Isotope composition and volume ofEarth’s early oceansEmily C. Popea,b,c,1, Dennis K. Birda,c, and Minik T. Rosinga,b,c

aDepartment of Geological and Environmental Sciences, Stanford University, Stanford, California 94305; bNatural History Museum of Denmark,University of Copenhagen, Øster Voldgade 5-7, 1350 København K, Denmark; and cNordic Center for Earth Evolution, Øster Voldgade 5-7,1350 København K, Denmark

Edited by Robert N. Clayton, University of Chicago, Chicago, IL, and approved January 19, 2012 (received for review September 23, 2011)

Oxygen and hydrogen isotope compositions of Earth’s seawaterare controlled by volatile fluxes among mantle, lithospheric (ocea-nic and continental crust), and atmospheric reservoirs. Throughoutgeologic time the oxygenmass budget was likely conservedwithinthese Earth system reservoirs, but hydrogen’s was not, as it canescape to space. Isotopic properties of serpentine from the approxi-mately 3.8 Ga Isua Supracrustal Belt in West Greenland are used tocharacterize hydrogen and oxygen isotope compositions of ancientseawater. Archaean oceans were depleted in deuterium [expressedas δD relative to Vienna standard mean ocean water (VSMOW)]by at most 25� 5‰, but oxygen isotope ratios were comparableto modern oceans. Mass balance of the global hydrogen budgetconstrains the contribution of continental growth and planetaryhydrogen loss to the secular evolution of hydrogen isotope ratiosin Earth’s oceans. Our calculations predict that the oceans of earlyEarth were up to 26%more voluminous, and atmospheric CH4 andCO2 concentrations determined from limits on hydrogen escape tospace are consistent with clement conditions on Archaean Earth.

Archaean ∣ faint early sun ∣ hydrogen escape ∣ hydrosphere ∣ serpentine

The geologic record provides scant definitive evidence of earlyArchaean seawater chemistry. Oxygen and hydrogen isotope

compositions of oceanic sedimentary rocks provide a commonlyused proxy record, but interpretations remain inconclusive (1–4).Average δ18O values of Archaean marine sediments (cherts andcarbonates) are tens of per mil lower than present-day values, andsystematically increase over geologic time (2, 3). This trend hasbeen attributed either to lower δ18OSEA WATER in the Archaean(1, 3), or to ocean temperatures up to approximately 70 °C (5–8).Both interpretations have been questioned based on the suscept-ibility of chemical sediments to diagenetic alteration over geolo-gic time (2, 4), and for the latter, a lack of geologic evidence forthe extreme greenhouse gas concentrations required to sustainsuch high temperatures under a less luminous young Sun (9).

Minerals formed by metasomatic reactions between seawaterand oceanic lithosphere provide an alternative proxy for earlyocean chemistry (10–14). Here we report hydrogen and oxygenisotope compositions of 114 silicate mineral separates from theca. 3.8 Ga Isua Supracrustal Belt (ISB) ofWest Greenland, includ-ing metamorphosed basalts, gabbros, and ultramafic rocks thatrepresent fragments of Eoarchaean oceanic crust (see Fig. S1,Table S2). Minerals formed during heterogeneous late-stageCO2- or meteoric-dominated metamorphic overprinting are distin-guished by characteristic trends in their coupled δD and δ18Ovalues (Fig. S2). In contrast, serpentine minerals formed by reac-tion of seawater with ultramafic rocks (peridotites) preserved in alow-strain enclave of the ISB (15) define a different isotopic trend.Here we use these serpentines to constrain D∕H and 18O∕16Oratios of the Eoarchaean ocean.

Serpentine Isotope GeochemistrySerpentine-group minerals are hydrous magnesian silicates com-monly formed by hydrothermal infiltration of seawater intoultramafic oceanic crust, hydration of the mantle wedge by slab-

derived fluids, or fluid-rock interaction in continentally emplacedultramafic lithologies (16–18). Serpentinized oceanic peridotitesare dominantly composed of low-temperature (<300 °C) poly-morphs lizardite and chrysotile, and to a lesser extent antigorite,which forms at temperatures >250 °C (17). In modern oceanicserpentinites, antigorites have δD between −46 and −30‰(Figs. 1 and 2A, filled squares), values that indicate equilibriumwith modern-day seawater at temperatures between 150 and300 °C (19). In contrast, chrysotiles and lizardites have δD pre-dominantly <50‰ (Figs. 1 and 2A, open squares). They likelyformed by recrystallization in the presence of low-δD fluidsoriginating from dehydration reactions in hydrothermally alteredoceanic crust (20, 21). Due to its more refractory nature (12, 22,23), antigorite retains its primary δD value, even when lizarditeand chrysotile reequilibrate (21) (Fig. 1).

Oxygen isotopes in oceanic serpentines range from þ0.8 toþ12.4‰, with the majority between þ3 and þ7‰ (Fig. 2B).Isotopic variation is due to the temperature dependence of ser-pentine-water δ18O fractionation (19) and the exchange of oxygenisotopes between the fluids and basaltic ocean crust (24). This canbe seen in the δ18O of fluids venting at mid-ocean ridge hydro-thermal systems, which are typically enriched in 18O relative toseawater by 2 to 3‰ (24–26).

Serpentinites in ophiolites range in age from Tertiary to theEoarchaean, and show a wider range in δD than their present-day oceanic counterparts (Fig. 3), often due to postemplacementisotope exchange with crustal fluids, including meteoric waters.The highest δD values, particularly those of antigorite, arecommonly considered to have preserved their original isotopiccompositions established during serpentinization by seawaterin oceanic environments (10, 12, 27), and have been used to makeinferences about the extent to which δDSEA WATER changes overgeologic time (11).

