Earth’s water may have been inherited from material similar to enstatite chondrite meteorites

See allHide authors and affiliations

Science  28 Aug 2020:
Vol. 369, Issue 6507, pp. 1110-1113
DOI: 10.1126/science.aba1948

An unexpected source of Earth's water

The abundances of Earth's chemical elements and their isotopic ratios can indicate which materials formed Earth. Enstatite chondrite (EC) meteorites provide a good isotopic match for many elements but are expected to contain no water because they formed in the hot inner Solar System. This would require Earth's water to be from a different source, such as comets. Piani et al. measured hydrogen contents and deuterium/hydrogen ratios (D/H) in 13 EC meteorites (see the Perspective by Peslier). They found far more hydrogen than is commonly assumed, with D/H close to that of Earth's mantle. Combining these data with cosmochemical models, they show that most of Earth's water could have formed from hydrogen delivered by EC meteorites.

Science, this issue p. 1110; see also p. 1058


The origin of Earth’s water remains unknown. Enstatite chondrite (EC) meteorites have similar isotopic composition to terrestrial rocks and thus may be representative of the material that formed Earth. ECs are presumed to be devoid of water because they formed in the inner Solar System. Earth’s water is therefore generally attributed to the late addition of a small fraction of hydrated materials, such as carbonaceous chondrite meteorites, which originated in the outer Solar System where water was more abundant. We show that EC meteorites contain sufficient hydrogen to have delivered to Earth at least three times the mass of water in its oceans. EC hydrogen and nitrogen isotopic compositions match those of Earth’s mantle, so EC-like asteroids might have contributed these volatile elements to Earth’s crust and mantle.

The origin of Earth’s water is debated. The isotopic composition of Earth suggests that it is composed of material from the inner Solar System, such as enstatite chondrite (EC) meteorites (13). The inner Solar System was too warm to have retained water ice, so terrestrial water is thought to have been supplied by hydrated materials that originally formed in the outer Solar System before migrating inward (4, 5).

Hydrogen isotopic compositions are conventionally expressed as δD ≡ [(D/H)sample/(D/H)SMOW – 1] × 1000, where D/H is the ratio of 2H (deuterium) to 1H (protium) and standard mean ocean water (SMOW) is the mean value of Earth’s oceans. δD varies greatly among Solar System objects. The bulk protosolar nebula had δD = –865 per mil (‰), estimated from the solar wind and remnant H2 in the atmospheres of the giant planets (6). Compared with this protosolar value, all other materials in the Solar System are enriched in deuterium, with inner Solar System objects having intermediate values [e.g., δD ≡ 0‰ in terrestrial oceans and –165 to +800‰ in asteroids, as recorded in primitive meteorites (7)]. Outer Solar System objects, such as comets, are even more enriched in deuterium [e.g., up to δD = 2400‰ in comet 67P/Churyumov-Gerasimenko (8)]. Among meteorites, water-rich carbonaceous chondrites (CCs) of Ivuna-type (CI) and Mighei-type (CM) are potential sources of terrestrial water because their δD values are distributed around a value coincident with Earth’s oceans—i.e., δD ≡ 0‰ (9). Dynamical simulations suggest that C-type (carbonaceous) asteroids, whose optical properties suggest a link to water-rich CCs, could have been scattered into the inner Solar System from beyond the orbit of Jupiter (10), potentially delivering water to Earth during or after its main phase of accretion, in the first few million years of Solar System formation.

An alternative possibility is that EC-like materials—which have similar isotopic abundances to those of Earth for elements including oxygen, titanium, calcium, and molybdenum (3)—could contain enough hydrogen with the appropriate isotopic composition to provide water to the growing proto-Earth. Efforts to test this scenario have been hampered by difficulties in measuring H concentrations and isotopic ratios in ECs, owing to their assumed low H abundances and the potential for terrestrial contamination (7).

