The oxygen isotope composition of earth's oldest rocks and evidence of a terrestrial magma ocean



[1] Analysis of Hadean and Archean rocks for 16O-17O-18O isotopes demonstrates that the Terrestrial Mass Fractionation Line of oxygen isotopes has had the same slope and intercept for at least the past 4.0 and probably for as long as 4.2 Ga. The homogenization of oxygen isotopes required to produce such long-lived consistency was most easily established by mixing in a terrestrial magma ocean. The measured identical oxygen isotope mass fractionation lines for Earth and Moon suggest that oxygen isotope reservoirs of both bodies were homogenized at the same time during a giant moon-forming impact. But other sources of heat for global melting cannot be excluded such as bolide impacts during early accretion of proto-Earth, the decay of short-lived radioactive isotopes, or the energy released during segregation of core from mantle.

1 Introduction

[2] Magma oceans offer a widely accepted hypothesis for explaining the feldspathic Lunar Highlands and understanding the segregation of planetary cores from their mantles but, by their nature, magma oceans are notably successful in erasing outcrop evidence of their existence. There are no chilled margins, no xenoliths, no hornfels, no wall rocks, familiar proofs of the former presence of molten magma, because, in a global magma ocean, everything is melted. The hypothesis survives, however, because it has passed permissive geochemical and petrological tests. For Earth's Moon, for example, plagioclase flotation in a differentiated, mafic, lunar magma ocean has been replicated in laboratory experiments, confirmed by measured complementary enrichments of incompatible elements such as potassium, rare earth elements, and phosphorus (KREEP) and validated by the observation that anorthositic and KREEP basaltic lunar rocks share ancient crystallization ages [Warren, 1985; Shearer et al., 2006; Elkins-Tanton, 2012; Yamamoto et al., 2012]. Here original data on the 16O-17O-18O oxygen isotopic compositions of some of Earth's oldest rocks are used to provide a permissive test of the Terrestrial Magma Ocean hypothesis.

[3] It was not difficult to melt planets in the early Solar System. Collisions of Mercury-sized planetesimals produced planetary melting during the culminating paroxysm of solar system evolution some 107 years after the formation of refractory calcium aluminum inclusions, the oldest solids in the solar system [Tonks and Melosh, 1993; Chambers, 2007; Halliday, 2007]. Heat from decaying short-lived radioactive nuclides such as 26Al, still present in a youthful solar system, would melt growing planetesimals during early accretion [McKeegan and Davis, 2007]. Gravitational separation of immiscible molten metal from a silicate magma ocean to form planetary cores would generate even more heating [Halliday, 2007]. A key consequence of magma ocean formation from the perspective of oxygen isotope cosmochemistry is the thorough mixing of components of a planetary body owing to vigorous convection in the melt, driven by thermal gradients from liquidus temperatures to the black cold of space. Impacts and impact heating would be effective in mixing as well. Violent collisions of planetesimals are predicted to produce molten disks homogeneous in oxygen isotopes in times as short as 102 to 103 years by turbulent mixing of impactor and target [Pahlevan and Stevenson, 2007].

[4] The degree of oxygen isotopic heterogeneity of the planetary embryos accreting to form early Earth is unknown. Meteorites and returned samples from the solar wind, Moon, Itokawa, Comet 81P/Wild 2, and interplanetary dust particles are the only non-terrestrial samples available for direct analysis. Many meteorites have ages older than those of Earth's oldest rocks. Oxygen isotope analyses of meteorites and their solar system components differ by up to hundreds of per mil in δ17O, δ 18O, and Δ17O in respect to Earth's mass fractionation line. Of all the meteorite groups, only enstatite chondrites are virtually identical in oxygen isotopic composition to bulk Earth [Javoy et al., 2010]. Surveys of meteorite groups demonstrate that Earth can be assembled from combinations of chondritic meteorites [Lodders, 2000; Burbine and O'Brien, 2004]. Recently published analyses of Si-isotopes in terrestrial and extraterrestrial materials, however, do not support the hypothesis of Earth originating from enstatite chondrites [Fitoussi and Bourdon, 2012]. The solar wind which samples the Sun, the largest reservoir of oxygen in the solar system, is enriched in 16O by some 30‰ (Δ17O = −30‰) relative to Earth [McKeegan et al., 2011]. Taken together, available evidence reveals a solar nebula heterogeneous in oxygen isotopic composition from which rocky planets and asteroids accreted to form the inner solar system. Accordingly, we assume the planetesimals that accreted to form early Earth were heterogeneous in oxygen isotopic composition but their degree of heterogeneity is unknown.