Serpentines in the Isua Supracrustal BeltThe Isua Supracrustal Belt represents the best preserved oceaniccrust of Eoarchaean age (15, 28, 29). It has undergone regionalgreenschist- to amphibolite-facies metamorphism (SI Text), butboth geochemical (15) and structural preservation (14) of proto-lith material occurs within localized low-strain exposures. Forexample, original bedding is visible in low-δ13C carbon-rich mar-ine sediments (30), and pillow lavas and sheeted dike structuresassociated with metagabbroic and serpentinized ultramafic rocksindicate that the ISB contains preserved fragments of oceaniccrust (14, 29).

Author contributions: E.C.P., D.K.B., and M.T.R. designed research; E.C.P. performedresearch; E.C.P., D.K.B., andM.T.R. analyzed data; E.C.P., D.K.B., andM.T.R. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: emily@snm.ku.dk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1115705109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1115705109 PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ 4371–4376

EART

H,AT

MOSP

HER

IC,

AND

PLANETARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 2

3, 2

021

Hydrogen and oxygen isotopes of the analyzed minerals distin-guish two metasomatic events that have selectively altered ISBlithologies (Fig. S2). First, CO2-rich fluids likely associated withca. 3.81 − 3.74 Ga granitoid intrusions (14, 15, 28) intenselymetasomatized ultramafic lithologies and adjacent amphibolitesand felsic schists along the western and southern portions of theISB. A second, later-stage meteoric-overprinting in the ISB isrestricted to highly deformed regions and along contacts withlate-stage intrusions (14). Isotopically, the first metasomaticevent is distinguished by a large increase in δ18O (up to þ18‰)with an insignificant change in δD, and the second resulted in aprogressive decrease in δD (as low as −129‰) with very littlechange in δ18O. The isotopic trends associated with both eventsapparently originate at a common protolith isotopic composi-tion (Fig. S2).

Serpentinites are preserved in regions of the ISB not affectedby the CO2-metasomatic event (Fig. S1) (28). Antigorite, lizar-dite, and chrysotile occur with varying modes, and have a cumu-lative range in δD of −99 to −53‰ and δ18O of þ0.1 to þ5.6‰(Table 1; Fig. 1, purple circles). Serpentines with the highest δD(samples from low-strain outcrops) exhibit little variation in δ18O,but samples with δD≲ −80‰ (highly deformed samples or thosecomposed predominantly of chrysotile/lizardite) have δ18O thatdecrease with decreasing δD. This hockey-stick trend in valuesof δD and δ18O of hydrothermal minerals is characteristic of pro-gressive meteoric-hydrothermal alteration (31), where decreasingδD and δ18O values of crustal rocks reflects increasing water-rock(W∕R) ratios. Hydrogen isotopes in the rock are more significantlyaffected by secondary water-rock reaction than oxygen isotopesbecause of the greater oxygen content in rock-forming minerals.Thus at low W∕R mass ratios, δDROCK will reflect isotope ex-change with meteoric fluids but δ18OROCK will not, whereas at ele-vated W∕R ratios, both the δ18OROCK and δDROCK will change.

Hypothetical water-rock reaction paths between the most deu-terium-rich serpentine (δD ¼ −53‰, δ18O ¼ þ4.8 to þ5.6‰)and meteoric water (δD ¼ −90‰, δ18O ¼ −12.5‰) are shownin Fig. 4. δDMETEORIC-WATER was estimated based on approxi-mately 40‰ fractionation (an approximate ΔDMINERAL-H2Ofor Fe-Mg minerals at 100–300 °C; SI Text) between meteoricwater and ISB mineral samples from regions associated withlate-stage intrusions or mylonitization. The lowest δD value ofall minerals analyzed in the ISB is −129‰ (Table S2).δ18OMETEORIC−WATER was calculated using the equation for themeteoric water line (32). With the exception of two outliers(16-05, 07-07, Table 1), the hydrogen and oxygen isotope proper-ties of ISB serpentines are well approximated by a W∕R reactionpath where temperatures are approximately 100–200 °C (Fig. 4).For samples with δD > −85‰, W∕R ratios were likely ≲0.1 byweight, regardless of the temperature of isotope exchange. Inview of these observations, we interpret serpentines withδD > −85‰ as preserving oxygen isotope values close to theiroriginal composition.

Fig. 1. Hydrogen and oxygen isotope values of Isua serpentines. ISBserpentines ¼ purple circles (Table 1). Shown relative to: modern ocean ser-pentinites [gray region; filled squares—antigorite, open squares—lizarditeand chrysotile, hatched squares—undifferentiated serpentine; from HessDeep [East Pacific Ridge] (52), the Marianas Trench (23, 53), the Puerto RicanTrench, the Mid-Atlantic Ridge and the Blanco Fracture Zone (16)], themeteoric water line (MWL) and modern seawater composition (VSMOW),and the isotope composition of serpentine that would be in equilibrium with(1) modern-day seawater (black line), (2) Archaean seawater from this study(δD ¼ −25� 5‰; δ18O ¼ 2.3‰; purple region), and (3) Archaean seawater asproposed by Hren et al. (3) (δD ¼ −60‰; δ18O ¼ −10‰; gray line). Based onthe temperature-dependent fractionation of Saccocia et al. (19). Errors forIsua serpentines are 1σ (SD) or analytical error, whichever is larger (see Table 1,Table S1 and Analytical Methods; N ¼ 2–4 for δD, N ¼ 1–2 for δ18O).