We measured H abundances and isotopic compositions for a suite of 13 EC meteorites that spans the full range of thermal metamorphism degrees [from petrologic type EH3 (the least metamorphosed) to EH6 (those most altered by thermal metamorphism on their parent asteroid)]. We covered this range to evaluate potential effects of parent-body processing, which could have altered the original water content and H isotopic composition (11). We also analyzed one aubrite meteorite (Norton County), which represents an evolved planetary body of enstatite meteorite composition for which metal-silicate segregation (differentiation) had occurred (11). Hydrogen measurements were performed at the Centre de Recherches Pétrographiques et Géochimiques (CRPG, Nancy, France), either on bulk samples with an elemental analyzer (EA) coupled to a stable isotope ratio mass spectrometer (IRMS) or at the micrometer scale by secondary ion mass spectrometry (SIMS) (11).

We found that the ECs have bulk hydrogen contents (reported as water equivalents) ranging from 0.08 to 0.54 wt % H2O (Fig. 1A and table S2). The EH3 meteorites are generally richer in H (0.44 ± 0.04 wt % H2O; all uncertainties are one standard deviation unless otherwise noted) than the more metamorphosed groups EH4 (0.2 ± 0.1 wt % H2O), EH5 (0.3 ± 0.1 wt % H2O), and EL6 (0.3 ± 0.3 wt % H2O). The Norton County aubrite is 0.3 ± 0.2 wt % H2O. These EC water contents are well below those measured in the water-rich CCs Orgueil (CI), Alais (CI), and Murchison (CM), which are 7.2 to 9.1 wt % H2O (Fig. 1A and table S2). However, all analyzed ECs contain hydrogen concentrations above the EA-IRMS detection limit [~0.05 wt % H2O (11)]. The least-altered EH chondrites (EH3 and EH4) have δD values averaging δD = –103 ± 3‰, systematically below the current value of Earth’s oceans, whereas metamorphosed ECs (EH5 and EL6) and the aubrite have even lower values, averaging δD = –127 ± 15‰ (Fig. 1B and table S2).

Fig. 1 Hydrogen contents and isotopic compositions of ECs.

(A) Bulk-rock hydrogen content, reported as water equivalent by weight on a logarithmic scale. (B) Bulk-rock deuterium abundance, expressed as an offset from standard mean ocean water. (C) Mesostasis hydrogen content. (D) Mesostasis deuterium abundance. The CM- and CI-type CCs Orgueil, Alais, and Murchison, as well as the aubrite Norton County (aub.), are included in (A) and (B) for comparison. Data for the primitive EC Sahara 97096 (SAH97096; EH3.1-3.4) and the two CV-type CCs Vigarano and Kaba are shown in (C) and (D) for comparison (11). Different symbols in (C) and (D) indicate different chondrules (labeled CH). Error bars are 2σ and are smaller than the symbol size in (B). The data are listed in tables S2 and S4.

If the hydrogen contents had been entirely due to atmospheric contamination during sample preparation, constant D/H ratios would be observed despite variable H concentrations. Instead, our data follow a negative trend in a diagram of δD versus 1/H (fig. S5), indicating possible loss of D-enriched hydrogen-bearing organic molecules (12) during thermal metamorphism within the EC parent asteroid(s). In our least-metamorphosed sample, the pristine EH3 meteorite Sahara 97096 [EH3.1-3.4 (13); a fragment of the same meteorite as Sahara 97116, see table S2], only a thin (≈500 μm) surface layer of terrestrial alteration is visible (11). Its inner portions remain unaffected, as attested by the presence of unaltered grains of oldhamite (CaS), a highly hydrophilic mineral (14, 15). The fresh interior of Sahara 97096 has lower (and constant) H contents (0.5 ± 0.1 wt % H2O) than the altered parts (3.0 ± 1.1 wt % H2O), and their H isotopic compositions are also different (fresh interior δD = –103.6 ± 0.7‰, altered surface δD = –136.3 ± 3.9‰; fig. S4 and table S2). Similar features have been observed in martian meteorites subjected to desert alteration, with the H content and isotopic characteristics being modified only near the surface (i.e., <2 mm) (16). In addition, the hydrogen contents and D/H ratios measured for the two meteorite finds (Sahara 97096 and Yamato 791790) are very close to the ones measured for the Indarch chondrite that was collected after its observed fall and are consistent with a metamorphic evolution of the H signatures in ECs (table S2 and fig. S5B). These observations suggest that the H contents of ECs were established during the early Solar System and have experienced minimal modifications on Earth.

