Identification of the giant impactor Theia in lunar rocks

See allHide authors and affiliations

Science  06 Jun 2014:
Vol. 344, Issue 6188, pp. 1146-1150
DOI: 10.1126/science.1251117

An analysis of motes of the Moon maker

How did the Moon form? According to the prevailing hypothesis, a Mars-sized body known as Theia smashed into Earth. Herwartz et al. analyzed fresh basalt samples from three Apollo landing sites and compared them with several samples of Earth's mantle. The oxygen isotope values measured in these lunar rocks differ significantly from the terrestrial material, supporting the giant-impact hypothesis.

Science, this issue p. 1146


The Moon was probably formed by a catastrophic collision of the proto-Earth with a planetesimal named Theia. Most numerical models of this collision imply a higher portion of Theia in the Moon than in Earth. Because of the isotope heterogeneity among solar system bodies, the isotopic composition of Earth and the Moon should thus be distinct. So far, however, all attempts to identify the isotopic component of Theia in lunar rocks have failed. Our triple oxygen isotope data reveal a 12 ± 3 parts per million difference in Δ17O between Earth and the Moon, which supports the giant impact hypothesis of Moon formation. We also show that enstatite chondrites and Earth have different Δ17O values, and we speculate on an enstatite chondrite–like composition of Theia. The observed small compositional difference could alternatively be explained by a carbonaceous chondrite–dominated late veneer.

Earth’s Moon is distinct among the >150 moons of our solar system (1). Most other moons are either captured planetesimals, or they formed along with the planet in a common accretion disc. In contrast, it is hypothesized that our satellite formed ~4.5 billion years ago from the debris of a giant collision between the proto-Earth and another smaller proto-planet [giant impact hypothesis (2, 3)]. Some of the distinct features of the Moon—such as the depletion in moderately volatile elements and water, the small lunar core, and the angular momentum of the Earth-Moon system—are interpreted as products of the energetic collision with Theia (1).

Most numerical models of the collision assume that Theia was about the size of Mars and collided with the proto-Earth at an oblique angle. These classic collision models predicted that the Moon is made of 70 to 90% Theia [mass fraction of impactor (Mi)] and only 10 to 30% proto-Earth [mass fraction of proto-Earth (M)] material (2, 3). Such a large fraction of Theia in the Moon, however, is difficult to reconcile with the observed isotopic similarity between the Moon and Earth (4). Recent simulations take this into account and aim to decrease the compositional difference between Earth and the Moon (see below).

Measurements of isotope ratios of terrestrial, martian, and asteroidal samples show that the bodies in the early solar system were isotopically heterogeneous (5). It is therefore expected that Theia and proto-Earth were isotopically distinct. If the Moon formed predominantly from fragments of Theia, as predicted by most numerical models, the Moon and Earth should differ in their isotopic composition. However, no isotopic differences between Earth and the Moon have yet been recognized; for instance, for O (610), Ti (11), Ca (12), Si (13), or W (14). We argue herein that careful reinvestigation of the published data sets for O and Ti hint at small variations between Earth and the Moon.

Three explanations exist for the paradox of identical isotopic compositions of Earth and the Moon: (i) formation of proto-Earth and Theia at similar heliocentric distances from the same isotopic reservoir, resulting in identical compositions of proto-Earth and Theia (8); (ii) isotopic reequilibration in the aftermath of the giant impact that has obliterated the initial heterogeneity (4); or (iii) less compositional difference between Earth and the Moon than predicted by classic numerical simulations (1517).

Recent collision models that aim to be consistent with isotope measurements proposed a larger (15), smaller (16), or faster impactor (16, 17). In models with small impactors that assume a fast-spinning proto-Earth (16), the Theia component in the Moon can be reduced down to ~8 weight percent (wt %), whereas Earth receives ~2 wt % (16); the Moon and the postimpact mantle then differ by only a few weight percent. Simulations with very large impactors can produce a Moon and Earth that are each composed of ~50% Theia material, and in such cases, the Moon and the postimpact Earth are compositionally identical down to the 0.1–wt % level (15).

The size of Theia is unknown, so to report compositional variation between Earth and the Moon independent of that size, we introduce the value δfTEmbedded Imagewhere F is the mass fraction of proto-Earth material (target) in the Moon (FM,tar) and in the final planet (FP,tar), respectively. The δfT quantifies the percent compositional deviation of the Moon (or disk) from Earth (15, 17). Most models predict fractionally more impactor material in the Moon than in Earth; thus, δfT is usually negative. However, some simulations with large impactors also predict positive δfT and, thus, larger fractions of Theia in Earth than in the Moon (15).