Fig. 2. Isotopic features of modern ocean serpentine. A. Hydrogen isotopecomposition of serpentine from various discrete locations (16, 21, 23, 52, 53).B. Histogram of δ18O values of modern ocean serpentinites (blue) (16, 21, 23,52–56). Also shown is the distribution of δ18O of ISB serpentinites (purple).

4372 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1115705109 Pope et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

3, 2

021

The highest δDSERPENTINE value recorded in the ISB provides aminimum estimate of their primary hydrogen isotope composition.However, it is possible that the serpentines with δD ≈ −53‰do not represent the initial rock composition as we model inFig. 4, but have also been altered by meteoric water interaction.To test this, reaction paths between the same meteoric fluid and ahypothetical starting rock with δD ¼ −30‰ (uppermost δD ofmodern ocean antigorites) were also calculated. These paths areshown in Fig. S4A, and are less consistent with the Isua serpentinedata than those in Fig. 4, but cannot be unequivocally ruled out.

We conclude that samples with the highest δD (−56 to−53‰) andδ18O betweenþ4.1 andþ4.9‰ (Fig. 4), have undergone minimallate-stage metasomatism (W∕R≲ 0.01), and retain an isotope sig-nature of seawater serpentinization of Eoarchaean oceanic crust.

Critical to our interpretation are the paragenetic, petrofabric,and geochemical features of the most deuterium-rich ISB ser-pentine samples. These samples exhibit primary pseudomorphtextures after peridotitic olivine (Fig. S5), and contain modallyabundant antigorite that has a tendency to seal pathways for sec-ondary fluids and inhibit hydrogen isotope exchange (SI Text) (12,22, 23). We note that the four samples with δD ≥ −56‰ are allfrom a region in the ISB that has been structurally and geochemi-cally characterized as a well-preserved fragment of oceanic crust(Figs. S1, S3) (14, 29). Pons et al. (33) compared the sourceregion for these serpentinized ultramafics to serpentinite mudvolcanoes of the Mariana forearc on the basis of zinc isotopes.The presence of magnetite in the ISB serpentinites also indicatesformation by low-temperature (50–300 °C) serpentinization onthe ocean floor, as opposed to formation during high-gradeserpentinization of the mantle wedge during crustal accretion(18, 34) (SI Text).

Over the range of temperatures typical for antigorite forma-tion in oceanic settings (250–300 °C, SI Text), samples 15-05,03-10A, 03-10B, and 03-10C (Table 1) would form in equilibriumwith fluids having a δD between −28 and −21‰ (−25� 5‰,incorporating analytical error) and δ18O between þ0.8 and3.8‰ (Fig. 1, blue cross) (19). We consider these values to re-present a minimum (the most negative possible) estimate forδD of Eoarchaean seawater. Additionally, we note our modeledoxygen isotope ratios for early Archaean oceans (Fig. 1) are notsignificantly different from those of modern-day seawater(δ18OVSMOW ¼ 0‰). As in modern hydrothermal alteration ofocean crust, ISB serpentine-forming fluids may be up to approxi-mately 2.8‰ higher than starting seawater δ18O, because ofisotopic exchange with oceanic crust (25). From Fig. 2B, it is clearthat the δ18O distribution in ISB serpentinites (purple values) isconsistent with that of modern ocean serpentinites (blue values),particularly for samples where δD > −85‰.

Models for a Global Hydrogen BudgetGeologic processes that control the secular increase inδDSEA WATER from −25� 5‰ in the Eoarchaean to 0‰ in mod-ern oceans can be evaluated through mass balance constraints on

Fig. 3. δD of serpentine in modern ocean crust and in ophiolites as afunction of time. Antigorite ¼ solid squares, lizardite and chrysotile ¼open squares, whole rock data ¼ crossed squares. Oceanic serpentines arefrom the same references as Fig. 1. Older serpentines are from ophioliteor greenstone belt sequences (10–12, 16, 27, 57–59, this study). Serpen-tine-forming fluids (δD ¼ −25� 5‰) at 3.8 Ga are the minimum estimateof Archaean seawater, based on equilibrium fractionation of unaltered ISBantigorite serpentine. Blue line represents the approximate increase in sea-water δD with time (calculated from average of values in Table 2) resultingfrom gradual, steady-state continental growth, and hydrogen escape beforethe rise of oxygen. The different slopes before and after approximately2.4 Ga represent the elimination of appreciable hydrogen escape after therise of oxygen (Fig. S7). We note that seawater evolution as representedby this line is approximately parallel to maximum δD of serpentine over geo-logic time (purple line). Phan ¼ Phanerozoic.

Table 1. Isotope analyses of Isua serpentinites

Sample Mineral* δD 1σ† δ18O 1σ

03-10B A −53 ± 0.5 4.6 ± 0.2803-10C A+L −53 0.7 4.1 -‡

15-05 A −56 1.9 4.9 0.0703-10A L −56 1.3 5.3 1.1301-10 L −61 1.8 4.1 -28-05 U −69 2.2 5.1 0.42940094 A+L −69 2.2 3.5 -07-08A U −73 1.1 4.8 0.35940095 U −74 0.2 4.5 0.3509-08 A+L −78 1.3 5.2 0.4220-07A L −84 4.0 4.1 -16-05 A+L −85 2.1 5.6 -08-08 L+C −91 1.5 2.9 -810000 A −95 1.9 4.4 0.7807-07 U −98 1.2 0.1 0.4206-07AA A+C −99 0.2 2.7 0.42

Relative to VSMOW.*Primary phase: A ¼ Antigorite, L ¼ Lizardite, C ¼ Chrysotile, U ¼ allserpentine polymorphs.