[5] The existence of highly correlated loci of δ18O vs. δ17O data, termed mass-dependent fractionation lines, are a predictable outcome of the igneous differentiation of a planetary reservoir of initially uniform oxygen isotopic composition [Matsuhisa et al., 1978, Appendix I; Young et al., 2002]. Such mass fractionation lines are a fundamental characteristic of solar system planetary bodies and may date from the earliest stages of accretion although, as in the Earth, subsequent events may obscure their ancient age. Analyses of shergottites, nakhlites, and chassignites, rare achondritic meteorites that were blasted off Mars by impacts and fell to Earth, demonstrate a linear plot of data with a slope of approximately 0.52 on a graph of δ18O (horizontal) versus δ17O (vertical) offset by +0.3‰ Δ17O in relation to Earth [Clayton and Mayeda, 1983; Franchi et al., 1999]. Linear plots of oxygen isotope ratios with slopes of 0.52x are also known for the Moon [Wiechert et al., 2001; Spicuzza et al., 2007], angrite meteorites [Greenwood et al., 2005], howardite, eucrite, and diogenite meteorites [Wiechert et al., 2004; Greenwood et al., 2005], and the silicate inclusions of IVA iron meteorites [Wang et al., 2004]. Groups of meteorites sharing the same mass fractionation line are considered to have originated from the same parent body even though the body itself may be unknown [Clayton, 2007]. The angrites, for example, all share the same planetary oxygen isotope mass fractionation line offset by −0.07 in Δ17O in respect to Earth and record basaltic volcanism on a small planet taking place over the interval 4556.60 (±0.26) to 4563.37 (±0.25) Ma but the parent body is unknown [Greenwood et al., 2005; Amelin, 2008; Brennecka and Wadwha, 2012].

[6] Oxygen isotope analyses of terrestrial rocks and minerals plot on a slope 0.52x line coincident with that of the Moon [Matsuhisa et al., 1978; Clayton and Mayeda, 1983; Robert et al., 1992; Wiechert et al., 2001; Spicuzza et al., 2007]. Not only rocks and minerals but also meteoric waters, from the ice of polar regions to tropical monsoon rain, plot on mass-dependent fractionation lines [Li and Meijer, 1998; Barkan and Luz, 2005]. The only known deviations from terrestrial mass-dependent fractionation lines are found in tropospheric and stratospheric ozone, stratospheric CO2, the N-, S-, and Cl-oxy-anions of atmospheric reactions in which ozone is the oxidant, and in aerosol particles of sulfate, perchlorate, and nitrate resulting from oxidation by ozone. The oxygen we are breathing has a small excess of 16O in relation to the mass fractionation line defined by meteoric water [Luz et al., 1999]. Surface deposits of sulfate and nitrate minerals in arid desert regions record mass-independent oxygen isotope fractionations inherited from accumulated aerosol particles [Bao et al., 2000; Michalski et al., 2003; Thiemens, 2006]. The carbonate-associated sulfate of Neoproterozoic carbonate sediments is known to record anomalous, mass-independent deficiencies in 17O/16O [Bao et al., 2008]. Microscopic carbonate mineral grains found in atmospheric aerosol particles show excess 17O/16O [Shaheen et al., 2010]. Although of deep interest to atmospheric scientists, terrestrial occurrences of mass-independently fractionated oxygen isotopes are infinitesimal in abundance compared to the whole Earth.