The predicted range of δfT is large, and from the numerical models alone it is impossible to decide which one is the most likely. Independent estimates of δfT or Mi can thus help to constrain the kinematics of the collision. To our knowledge, only one independent estimate exists, which is based on the variable Nb/Ta ratio in Earth’s mantle (14.0 ± 0.3), the Moon (17.0 ± 0.8), and chondrites (19.9 ± 0.6) (18). The depletion in Nb relative to Ta in the bulk silicate Earth probably results from the more siderophilic nature of Nb over Ta at high pressures and sequestration of some Nb in Earth’s core. Assuming that Theia was Mars-sized and had a chondritic Nb/Ta ratio, the most likely mass fraction of Theia in the Moon was estimated to Mi ~ 30 to 50 wt % (18). This translates to δfT ~ –21 to –44 for the assumed Mars-sized impactor.

If the giant impact hypothesis holds, the isotopic composition of lunar rocks is a function of impactor composition and impactor mass fraction. Large isotopic variations among solar system bodies are observed for O (5, 7). Oxygen has three stable isotopes (16O, 17O, and 18O) and displays large mass-independent variations among solar system bodies that are expressed in the form of the Δ17O value (5). Oxygen isotopes are therefore ideal for tracing the isotope composition of Theia in lunar rocks. Previous oxygen isotope studies did not find any difference between Earth and the Moon (610), and Wiechert et al. (8) concluded from their high-precision data that Earth and the Moon are identical within 5 parts per million (ppm) [0.005 per mil (‰)] in Δ17O. However, careful reinvestigation of the three high-precision data sets (810) reveals that lunar samples are elevated, on average, by 9 ppm compared with the terrestrial UWG-2 standard used in all three studies (fig. S1). With an improved analytical technique (19), we are able to measure variations in parts per million (0.00x‰) in Δ17O (see supplementary materials and methods), allowing us to detect the isotopic differences between Earth and the Moon.

We first attempted to use lunar meteorites to determine the isotope composition of the Moon. Our data show that terrestrial weathering modified the Δ17O of the studied lunar meteorites, making identification of small variations impossible (fig. S2). Therefore, we have analyzed fresh basalt samples from three Apollo landing sites that were provided by NASA. We compared the composition of the lunar basalts with the composition of Earth mantle xenoliths and mantle-derived melts [mid-ocean ridge basalt (MORB)] (Table 1 and Fig. 1).

Table 1 Triple oxygen isotope analysis of terrestrial, lunar, and enstatite chondrite samples.

Delta prime notations (δ′) are calculated according to (19), and Δ17O is equal to δ′17O – 0.5305 * δ′18O. The enstatite chondrite samples identified by “MS” (from Sudan) represent individual meteorite fragments of high- and low-iron enstatite chondrites of various petrological types from the Almahata Sitta polymict breccia (31). n indicates the number of single analysis.

View this table:
Fig. 1 Terrestrial (squares) and lunar (circles) samples and ECs (triangles) in δ′18O versus ∆17O space.

VSMOW, Vienna standard mean ocean water. Error bars denote 1σ SEM.

The bulk silicate Earth (BSE) is constrained to a Δ17O value of –0.101 ± 0.002‰ [1σ SEM, n = 65 measurements (19)] from mantle xenoliths and MORB from seven localities around the world [for definitions and analytical details, see (19, 20)]. Earth mantle minerals and MORB fall on a common mass-fractionation line, with a slope of θ = 0.532 ± 0.006, in the δ′17O versus δ′18O space. This slope is, within uncertainty, identical to the high-temperature approximation for equilibrium oxygen isotope fractionation of 0.5305 (21). The Δ17O values of Earth mantle minerals and MORB are identical within the uncertainty. This is in agreement with mass fractionation upon melt extraction along a slope of ~0.53 for high-temperature processes (19, 21). Because the same slope must apply to melt extraction from the lunar mantle, we can safely use the lunar basalts as analogs for the bulk silicate Moon (BSM) with respect to Δ17O.

The three lunar basalts span a small range in δ18O and have an average Δ17O = –0.089 ± 0.002‰ (1σ SEM, n = 20). We suggest that this value is representative of the BSM. Thus, the Δ17O of the Moon is 12 ± 3 ppm (0.012 ± 0.003‰) higher than that of Earth (Fig. 2). This unequivocally identifies an isotopic difference between Earth and the Moon and supports the view that the Moon formed by a giant collision of the proto-Earth with Theia.

Fig. 2 Δ17O composition for terrestrial and lunar samples.