†1σ ¼ 1SD for N ¼ 2–4 samples. Does not include analytical error.‡1σ not applicable; N ¼ 1. Further details of analyses in Table S1.

Fig. 4. ISB serpentine data shown relative to hypothetical reaction pathsbetween serpentines and meteoric water. Starting composition of rock isδD ¼ −53‰ and δ18O ¼ þ4.1 to þ5.3‰ (highest δD and δ18O values of ISBserpentines) and meteoric water is δD ¼ −90‰ and δ18O ¼ −12.5‰. Reactionpaths shown for varying temperature conditions, with water-rock ratios(W∕R) denoted. Errors for Isua serpentines are 1σ (SD) or analytical error,whichever is larger (see Table 1, Table S1 and Analytical Methods; N ¼ 2–4for δD, N ¼ 1–2 for δ18O).

Pope et al. PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ 4373

EART

H,AT

MOSP

HER

IC,

AND

PLANETARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 2

3, 2

021

fluxes between global reservoirs and hydrogen loss to space (35).Global reservoirs include water contained in the modern ocean,trapped in hydrous minerals of the continental crust and sedi-ment, in surficial and glacial waters, in the biosphere, in thehydrated ocean crust and in the mantle (Table S3). Relative tothe other hydrogen reservoirs, the volume of water trapped inthe biosphere is negligibly small. We assume that the extent ofhydration of ocean crust by reaction with seawater has likely re-mained approximately constant over geologic time. In addition,we tentatively consider a net zero flux of water into and out ofthe mantle, as modern ingassing fluxes in subduction zonesand outgassing fluxes at mid-ocean ridges are comparable (seeTable S4, SI Text).

Three model scenarios that consider fluxes among the remain-ing hydrogen reservoirs are presented in Fig. 5 as a function oftotal ocean mass and δDSEA WATER, and are compared with valuesinferred from ISB serpentinites. In scenarios A and B, Earthbehaves as a closed system for hydrogen. In A, all the hydrogencurrently sequestered in the cryosphere, biosphere, surface- andground-waters, sediments, and continental crust is allocated tothe ocean reservoir, characterizing seawater before continentalgrowth sequestered deuterium-poor fluids from the oceans intohydrous minerals (SI Text). In scenario B, 100% of modern con-tinental crust is present, but it assumes nonglacial conditions. Inthese two scenarios, the ocean would have been 4 to 20% largerthan modern oceans, and lower in deuterium by 9–18‰. Modelscenario C is an estimate of the effect of hydrogen escape to spacevia methane photolysis in the preoxygenic Archaean atmosphere(36), given ΔDCH4-H2O ¼ −150 to −300‰ (37), and diffusion-limited hydrogen escape, where no D∕H fractionation occurs(36, 38) (SI Text). The amount of hydrogen escape required tooxidize Earth’s crust indicates that preescape oceans would havebeen approximately 6% larger, and have a δD 10 to 20‰ lowerthan Vienna standard mean ocean water (VSMOW).

Here we consider the effects of both hydrogen sequestrationin terrestrial reservoirs and hydrogen escape to space to account

for the secular increase of δDSEA WATER from −25� 5‰ atca. 3.8 Ga (see SI Text for model parameters, calculations, anduncertainties). The calculated amount of hydrogen that must belost via escape to accommodate the estimated maximum andminimum δDEOARCHAEAN OCEAN values based on ISB serpenti-nites (−30 and −20‰) are compiled in Table 2, assuming differ-ent amounts of continental crust at approximately 3.8 Ga. Ifsignificant continental mass existed at approximately 3.8 Ga,the isotopic shift of seawater since that time would be dominantlycontrolled by hydrogen escape. If no continental crust existed atapproximately 3.8 Ga, less hydrogen would have to escape tospace to result in the observed change in δD, as some low-δDfluids would be incorporated into gradually growing continentalcrust. For our modeled range of δDSEA WATER, ocean volumeswould have been 9 to 26% greater than present-day, requiring amass loss of H to space of between approximately 2 × 1021 moland 18 × 1021 mol (Table 2; Fig. 5). Such a difference in massmay have a significant impact on current hypotheses of earlyEarth processes, including shallow mid-ocean ridge-crests andassociated models for δ18O buffering of the Archaean ocean(1), continental freeboard in early Earth, and rates of heat andvolatile transport via ocean spreading ridges (39).

We consider the values determined from our hydrogen massbalance model to be reasonable first order approximations basedupon average masses and isotope compositions of major hydro-gen reservoirs on Earth (SI Text), which should be supported byadditional evidence from the geologic record. Namely, changes inthe isotopic composition of seawater with time will be apparent inother well-preserved rock suites. In Fig. 3, an extrapolation ofδDSEA WATER from present day to approximately 3.8 Ga basedon the average values from Table 2 is shown as a blue line (SIText). If Archaean δDSEA WATER were lower; e.g., approximately−60‰ as is proposed by an alternative study based on the 3.4 GaBuck Reef Cherts in South Africa (3), the slope of this line wouldbe much steeper; if δDSEA WATER had remained constant withtime, the line would be horizontal. The average isotope composi-tion of serpentine that would be in equilibrium with our modeledδDSEA WATER is denoted by the parallel purple line, which is clo-sely correlated to the maximum measured (and thus the best-preserved) δD values of serpentinite over geologic time.