To confirm the indigenous nature of H in the interiors of ECs, we used SIMS to analyze the H contents and isotopic compositions of the glass fraction (mesostasis) in four chondrules—submillimeter silicate spherules that correspond to the main mineral assemblages in ECs—of Sahara 97096, the least-metamorphosed EC in our sample (11, 12). For comparison, we performed similar measurements in two CV chondrites (Kaba and Vigarano) that have the same degree of metamorphism (17). The glassy mesostasis of Sahara 97096 contains between 2700 and 12,300 wt parts per million (ppm) H2O equivalent, with an average value of 7560 ± 1546 wt ppm H2O (two standard deviations, 2σ) for 16 measurements (Fig. 1C and table S4). Although their metamorphic grades are similar to that of Sahara 97096, the CV chondrites Vigarano and Kaba have lower H contents of 330 ± 140 and 210 ± 120 wt ppm H2O, respectively (Fig. 1C and table S4). The H isotopic composition of Sahara 97096 mesostasis is homogeneous, with an average δD value of –147 ± 16‰ (2σ; Fig. 1D and table S4), whereas Vigarano and Kaba have δD = –261 ± 25 and +17 ± 95‰, respectively (2σ; Fig. 1D and table S4). Among these three meteorites, only Kaba underwent aqueous alteration of its chondrules (18), which accounts for the variable D/H ratio of its mesostasis. Vigarano contains traces of incipient aqueous alteration restricted to its fine-grained matrix (19). Because Sahara 97096 does not show any evidence of aqueous alteration, the abundant H contents of its chondrule mesostases must have been acquired before the EC parent asteroid evolution, possibly during the period of chondrule formation.

Considering the H concentration of Sahara 97096 mesostasis, the modal abundance of mesostasis in Sahara 97096 chondrules [16.2 vol % (15)], and the modal abundance of chondrules in ECs [~70 vol % (20)], H in chondrule mesostases accounts for 702 ± 117 wt ppm H2O of the bulk rock, or ~13% of the bulk-rock H content. Another potential source of H in ECs is insoluble organic matter (IOM). However, given the low abundance of IOM in ECs (0.6 wt %) and the H concentration in EC organics [0.7 wt % (12)], the contribution of IOM to the bulk-rock H content is only 380 ppm H2O, or 7.7% of the total H content of the chondrite. Combined, H hosted in chondrule mesostases and IOM can account for only ~20% of the total H measured in the Sahara 97096 bulk rock.

H concentrations have been reported in low-calcium pyroxenes from non-CC materials from (i) the S-type asteroid Itokawa (700 to 1000 wt ppm H2O) brought back by the Hayabusa sample-return spacecraft and (ii) the metamorphosed ordinary chondrite (OC) Larkman Nunatak 12036 (600 to 1300 wt ppm H2O) (21). We found an even higher H concentration (5300 wt ppm H2O; table S2) for the pyroxene fraction of the Norton County aubrite. The modal abundance of low-calcium pyroxene in ECs is ~50 vol % (15), so enstatite pyroxenes could account for ~15 or ~58% of the bulk H content of ECs, considering the highest H content of OC pyroxenes (1300 wt ppm H2O) or the aubrite pyroxene fraction, respectively. The remaining ~65 to 20% of H measured in the bulk rock may be contributed by other unknown carriers such as sulfur-rich carbon-bearing porous amorphous silica (22), owing to an underestimation of the H content in EC pyroxenes, and/or could be derived from some pervasive terrestrial contamination, although a limited amount of contamination is expected (otherwise, it would have erased the thermal metamorphism effect visible on the diagram of δD versus 1/H; fig. S5). Our SIMS measurements demonstrate that at least 20% of the H in ECs has an identified carrier phase (mesostasis or IOM) and that most of the remaining 80% is probably of indigenous origin. This implies that terrestrial hydrogen could be derived from ECs without invoking additional sources.