A slope of 0.5305 and zero intercept (VSMOW) is used to calculate Δ17O (19, 20). Error bars are 1σ SEM. Solid vertical lines denote mean values for the BSE and the Moon. Gray shaded areas represent 1σ SEM, and dotted lines represent 2σ SEM.

The reevaluation of lunar O isotope data from three previous studies (810) is consistent with our finding within the respective uncertainties (fig. S1) (20). Thus, our new data are not in conflict with previous studies, and the slightly elevated Δ17O composition of the Moon was already present in these data sets (810).

Carbonaceous chondrites and BSE differ considerably, not only in Δ17O, but also in their Ti, Cr, and Ni isotope composition (22), making a carbonaceous chondrite composition of Theia unfeasible. Rather, Theia formed from the same large noncarbonaceous chondrite (22) reservoir as Earth, Mars, ordinary chondrites (OCs), enstatite chondrites (ECs), and other noncarbonaceous chondrites and achondrites (22). All numerical simulations, except some with very large impactors, predict that the Moon received fractionally more impactor material than Earth (negative δfT). Hence, the Δ17O of Theia was most likely higher than that of Earth and the Moon. The solar system bodies from the noncarbonaceous chondrite (22) reservoir with a higher Δ17O than that of Earth are Mars and the parent asteroids of OC and R-chondrites (7). Admixing only 4% of material isotopically resembling Mars would be sufficient to explain the observed 12 ppm difference between Earth and the Moon. For an OC or R-chondrite composition, less than 2% would be required in the Moon. Such small proportions are inconsistent with most numerical models (2, 3, 1517) that generally suggest larger fractions of Theia in the Moon (Fig. 3A). This implies that the composition of Theia was only slightly higher in Δ17O than that of Earth.

Fig. 3 Mass balance between possible impactors and proto-Earth.

(A) The percent compositional difference (δfT) between the Moon and Earth, as predicted by classical (2, 3) and more recent (1517) numerical simulations. Estimates derived from Nb/Ta systematics (18) are displayed for comparison (assuming a Mars-sized impactor). Black lines represent the range of most simulations; black bars represent typical values for seemingly realistic model runs. (B) Mass balance estimates for oxygen isotopes between Earth and several possible impactors (LL and H ordinary chondrites, CI, Mars, EL, and EH). Because δfT and Mi are related for a given impactor size, both quantities can be displayed on the same figure. A Mars-sized impactor with mTheia/m = 0.12 is used to estimate Mi; thus, Earth contains ~11% of Theia component. Mixing lines (black lines) between the possible impactors and proto-Earth are defined by their Δ17O isotopic composition (i.e., Δ17O at 100% Mi) and the fraction of Theia material in Earth (i.e., Δ17O = 0.101‰ at 11% Mi). Extrapolation of the mixing lines to 0% Mi indicates potential Δ17O proto-Earth compositions for the given Mars-sized impactor. Shaded areas represent 1σ SEM. A Mars-sized impactor with EH composition would result in a moon within the dark grey ellipse with Mi ~ 43%.

We have obtained new data on ECs that were previously assumed to be identical to Earth in Δ17O (23, 24). Our data show a difference of 59 ± 8 ppm (1σ SEM, n = 14) between Earth and EL (low iron) enstatite chondrites and 35 ± 10 ppm (1σ SEM, n = 10) between Earth and more metal-rich EH (high iron) enstatite chondrites. If the oxygen isotopic composition of Theia resembled that of EL or EH chondrites, δfT is –21 ± 9% for EL and –36 ± 15% for EH, respectively (Fig. 3B). Assuming a Mars-sized impactor (mTheia/m = 0.12), this translates to Mi = 30 ± 8% (EL) or 43 ± 13% (EH), respectively (Fig. 3B). These estimates are in good agreement with the more recent numerical models (1517) and estimates from Nb/Ta mass balance considerations (Fig. 3A) (1518).

If Theia compositionally resembled EC, one would expect the Moon to fall on a mixing trend for isotope systems that show differences between ECs and Earth. High-precision Ti isotope measurements suggest that Earth and ECs are different (11). The datum for the Moon indeed falls between Earth and ECs (fig. S3) (20). [Note that Zhang et al. (11) argued for an identical 50Ti composition of Earth and the Moon.] Earth and ECs also differ slightly in Ca (12) and Si isotopes (13), but variations between Earth and the Moon were not detected. Thus, only the O and possibly the Ti isotopes indicate that Theia may have resembled ECs. The Moon is a mixture between Theia and proto-Earth, not only isotopically, but also chemically. For refractory lithophile elements, mass balance modeling shows that an EC composition of Theia is feasible (fig. S4). The elevated FeO content in the lunar mantle, however, is not consistent with the reduced state and, thus, low FeO content of ECs, although their bulk Fe content is high (EH = 33 wt % and EL = 24 wt %) (25).