Although an ambitious study, the Buck Reef Cherts analyzedby Hren et al. (3) (also shown in Fig. 5 as scenario D) are a lesswell-suited proxy for seawater δD than serpentine, resulting in thediscrepancy between our two datasets. δD of chert is measuredfrom small concentrations of water (typically less than 1.4 wt%compared to up to 14 wt% in serpentine) that are susceptibleto contamination with organic matter, clay minerals, or secondaryfluids (40), and the temperature dependence of H∕D fractiona-tion between chert and water is not straightforward (3, 41).

Another important consideration is the relationship betweenthe δD and mass of Archaean oceans and the extent to whichmethane photolysis controlled oxidation of Earth’s surface (36).δDSEA WATER of −60‰ (3) would result in early oceans that were25 to 67% larger than their present size (Fig. 5), thus releasing4 to 10 times as much oxygen by H escape from photolysis of bio-genically produced methane than in Catling et al. (36)’s calcula-tion for oxidation of the crust, summarized by the reaction:

CO2 þ 2H2O → CH4 þ 2O2 → CO2 þO2 þ 4Hð↑ spaceÞ: [1]

An additional oxygen sink, such as oxidation of the mantle viasubduction, would be necessary to accommodate the oxidation as-sociated with such an extreme hydrogen flux to space over Earthhistory. In contrast, the amount of oxygen released due to hydro-gen escape from our models in Table 2 is 2.5� 1.9 × 1021 mol O2,a range consistent with the estimated 2.0 to 2.9 × 1021 mol O2-equivalent necessary to oxidize the Earth’s crust before the riseof oxygen at ca. 2.4 Ga (36).

Fig. 5. Hydrogen isotope composition and mass of ocean in the earlyArchaean. The terrestrial hydrogen budget [A, B; e.g., Lécuyer et al. (35)] andCatling et al. (36) data (C) represents the required δD and mass of the oceanto satisfy the parameters of their models. The terrestrial hydrogen budgetmodel represents a primordial ocean in which either (A) continental crust isnot yet present (δD ¼ −18‰), or (B) where 100% of modern continental massexists (δD ¼ −9‰). Model (C) estimates that oxidization of all modern conti-nents, oceans and the atmosphere due to H-escape caused 6%mass loss fromthe ocean and enriched the ocean in deuterium by approximately 10 to 20‰(36). Point (D) shows the mass of the ocean required to satisfy the ArchaeanδDSEAWATER values predicted by chert analyses from Hren et al. (3), givenhydrogen loss via escape to space from photolysis ofmethane (range in oceanmass results from uncertainty in methane-water fractionation). These dataare shown relative to the minimum δD values of the Archaean ocean deter-mined from serpentine analyses (this study, Table 2), and the ocean mass as afunction of the volume of continents present at 3.8 Ga (0 to 70%).

4374 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1115705109 Pope et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

3, 2

021

Finally, our model indicates that hydrogen escape is limited toat most 18 × 1021 mol H, releasing approximately 4.5 × 1021 molof free O2 to the Earth system (Table 2). If this much oxygen wasreleased over approximately 1.4 billion years (3.8 to 2.4 Ga), thenatmospheric CH4 at 3.8 Ga must have been at most 480 ppmv(64 to 480 ppmv for the range of O2 released in Table 2; SI Text)(36). Haqq-Misra et al., (42) has shown that at CH4∶CO2 ratiosabove 1.0 or 0.2 (depending on atmospheric model), a climati-cally cooling organic haze would form, lowering surface tempera-tures. For an atmospheric mixing ratio of CH4 in the Archaean ofapproximately 64 to 480 ppmv, the pCO2 needed to maintain ahaze-free atmosphere would be between 10−4.2 and 10−2.6 bar, orat a minimum approximately 0.2 to 6 times present atmosphericlevels. These values are in the pCH4 and pCO2 ranges suggestedby Precambrian paleosols, methanogenic metabolic constraints,and magnetite/siderite stability in early Archaean banded ironformations (9, 43, 44) (Fig. S7).

δ18O of Eoarchaean SeawaterThe δ18OSEA WATER of Archaean oceans we calculate from ISBserpentinites (þ0.8 to 3.8‰; Fig. 1) is significantly higher thanestimates based on carbonate and chert analyses that assume atemperate Archaean climate, which are as low as −13.3‰ (2, 3).Our values are, however, consistent with observations made fromother Archaean volcanic rocks (13), including those from nearbypillow basalts in the ISB (14), from biogenic phosphates preservedin the 3.5 to 3.2 Ga Barberton Greenstone Belt (4), and from theophiolite record of the past approximately 3.5 Ga (10).

The progressive increase in δ18O of chemical sediments overgeologic time to their modern average value of approximately 0‰[PDB] we attribute to postdepositional exchange with hydrother-mal or pore water fluids on the seafloor in accord with conclusionsdrawn from the Barberton phosphate data (4). In modern oceans,sediments are on average approximately 500–1;000 m thick (45),and the near-surface geothermal gradient is about 25 °C∕km. If theEarth’s heat flow in the early Archaean was approximately threetimes greater than at present (46, 47), then a thick package ofseafloor sediments on the Archaean ocean floor could isotopicallyexchange with seawater-derived pore waters at temperatures ap-proaching 70 °C, as on average, the temperature gradient of theEarth’s crust will scale linearly with heat flow. A decline in Earth’sheat flow with geologic time could explain the progressive increase inδ18O of cherts and carbonates, although we note that such a decreasein heat flow over geologic time has been disputed (48, 49).