We incorporated the H concentrations measured in ECs (tables S2 and S4) into theoretical models of Earth’s formation that invoke mixing of chondrite-like materials. These models use various proportions of EC-like, OC-like, and CC-like materials (13, 5, 23), so we explored three end-member compositions: (i) 100% EC-like materials (1); (ii) 68% EC-like and 32% CC-like materials (2, 23), although these proportions could be inconsistent for the isotopic composition of Earth’s mantle (11); and (iii) ~70% EC-like, ~25% OC-like, and ~5% CC-like materials (3) (Fig. 2). In all cases, EC-like materials contribute substantially to Earth’s water budget, supplying 3.4 to 23.1 times the mass of Earth’s oceans (1.4 × 1021 kg) once our H content measurements are included (table S2). These estimates are upper limits because loss of volatiles might have occurred during Earth’s accretion (24, 25). Considering only identified H-bearing phases (mesostasis and IOM), EC-like materials alone contribute ~3 to 4.5 times the mass of Earth’s oceans (Fig. 2). If Sahara 97096 is representative of the most pristine ECs (Fig. 2), its bulk-rock hydrogen content (4700 wt ppm H2O or ≥14 ocean masses) is sufficient to explain the highest estimates of the water content of Earth’s surface and mantle, which are 1000 to 3900 ppm H2O (5, 26). Mesostasis and IOM H contents alone can account for three to five times the total mass of the oceans, corresponding to intermediate estimates of the water content of Earth’s mantle.

Fig. 2 Contributions of accreting materials to Earth’s hydrogen budget in three different models.

Values are expressed in equivalent ocean masses, where one ocean = 1.4 × 1021 kg H2O. (A) A model Earth comprising only ECs (1). (B) A model Earth comprising 68% ECs and 32% CCs (2). (C) A model Earth comprising 71% ECs, 24% OCs, and 5% CCs [the most extreme case from (3)]. The consistency of such models for the isotopic composition of the mantle is discussed in (11). SAH97096, Sahara 97096 (EH3); Bulk, bulk rock; OM, organic matter; Sem., Semarkona (LL3.0). CCs are divided into water-rich (CI-type) and water-poor (CV- and CO-type) groups. Water contributions from ECs are shown in purple, from OCs in red, and from CCs in blue. The water content of Earth’s mantle (green band) is between 1 and ≥10 ocean masses (26). We consider 10 ocean masses in the mantle as a maximum value. Regardless of the model, the water provided by ECs alone lies above the maximum value of 10 ocean masses and thus substantially contributes to Earth’s water budget. Data are from (5, 7, 11, 21) and this study (tables S2 and S4). Error bars are 2σ.

The hydrogen isotopic composition of ECs is depleted in deuterium relative to that of Earth’s oceans but is still within the range of mantle values (Fig. 3). ECs are not the only group of meteorites with D-poor values consistent with those of Earth’s mantle, as CM-type CCs also span the range of mantle H isotopic compositions (Fig. 3). Nitrogen isotopes provide additional constraints (27), as the mantle is depleted in both D and 15N relative to the surface reservoirs (atmosphere and oceans). Combining H and N isotopic compositions allows us to identify potential chondritic precursors (Fig. 4). Only ECs are compatible with both the δD and δ15N values measured in mantle-derived rocks (Fig. 4). D-poor solar hydrogen [δD = –865‰ (6)] could also have contributed to the mantle composition (28), although it would require extreme enrichments of solar H relative to solar Ne because mantle neon is solar-like. The chondritic H/Ne ratio [4 × 108 (29)] is five orders of magnitude higher than the solar ratio [7.9 × 103 (30)]. Therefore, we favor the incorporation of EC-like material in the silicate Earth (1).