[7] The objective of the present study is to determine the oldest date by which a uniform, mass-dependently fractionated oxygen isotope reservoir had been established on Earth. The data on 16O-17O-18O ratios of rocks and minerals presented below were measured on rocks ranging in age from 3.8 Ga (Isua, Greenland) to 4.0 Ga (Acasta, Canada). Their oxygen isotope mass fractionation lines are indistinguishable from those of younger rocks, within analytical error. The oxygen isotope composition of the accessible Earth, its crust of sediments, volcanics, and intrusives, as well as the available mantle sample, evolved from an ancient reservoir homogeneous in oxygen isotopes that must have been present prior to 4.0 Ga and probably as long ago as 4.2 Ga. That reservoir most likely was a terrestrial magma ocean melted by the combined heating effects of bolide impacts during accretion, radioactive decay, core segregation, and a giant impact that produced the Moon.

2 Mass-Dependent Fractionation of Oxygen Isotopes

[8] The origin of mass-dependent fractionation lines with slopes of approximately 0.52 has been investigated by Matsuhisa et al. [1978, appendix I, p. 180] and Young et al. [2002]. Recalling the definition, δ18O = {[(18O/16O)sample/(18O/16O)reference] − 1} × 1000‰ (per mil) and similarly for δ17O, the final composition of a substance participating in oxygen isotope exchange reactions is given in terms of its initial composition by

display math(1)

where the exponent, λ, is

display math

for equilibrium isotope exchange and miO are exact atomic weights of the isotopes of oxygen [Young et al., 2002, equation 37, p. 1099]. The calculated numerical value of λ is 0.530, in good agreement with the measured equilibrium fractionation of oxygen isotopes between liquid water and water vapor, i.e., 0.529 (±0.001) [Barkan and Luz, 2005].

[9] The implication of equation (1) for understanding planetary mass fractionation lines is that for a planetesimal initially homogeneous in oxygen isotopes the products of its subsequent differentiation all should lie along a line defined by its initial composition and the value of λ. Whether the planetary body considered is Earth, Mars, Moon, or Vesta, its rocky differentiates should define a single mass fractionation line in δ18O versus δ17O coordinates. The parameters of mass fractionation lines are independent of the temperature-pressure histories of planetary bodies except insofar as changes in environmental conditions accelerate or slow the rate of chemical reactions and accompanying isotope exchange. The strongest control on the slope of mass fractionation lines is by reaction kinetics [Young et al., 2002, compare equation 15, p. 1097 and equation 25, p. 1098]. Kinetic isotope fractionation between stem and leaf water of plants by a mechanism of evapotranspiration results in measured λ values of 0.5111–0.5204 [Landais et al., 2006].

[10] Miller [2002, equation 4, p. 1882; cf. Hulston and Thode, 1965] emphasized that for mass fractionation lines, values of the exponent, λ, and the y-axis intercept, ka,b, could be obtained expeditiously by least squares regression of analyses of related groups of natural samples or experimental run products using the formula

display math

where conventional deltas are converted to a logarithmic form. In the data analysis of this paper, the formulation of Miller [2002] is used, together with the ISOPLOT program of Ludwig [2003], to perform linear least squares regression of measured oxygen isotope data.

3 Earth's Oldest Rocks

[11] The samples of Earth's oldest rocks analyzed in this study were collected from the Acasta gneisses which are exposed along the western edge of the Slave craton, Canada, and from the gneisses and metasedimentary rocks of the Isua Greenstone Belt, southwestern Greenland. The Acasta gneiss consists of banded tonalities, granodiorites, granites, and amphibolites. The crystallization ages of the oldest Acasta orthogneisses are 4030–3940 Ma as determined by analysis by ion microprobe and laser ablation-inductively coupled plasma–mass spectrometry [Bowring et al., 1989; Bowring and Housh, 1995; Bowring and Williams, 1999; Sano et al., 1999; Iizuka et al., 2007]. Furthermore, the Hf isotopic compositions of zircons [Amelin et al., 2000; Iizuka et al., 2009] and the presence of the core of a zircon xenocryst from an Acasta tonalitic gneiss with an U-Pb age of 4203 ± 58 Ma (see oxygen isotope analysis of sample AC 012, Table 1) [Iizuka et al., 2006] indicate that parental magmas were generated by melting or assimilation of even older crustal materials. Note that oxygen isotope analyses reported in Table 1 were made on aliquots of the same Acasta rocks whose U-Pb ages had been measured previously. Sample numbers of Table 1 are the same as those used by the original authors. Sample BGXM (Table 1), for example, a banded gneiss from Acasta, is discussed by Bowring et al. [1989], Bowring et al. [1990], and Bowring and Housh [1995].

