupted the terrestrial magma ocean. A memory of the differentia- by hit-and-run collisions early in Solar System history. tion of Earth’s precursor planetary bodies is thus retained Chondrite parent bodies had a less violent history, perhaps in the chemical and isotopic composition of Earth’s mantle. because they formed a few million years later and further out in the Solar System. Although the chondrite parent Our understanding of how the Earth formed is thus bodies assembled late and in an orbit beyond Mars, some changing from the relatively simple picture of a chondritic chondritic meteorites contain CAIs that formed close to Earth that internally differentiated by a one-stage metal– the active young Sun. This leaves little doubt that there silicate equilibration process to a more complex picture of has been substantial outward transport, over several AU, a dynamically evolving planet involving collisional erosion of material. While we understand how CAIs were incorpo- and metal–silicate disequilibrium. The implications for rated into small bodies in the outer asteroid belt and compositional models of the Earth are profound and open beyond, the processes by which chondrules formed and up new avenues of research that will likely lead to a new then accreted to chondrite parent bodies are still unclear. understanding of how the Earth formed and fi rst evolved. Further progress in this area will require reconciling dynamical models of the solar nebula with results from ACKNOWLEDGMENTS meteorite chronometry. We thank reviewers Daniel Apai, Harold Connolly Jr, After the rapid formation of planetary embryos, accretion Michael Drake and Hap McSween, and guest editor Dante slowed down, and the Earth and Moon were fully formed Lauretta for their constructive comments and suggestions. no earlier than ~50 My after Solar System formation. Thorsten Kleine acknowledges funding by the Collisions during terrestrial planet formation may have led Schweizerischer Nationalfonds (PPOOP2_123470). REFERENCES Dahl TW, Stevenson DJ (2010) Turbulent Philosophical Transactions of the Royal mixing of metal and silicate during Society of London A 366: 4205-4238 Albarède F (2009) Volatile accretion planet accretion — An interpretation of Rubie DC, Nimmo F, Melosh HJ, Gerald S history of the terrestrial planets and the Hf–W chronometer. Earth and (2007) Formation of Earth’s core. In: dynamic implications. Nature 461: Planetary Science Letters 295: 177-186 1227-1233 Evolution of the Earth. Treatise on Halliday AN (2006) The origin of the Geophysics 9, Elsevier, Amsterdam, Alexander CMD (2005). From supernovae Earth: What’s new? Elements 2: 205-210 pp 51-90 to planets: the view from meteorites and interplanetary dust particles. In: Jacobsen SB (2005) The Hf-W isotopic Rudge JF, Kleine T, Bourdon B (2010) Krot et al. (eds) Chondrites and the system and the origin of the Earth and Broad bounds on Earth’s accretion Protoplanetary Disk. Astronomical Moon. Annual Review of Earth and and core formation constrained by Society of the Pacifi c Conference Series, Planetary Sciences 33: 531-570 geochemical models. Nature Geoscience 3: 439-443 pp 972-1001 Jacobsen B, Yin Q-Z, Moynier F, Amelin Amelin Y, Kaltenbach A, Iizuka T, Stirling Y, Krot AN, Nagashima K, Hutcheon ID, Stevenson DJ (2008) A planetary perspec- 26 26 CH, Ireland TR, Petaev M, Jacobsen SB Palme H (2008) Al- Mg and tive on the deep Earth. Nature 451: 207 206 (2010) U–Pb chronology of the Solar Pb- Pb systematics of Allende CAI: 261-265 Canonical solar initial 26Al/27Al ratio System’s oldest solids with variable Touboul M, Kleine T, Bourdon B, Palme 238 235 reinstated. Earth and Planetary Science U/ U. Earth and Planetary Science H, Wieler R (2007) Late formation and Letters 272: 353-364 Letters 300: 343-350 prolonged differentiation of the Moon Bottke WF, Nesvorný D, Grimm RE, Kleine T, Touboul M, Bourdon B, Nimmo inferred from W isotopes in lunar Morbidelli A, O’Brien DP (2006) Iron F, Mezger K, Palme H, Jacobsen SB, Yin metals. Nature 450: 1206-1209 QZ, Halliday AN (2009) Hf-W chro- meteorites as remnants of planetesimals Trinquier A, Elliott T, Ulfbeck D, Coath nology of the accretion and early evolu- formed in the terrestrial planet region. C, Krot AN, Bizzarro M (2009) Origin of tion of asteroids and terrestrial planets. Nature 439: 821-824 nucleosynthetic isotope heterogeneity Geochimica et Cosmochimica Acta 73: in the solar protoplanetary disk. Science Bouvier A, Wadhwa M (2010) The age of 5150-5188 the Solar System redefi ned by the oldest 324: 374-376 Krot AN and 15 coauthors (2009) Origin Pb–Pb age of a meteoritic inclusion. Villeneuve J, Chaussidon M, Libourel G and chronology of chondritic compo- Nature Geoscience 3: 637-641 (2009) Homogeneous distribution of nents: A review. Geochimica et 142 26Al in the solar system from the Mg Boyet M, Carlson RW (2005) Nd Cosmochimica Acta 73: 4963-4997 evidence for early (>4.53 Ga) global isotopic composition of chondrules. differentiation of the silicate Earth. Norman MD, Borg LE, Nyquist LE, Bogard Science 325: 985-988 DD (2003) Chronology, geochemistry, Science 309: 576-581 Wetherill (1986) Accumulation of the and petrology of a ferroan noritic anor- terrestrial planets and implications Brennecka GA, Weyer S, Wadhwa M, thosite clast from Descartes breccia concerning lunar origin. In: Hartmann Janney PE, Zipfel J, Anbar AD (2010) 67215: Clues to the age, origin, struc- 238 235 WK et al. (eds) Origin of the Moon. U/ U variations in meteorites: ture, and impact history of the lunar 247 Lunar and Planetary Institute, Houston, Extant Cm and implications for crust. Meteoritics & Planetary Science pp 519-550 Pb-Pb dating. Science 327: 449-451 38: 645-661 Wood BJ, Halliday AN (2010) The lead Caro G, Bourdon B, Halliday AN, Quitté Nyquist LE, Kleine T, Shih CY, Reese YD isotopic age of the Earth can be G (2008) Superchrondritic Sm/Nd ratios (2009) The distribution of short-lived explained by core formation alone. in Mars, the Earth and the Moon. radioisotopes in the early solar system Nature 465: 767-771 Nature 452: 336-339 and the chronology of asteroid accre- Chambers JE (2004) Planetary accretion tion, differentiation, and secondary Yang J, Goldstein JI, Scott ERD (2007) in the inner Solar System. Earth and mineralization. Geochimica et Iron meteorite evidence for early forma- Planetary Science Letters 223: 241-252 Cosmochimica Acta 73: 5115-5136 tion and catastrophic disruption of protoplanets. Nature 446: 888-891 Ciesla FJ (2010) The distributions and O’Neill HSC, Palme H (2008) Collisional ages of refractory objects in the solar erosion and the non-chondritic compo- nebula. Icarus 208: 455-467 sition of the terrestrial planets.

ELEMENTS 46 FEBRUARY 2011