ConclusionsPetrogenetic and geochemical evidence preserved in ISB serpen-tinites constrains a minimum δDSEA WATER of early Archaeanoceans to −25� 5‰ relative to present-day values (VSMOW).In the context of the metasomatic history of the ISB, the struc-tural setting of the samples considered, the secular trend inhydrogen isotope compositions of serpentinites over geologictime, and the corroboration with independent models of surfaceoxidation and atmospheric methane levels in the Eoarchaean,these data appear to genuinely reflect seawater-oceanic crust

interaction and are the best constraint thus far for δD of earlyArchaean seawater. Mass balance considerations demonstratethat low-deuterium waters were primarily sequestered in conti-nental crust during its progressive growth since the Hadean,and later also into groundwater and glacial reservoirs. At most,hydrogen lost to space is limited to 18 × 1021 mol H, releasing≲4.5 × 1021 mol of free O2 to the Earth system, and constrainingatmospheric methane concentrations at 3.8 Ga to ≲480 ppmv.This supports the argument that the combined greenhouse effectof atmospheric CO2 and CH4 cannot independently reconcilethe faint early sun paradox. Additional forcing, such as a lowerEarth-albedo (9), is necessary to maintain temperate conditionsin the early Archaean.

Oxygen isotope compositions of ISB serpentines suggest thatthey formed in the presence of seawater with a δ18O similar tomodern oceans, consistent with oxygen isotope studies of Archae-an biogenic phosphates (4), volcanic rocks from other Archaeangreenstone Belts (13, 43), and mafic pillow lavas and sheeteddikes in the ISB (14). We therefore suggest that the low δ18O ofArchaean chemical sediments is a result of postdepositionalexchange with shallow ground- or pore waters.

Analytical MethodsMineral separates were hand-crushed or powdered using a micro-drill for hydrogen and XRD analyses; rock samples were crushedand sieved to a size of approximately 0.5 mm, then handpickedusing a stereo microscope for oxygen analysis. Mineral puritywas visually assessed, and samples with better than 95% puritywere used for analysis. Hydrogen and oxygen isotope analyses ofmineral separates of serpentine, talc, tremolite, anthophyllite,hornblende, biotite, and muscovite from ultramafic lenses andadjacent amphibolite and felsic rocks within all major lithologiesin the ISB were performed at the Stanford University Stable Iso-tope Biogeochemistry Lab, following the methods of Sharp (50)and Sharp et al. (51). For hydrogen isotope analysis, 1–5 mg ofpowdered mineral separate were dried at vacuum for >24 h, thendropped into a Finnigan high temperature conversion elementalanalyzer (TC-EA) using an autosampler flushed with helium.H2 gas produced from sample combustion in a 1,450 °C carbonreduction furnace was introduced into a Finnigan DeltaPlusXLmass spectrometer in a helium gas stream. For oxygen isotopeanalysis, approximately 1 mg of mineral separates were dried atvacuum for >1 h in a nickel sample holder. Samples were heatedusing a CO2-infrared laser in a vacuum chamber containing theoxidizing agent BrF5. The oxygen gas was directly analyzed usinga dual-inlet Finnigan MAT 252 mass spectrometer. Isotopecompositions of the samples were corrected relative to NationalBureau of Standards samples as well as laboratory standards, andare correct within �3.4‰ for hydrogen and �0.2‰ for oxygenof accepted values. Results of these analyses are presented in thestandard delta notation as parts per thousand (‰), relative tothe VSMOW standard. X-ray diffraction was performed on ser-pentine samples to distinguish antigorite, lizardite, and chrysotilepolymorphs, at the Geballe Laboratory at Stanford University.

Table 2. Ocean volume at approximately 3.8 Ga based on δDSERPENTINE ¼ −54.5� 1.5‰

ArchaeanδDSEAWATER

% ContinentsFormed

H lost viaescape (mol H)*

O2 Equivalent(mol O2)

Ocean size:

(mol H2O) (kg H2O) (% modern)

−20 0 2.40 × 1021 0.60 × 1021 9.29 × 1022 1.67 × 1021 11915 3.43 × 1021 0.86 × 1021 9.16 × 1022 1.65 × 1021 11870 7.18 × 1021 1.80 × 1021 8.71 × 1022 1.57 × 1021 112

100 9.23 × 1021 2.31 × 1021 8.46 × 1022 1.52 × 1021 109−30 0 11.93 × 1021 2.98 × 1021 9.76 × 1022 1.76 × 1021 126

15 12.82 × 1021 3.21 × 1021 9.63 × 1022 1.73 × 1021 12470 16.11 × 1021 4.03 × 1021 9.16 × 1022 1.65 × 1021 118

100 17.91 × 1021 4.48 × 1021 8.90 × 1022 1.60 × 1021 114

*Elemental hydrogen.

Pope et al. PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ 4375

EART

H,AT

MOSP

HER

IC,

AND

PLANETARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 2

3, 2

021

ACKNOWLEDGMENTS. We thank P. Blisniuk and R. Jones for their assistance indata preparation and collection, and N. Sleep, C.P. Chamberlain, J. Stebbins,J.R. O’Neil, and anonymous reviewers for their constructive comments.This research was supported by the Danish National Research Foundation

through the Nordic Center for Earth Evolution, and endowment funds fromthe Department of Geological and Environmental Sciences at StanfordUniversity. Portions of the research were supported by the Allan C. CoxProfessorship to M.T.R.

1. Kasting JF, et al. (2006) Paleoclimates, ocean depth, and the oxygen isotopic compo-sition of seawater. EPSL 252:82–93.