Enstatite chondrites are sometimes regarded as the sole building blocks of Earth (25). This highly debated theory was supported by the isotopic similarity of EC and Earth in O (23, 24), Cr (26), or Ni (27), but it has been disputed, for instance, on the basis of Si isotopes (13). The differences between ECs and Earth found for Δ17O (this study), Ti (11), and Ca isotopes (12) are also inconsistent with ECs as the sole building blocks of Earth (25).

Hence, there is no chondrite group that can be regarded as the sole building blocks of Earth. Likewise there is probably no meteorite group that is identical in composition to Theia. Although Theia may have been compositionally similar to ECs, its composition was probably distinct and is not represented by any known meteorite group.

Pahlevan and Stevenson (4) have modeled likely compositions of potential impactors in terms of oxygen isotopic composition. In this model, >10% of all impactors would have a Δ17O value that is more similar to proto-Earth than to EH chondrites (fig. S5). This would allow more than 43 wt % Theia in the Moon and would give numerical models more elements of freedom (1517). If impact models that predict >70% impactor component in the Moon are correct, the Δ17O of Theia must have been <18 ppm higher in Δ17O than the proto-Earth. This conforms to ~6% of all modeled impactors (4). Thus, a small chance remains that numerical models predicting a large fraction of Theia in the Moon hold (2, 3).

The size of Theia is not known, and current assumptions for the impactor–to–proto-Earth mass ratio (mTheia/m) range over one order of magnitude (mTheia/m = 0.025 to 0.45) (15, 16). If a reservoir on Earth exists that has escaped equilibration with the Theia component in Earth, Theia’s size could be constrained. Theia has changed the composition of the proto-Earth as a function of mTheia/m and impactor composition. If Theia was Mars-sized (mTheia/m ~ 0.12) with an EC-like composition, proto-Earth was 4 ppm lower in Δ17O before mixing with Theia. For smaller impactors (mTheia/m = 0.025 to 0.048) (16), this effect shrinks to 1 to 2 ppm, as less material is added to the proto-Earth. In contrast, a large impactor with mTheia/m = 0.4 to 0.45 (15) would suggest a value of Δ17O that is 14 to 27 ppm lower for the proto-Earth than for BSE today (Fig. 3 and fig. S6). Currently, no reservoir on Earth is known to have escaped equilibration with Earth’s Theia component; thus, it is presently not possible to obtain information on Δ17O of the proto-Earth.

An alternative explanation for the isotope difference between Earth and the Moon is that the Δ17O value of Earth was modified by late-accreting material (late veneer) after the formation of the Moon. Such material may have had a Δ17O value lower than that of Earth. In the following scenario, Earth and the Moon had identical Δ17O compositions after the giant impact (4, 8, 1517), with the Moon now representing the composition of Earth before the late veneer.

From the overabundance of siderophile elements in Earth’s mantle, late veneer masses between 0.1 and 0.5% of BSE have been suggested (28). Higher estimates between 0.3 to 0.8% of BSE are derived from 182W isotope systematics (29). Here, we use 0.5% for mass balance considerations. The only groups of undifferentiated meteorites that have Δ17O much lower than that of Earth are carbonaceous chondrites (30). Assuming that the late veneer component had an oxygen isotope composition of CV carbonaceous chondrites (Δ17O = –4‰) (30), incorporation of 0.5% late veneer component would decrease the Δ17O of Earth by as much as ~20 ppm. Similar proportions of material resembling CO, CR, or CH carbonaceous chondrites would also be sufficient to explain the observed difference of 12 ppm between Earth and the Moon.

If this scenario is correct, the observed difference in Δ17O between Earth and the Moon points toward a late veneer that is dominated by carbonaceous chondrites. Among the carbonaceous chondrites, the most water-rich meteorites are found. It would thus be feasible that part of the ocean water was delivered by the late veneer impactors. Evidence for this scenario may be found in old crust that has escaped mixing with the late veneer component (29).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

Table S1

References (3241)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We thank NASA for providing samples; Z. Sharp, F. Wombacher, C. Münker, and H. Palme for discussions; three anonymous reviewers for constructive comments; and E. Barkan and B. Luz for analyzing our reference O2 relative to SMOW. All data are provided within the manuscript or the supplementary materials.
View Abstract

Navigate This Article