Table 1. Oxygen Isotope Analyses and U-Pb Crystallization Ages of Acasta Gneiss Samplesa
SampleRock Typeδ17O (‰)δ18O (‰)U-Pb Crystallization Age
  1. a

    O-isotope analyses performed on aliquots of samples used for U-Pb age dating. Sample numbers are those used in previous publications. O-isotope data given relative to nominal value of working standard without additional calibration to avoid introducing artifacts in least squares regression (see text). Add 0.4 (‰) to tabulated δ18O value to obtain approximate value calibrated to VSMOW.

  2. b

    Two samples of SA91-37-3 obtained at different settings of Frantz magnetic separator from same initial aliquot.

  3. c

    Bowring et al. [1989].

  4. d

    Bowring and Housh [1995].

  5. e

    Bowring and Williams [1999].

  6. f

    Sample contains 4.2 Ga zircon xenocryst; Iizuka et al. [2006].

  7. g

    Iizuka et al. [2007].

BGXMTonalitic/amphibolitic orthogneiss3.356.463964 ± 4 Mac-e
SB89-38Tonalitic orthogneiss2.885.583740 Mad
SAB91-37-3bTonalitic orthogneiss3.045.874012 ± 6 Mad, e
 Tonalitic orthogneiss3.556.83 
AC 012Tonalitic gneiss3.767.263942 ± 32 Maf
AC 461Quartz diorite gneiss (Pb lossg)3.827.32>3590 Mag
AC 478Tonalitic gneiss, light3.967.603974 ±26 Mag
 Tonalitic gneiss, dark3.176.16 
AY 066Granitic gneiss4.047.803586 ± 26 Mag

[12] Previous investigation showed the oxygen isotope composition of Acasta orthogneisses to reach δ18O values as high as +8.9‰ relative to ocean water, interpreted as evidence of the incorporation of 18O/16O-enriched crustal materials from an exogenic cycle still older than the 3960 Ma orthogneiss [Muehlenbachs and Bowring, 1992]. Values of δ18O > +7.5‰ in Hadean detrital zircons from the Jack Hills, Western Australia, have been taken as indirect evidence of the presence of liquid water on Earth's surface as early as 4.3–4.4 Ga [Wilde et al., 2001; Mojzsis et al., 2001]. Magmatic differentiation of mantle-derived rocks is incapable of enriching 18O/16O to the extent observed. The large isotope fractionations accompanying low-temperature processes such as weathering, diagenesis, and hydrothermal alteration however result in strong 18O/16O enrichments. Such high δ18O values in magmatic rocks or in minerals presumed to have crystallized from magmas indicate interactions between magmas and surface rocks through assimilation of wall rocks or partial melting of sediments.

[13] The Isua Greenstone Belt includes tonalitic gneisses, pillow basalts, banded iron formations, and turbidites, possibly representing a fragment of ancient seafloor [Rosing et al., 1996; Komiya et al., 1999]. The ages of Isua protoliths were first proposed to be between ca. 3.7 and 3.8 Ga [Moorbath et al., 1977; Crowley, 2003]. Systematic analysis of U and Pb in zircons collected from the terrane in which Isua supracrustals are embedded has revealed tonalites from 3690 to 3810 Ma, ultramafic rocks of 3810 Ma age or greater, and confirms ages of 3710–3800 Ma in Isua sediments and volcanics [Nutman et al., 2004].

[14] Oxygen isotope analyses of Isua samples are given in Table 2. Oxygen isotope analyses of a dunite body located 20 km south of the Isua supracrustal belt give δ18OVSMOW values of +4.5–4.9‰ for olivine, +5.7–5.9‰ for orthopyroxene, and +0.5–2.5‰ for chromite [Lowry et al., 2003]. Pillows from Isua basalts have cores with δ18O values of 6.5–9.9‰ and associated basaltic dikes from 5.7‰ to 6.9‰ [Furnes et al., 2007]. A much wider range in δ18O, from +8.0‰ to +18.0‰, is seen for in situ spot analyses of quartz from Isua banded iron formations measured with an ion microprobe [Heck et al., 2011].