Fig. 3 Hydrogen isotopic compositions of meteorites compared with that of Earth’s mantle and oceans.

Data are shown for ECs (bulk rock and mesostasis) (purple); OC pyroxenes and hydrous minerals (red); and CCs (bulk rock and CI, CM, and CV types) (blue). Mes, mesostasis; Pyrox, pyroxene. Meteorite data are from (7, 12, 21, 31) and this study (tables S2 and S4). Values for Earth’s mantle (δD = –220 to –20‰) (green shaded area) are from (26), with the most D-depleted values reported in (28). Primitive mantle δD values range from–130 to –75‰ (32, 33). Error bars are 2σ.

Fig. 4 Hydrogen and nitrogen isotopic compositions of meteorites compared with that of Earth’s mantle and surface.

Meteorite bulk-rock values are shown for ECs and CM-, CI-, CV-, and CO-type CCs; data are from table S2 and (7, 34, 35). Mantle δD values are from (26, 28). Primitive mantle δD values are from (32) and (33) and δ15N values are from the most negative values (–40‰) in (35). atm., atmosphere.

Earth’s surficial (oceans and atmosphere) H and N isotopic compositions (δD ≡ 0‰ and δ15N ≡ 0‰) cannot be easily derived from EC materials alone because they are richer in D and 15N by ~100 to 200‰ and 5 to 40‰, respectively. Atmospheric escape to space, often invoked for atmospheric species including water vapor, could potentially achieve such enrichments, but the compositions of light noble gases do not support this hypothesis. Ne-Ar isotopic variations in the mantle-atmosphere system are not consistent with the elemental and isotopic variations expected for atmospheric escape; rather, they are best reproduced by mixing between a solar-like component trapped in the solid Earth and a CC-like component with a low 20Ne/22Ne ratio (5). A contribution of CC-like materials, especially CIs, to the surface inventory would also be consistent with the observed H and N isotopic variations (Fig. 4) and can be computed independently from H and N isotopic ratios (11). We find a ~4% contribution from CI-like material to Earth’s surficial reservoir based on hydrogen abundances and isotopic compositions and a ~15% contribution based on equivalent data for nitrogen (11). Although nitrogen isotopes could also have contributions from cometary materials (11), these estimates are similar to those obtained from noble gases (5) and molybdenum isotopes (23). We cannot determine exactly when the CI-like material was delivered to the surficial reservoir, but it must have been sufficiently late during Earth’s formation to prevent a global rehomogenization of the mantle with the surface.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Tables S1 to S6

References (3663)

Data S1

References and Notes

  1. Materials and methods are available as supplementary materials.
Acknowledgments: We thank the Field Museum (Chicago, USA), the French National Museum of Natural History (Paris, France), the Japanese National Institute for Polar Research (Tokyo, Japan), the University of New Mexico (Albuquerque, USA), Natural History Museum (Vienna, Austria), and the CEREGE meteoritic collection (Aix en Provence, France) for providing the meteorite samples. We thank N. Bouden and the members of the Ion Probe Team Nancy (IPTN) for help with the SIMS. We thank E. Deloule, A. Gurenko, B. Luais, L. France, M. Broadley, D. Bekaert, and C. Cartier for fruitful discussions. This is CRPG contribution 2997. Funding: This work was supported by the French Research National Agency (grant ANR-19-CE31-0027-01 to L.P.) and the European Research Council (grant 695618 to B.M.). Author contributions: L.P. designed the study. L.P., T.R., L.G.V., and D.T. performed the bulk and in situ analyses. Y.M. and B.M. contributed to the experimental design. L.P., Y.M., and B.M. discussed the data and wrote the manuscript. All authors provided input to the data analysis and manuscript preparation. Competing interests: We declare no competing interests. Data and materials availability: All meteorites that we studied are deposited in public museums, as listed in the supplementary materials. Our measured H abundances and D/H ratios, for the meteorite samples and standards, are listed in tables S1 to S4 and data S1.

Stay Connected to Science


Navigate This Article