2. Jaffrés JBD, Shields GA, Wallmann K (2007) The oxygen isotope evolution of sea-water: A critical review of a long-standing controversy and an improved geologicalwater cycle model for the past 3.4 billion years. Earth-Science Rev 83:83–122.

3. Hren MT, Tice MM, Chamberlain CP (2009) Oxygen and hydrogen isotope evidencefor a temperate climate 3.42 billion years ago. Nature 462:205–208.

4. Blake RE, Chang SJ, Lepland A (2010) Phosphate oxygen isotopic evidence for a tem-perate and biologically active Archaean ocean. Nature 464:1029–1032.

5. Knauth LP, Epstein S (1976) Hydrogen and oxygen isotope ratios in nodular andbedded cherts. Geochim Cosmochim Acta 40:1095–1108.

6. Oskvarek JD, Perry EC, Jr (1976) Temperature limits on the early Archaean ocean fromoxygen isotope variations in the Isua supracrustal sequence, West Greenland. Nature259:192–194.

7. Knauth LP, Lowe DR (2003) High Archean climatic temperature inferred from oxygenisotope geochemistry of cherts in the 3.5 Ga Swaziland Supergroup, South Africa.Geol Soc Am Bull 115:566–580.

8. Robert F, ChaussidonM (2006) A paleotemperature curve for the Precambrian oceansbased on silicon isotopes in cherts. Nature 443:969–972.

9. Rosing MT, Bird DK, Sleep NH, Bjerrum CJ (2010) No climatic paradox under the faintearly sun. Nature 464:744–747.

10. Wenner DB, Taylor HP, Jr (1974) D∕H and 18O∕16O studies of serpentinization ofultramafic rocks. Geochim Cosmochim Acta 38:1255–1286.

11. Lécuyer C, Gruau G, Früh-Green GL, Picard C (1996) Hydrogen isotope composition ofEarly Proterozoic seawater. Geology 24:291–294.

12. Kyser TK, O’Hanley DS, Wicks FJ (1999) The origin of fluids associated with serpenti-nization processes: Evidence from stable-isotope compositions. Canadian Mineralo-gist 37:223–237.

13. Gregory RT (2003) Ophiolites and global geochemical cycles: Implications for theisotopic evolution of seawater. Geol Soc London Special Pub 218:353–368.

14. Furnes H, de Wit M, Staudigel H, Rosing M, Muehlenbachs K (2007) A vestige ofEarth’s oldest ophiolite. Science 315:1704–1707.

15. Frei R, Rosing MT, Waight TE, Ulfbeck DG (2002) Hydrothermal-metasomatic and tec-tono-metamorphic processes in the Isua supracrustal belt (West Greenland): A multi-isotopic investigation of their effects on the Earth’s oldest oceanic crustal sequence.Geochim Cosmochim Acta 66:467–486.

16. Wenner DB, Taylor HP, Jr (1973) Oxygen and hydrogen isotope studies of the serpen-tinization of ultramafic rocks in oceanic environments and continental complexes.Amer J Sci 273:207–239.

17. Evans B (2004) The serpentinite multisystem revisited: Chrysotile is metastable. IntGeol Rev 46:479–506.

18. Shaw AM, Hauri EH, Fischer TP, Hilton DR, Kelley KA (2008) Hydrogen isotopes inMariana arc melt inclusions: Implications for subduction dehydration and thedeep-Earth water cycle. EPSL 275:138–145.

19. Sacoccia PJ, Seewald JS, Shanks WC, III (2009) Oxygen and hydrogen isotope fractio-nation in serpentine-water and talc-water systems from 250 to 450 °C. Geochim Cos-mochim Acta 73:6789–6804.

20. Sheppard SMF (1980) Isotopic evidence for the origins of water during metamorphicprocesses in oceanic crust and ophiolite complexes. Colloques Internationaux duCNRS 272:135–147.

21. Barnes JD, Paulick H, Sharp ZD, Bach W, Beaudoin G (2009) Stable isotope (δ18O, δD,δ37Cl) evidence for multiple fluid histories in mid-Atlantic abyssal peridotites (ODPLeg 209). Lithos 110:83–94.

22. Kyser TK, Kerrich R (1991) Stable Isotope Geochemistry: A Tribute to Samuel Epsteinpp 409–422 The Geochemical Soc. Spec. Pub. No. 3.

23. Alt JC, ShanksWC, III (2006) Stable isotope compositions of serpentinite seamounts inthe Mariana forearc: Serpentinization processes, fluid sources and sulfur metasoma-tism. EPSL 242:272–285.

24. Mével C (2003) Serpentinization of abyssal peridotites at mid-ocean ridges. CRGeoscience 335:825–852.

25. Bowers TS, Taylor HP, Jr (1985) An integrated chemical and stable-isotope model ofthe origin of midocean ridge hot spring systems. J Geophys Res 90:12583–12606.

26. Jean-Baptiste P, Charlou J-L, Stievenard M (1997) Oxygen isotope study of mid-oceanridge hydrothermal fluids: Implication for the oxygen-18 budget of the oceans. Geo-chim Cosmochim Acta 61:2669–2677.

27. Burkhard DJM, O’Neil JR (1988) Contrasting serpentinization processes in the easternCentral Alps. Cont Min Pet 99:498–506.

28. Rose NM, Rosing MT, Bridgwater DB (1996) The origin of metacarbonate rocks in theArchaean Isua supracrustal belt, west Greenland. Am J Sci 296:1004–1044.

29. Friend CRL, Nutman AP (2010) Eoarchean ophiolites? New evidence for the debateon the Isua supracrustal belt, southern West Greenland. Am J Sci 310:826–861.

30. RosingMT (1999) 13C-depleted carbonmicroparticles in >3700-Ma sea-floor sedimen-tary rocks from West Greenland. Science 283:674–676.

31. Criss RE, Taylor HP, Jr (1986) Stable Isotopes in High Temperature GeologicalProcesses. Rev Mineralogy, eds JW Valley, HP Taylor, Jr, and JR O’Neil pp:373–424.

32. Craig H (1961) Isotopic variations in meteoric waters. Science 133:1702–1703.33. Pons M-L, et al. (2011) Early Archean serpentine mud volcanoes at Isua, Greenland,

as a niche for early life. PNAS 108:17639–17643.34. Evans BW (2010) Lizardite versus antigorite serpentinite: Magnetite, hydrogen, and

life (?). Geology 38:879–882.35. Lécuyer C, Gillet P, Robert F (1998) The hydrogen isotope composition of seawater

and the global water cycle. Chem Geol 145:249–261.36. Catling DC, Zahnle KJ, McKay CP (2001) Biogenic methane, hydrogen escape, and the

irreversible oxidation of early Earth. Science 293:839–843.37. Schoell M (1980) The hydrogen and carbon isotopic composition of methane from

natural gases of various origins. Geochim Cosmochim Acta 44:649–661.38. Catling DC (2006) Comment on “A hydrogen-rich early Earth atmosphere”. Science

311:38a.39. Kasting JF, Holm NG (1992) What determines the volume of the oceans? Earth Planet

Sc Lett 109:507–515.40. Knauth LP (1992) Isotopic Signatures and Sedimentary Records, Lecture Notes in Earth

Sciences #43, eds N Clauer and S Chaudhuri (Springer-Verlag, Berlin), pp 123–152.41. Kolodny Y, Epstein S (1976) Stable isotope geochemistry of deep sea cherts. Geochim

Cosmochim Acta 40:1195–1209.42. Haqq-Misra JD, Domagal-Goldman SD, Kasting PJ, Kasting JF (2008) A revised, hazy

methane greenhouse for the Archean Earth. Astrobio 8:1127–1137.43. Hessler AM, Lowe DR, Jones RL, Bird DK (2004) A lower limit for atmospheric carbon

dioxide levels 3.2 billion years ago. Nature 428:736–738.44. Sheldon ND (2006) Precambrian paleosols and atmospheric CO2 levels. Precamb Res

147:148–155.45. Divins DL (2010) NGDC total sediment thickness of the world’s oceans and marginal

seas, http://www.ngdc.noaa.gov/mgg/sedthick/sedthick.html. Retrieved 21 May 2010.46. Bickle MJ (1978) Heat loss from the Earth: A constraint on Archaean tectonics from

the relation between geothermal gradients and the rate of plate production. EPSL40:301–315.

47. Abbott DH, Hoffman SE (1984) Archaean plate tectonics revisited 1. Heat flow,spreading rate, and the age of subducting oceanic lithosphere and their effectson the origin and evolution of continents. Tectonics 3:429–448.

48. Korenaga J (2006) Archean Geodynamics and Environments, AGU GeophysicalMonograph Series, eds K Benn, J-C Mareschal, and K Condie pp:7–32.

49. Herzberg C, Condie K, Korenaga J (2010) Thermal history of the Earth and itspetrological expression. EPSL 292:79–88.

50. Sharp ZD (1990) A laser-based microanalytical method for in situ determinationof oxygen isotope ratios of silicates and oxides. Geochim Cosmochim Acta54:1353–1357.

51. Sharp ZD, Atudorei V, Durakiewicz T (2001) A rapid method for determination ofhydrogen and oxygen isotope ratios from water and hydrous minerals. Chem Geo178:197–210.

52. Agrinier P, Hekinian R, Bideau D, JavoyM (1995) O and H stable isotope compositionsof oceanic crust and upper mantle rocks exposed in the Hess Deep near the Galapa-gos Triple Junction. EPSL 136:183–196.

53. Sakai R, Kusakabe M, Noto M, Ishii T (1990) Origin of waters responsible for serpen-tinization of the Izu-Ogasawara-Mariana fore-arc seamounts in view of hydrogenand oxygen isotope ratios. EPSL 100:291–303.

54. Bonatti E, Lawrence JR, Morandi N (1984) Serpentinization of ocean peridotites:temperature dependence of mineralogy and boron content. EPSL 70:88–94.

55. Skelton ADL, Valley JW (2000) The relative timing of serpentinization and mantleexhumation at the ocean-continent transition, Iberia: constraints from oxygen iso-topes. EPSL 178:327–338.

56. Ribeiro da Costa I, Barriga FJAS, Viti C, Mellini M, Wicks FJ (2008) Antigorite indeformed serpentinites from the mid-atlantic ridge. Eur J Mineral 20:563–572.

57. Magaritz M, Taylor HP, Jr (1974) Oxygen and hydrogen isotope studies of serpenti-nization in the Troodos ophiolite complex, Cyprus. EPSL 23:8–14.

58. Yui T, Yeh H, Lee CW (1990) A stable isotope study of serpentinization in the Fengtienophiolite, Taiwan. Geochim Cosmochim Acta 54:1417–1426.

59. Liakhovitch VV, Quick JE, Gregory RT (2005) Hydrogen and oxygen isotope con-straints on hydrothermal alteration of the Trinity peridotite, Klamath Mountains,California. Int Geol Rev 47:203–214.

4376 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1115705109 Pope et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

3, 2

021