Table 2. Oxygen Isotope Analyses of Rocks and Minerals From Isua Supracrustal Belta
SampleRock typeMineralδ17O (‰)δ18O (‰)
  1. a

    Sample numbers are those used in previous publications. O-isotope data given relative to nominal value of working standard without additional calibration to avoid introducing artifacts in least squares regression (see text). Add 0.4‰ to tabulated δ18O value to obtain approximate value calibrated to Vienna SMOW. Quoted data are averages of duplicate and triplicate analyses of separate aliquots of samples. BIF, banded iron formations.

  2. b

    Samples supplied by M. Rosing.

  3. c

    cf. Rosing [1999, Table 1, p. 674].

  4. d

    Samples supplied by T. Komiya [Komiya et al., 1999].

  5. e

    Samples from Lepland et al. [2002, Table 1, p. 226].

  6. f

    cf. van Zuilen et al. [2003, Table 1, p. 334].

460516bGraphitic schist 5.2410.09
810025bFuchsite quartzite 6.0211.56
810196bSandy turbidite 3.927.54
810213b, cGraphitic Slate 4.298.31
930063bGarnet-hornblende schist 5.6610.86
990200bWhole rock 7.3214.11
56-AdBIF, magnetite-richmagnetite2.524.92
56-BdBIF, silicate-richsilicate5.3410.30
 BIF, silicate-richmagnetite2.575.01
193-AdBIF, silicate-richsilicate6.1111.77
193-BdBIF, magnetite-richmagnetite1.232.37
230-AdBIF, magnetite-richmagnetite0.972.02
232-AdBIF, magnetite-richmagnetite1.102.19
232-BdBIF, silicate-richsilicate5.2310.08
 BIF, silicate-richmagnetite1.452.74
IPL 18-BdPillow basaltmatrix3.426.58
IPL 18-CdPillow basaltbreccia2.855.49
IPL 18-FdPillow basaltcore3.326.40
AL10-3 eBIFquartz7.1313.67
AL15-Be, fBIFquartz6.8713.19

4 Samples and Sample Analysis

[15] Analyzed rock and mineral samples are aliquots of samples of Earth's oldest rocks for which data have been previously published. The samples were received primarily in the form of 10 cm pieces of whole rocks. Purified mineral separates were contributed to the study as well and were obtained by standard magnetic and density separation methods. Whole rock samples were slabbed with a diamond-bladed rock saw, photographed, and then cut into lithologic-specific billets. In a 12 cm section of a metamorphosed pillow basalt from greenstone in Isua, each layer of pillow matrix, pillow rim, and pillow core was sawn apart and analyzed separately (samples IPL 18-B, IPL 18-C, and IPL 18-F, Table 2). Banded iron formations were cut parallel to dominant layering to isolate magnetite-rich from silicate-rich rock types and analyzed separately. Dark biotite- or amphibole-rich layers were cut apart from lighter-colored ones in gneissose samples. Once billets of individual rock types had been cut, the samples were crushed in a percussion mortar and pestle, sieved to pass 100 mesh and retain on 200 mesh, washed in deionized (DI) distilled H2O, briefly ultrasonicated in dilute HCl to remove weathering products, washed once again in DI distilled H2O, and dried. For magnetite-bearing samples, a separation was made between grains adhering to a hand magnet (magnetite plus inevitable composite grains) and grains that were not attracted by the magnet (silicate fraction primarily). The sample preparation was adopted to provide a spectrum of sample types including lithologic-specific whole rocks and easily separated minerals. Mineral separates were used to test for oxygen isotope exchange between minerals from the same lithologic type.

[16] It is known from previous studies of coesite-eclogite and diamond-eclogite facies rocks metamorphosed under the severe pressure-temperature conditions of the upper mantle that primary oxygen isotope signatures of crustal whole rocks are typically preserved [Rumble and Yui, 1998; Masago et al., 2003]. Mantle peridotites sandwiched between 18O-depleted, coesite-eclogite crustal rocks retained pristine mantle O-isotope values in both olivine and pyroxene despite exhumation from depths of 100 km [Zhang et al., 2005]. Thus, the strategy of analyzing whole rocks may be reasonably expected to yield useful data on the primary oxygen isotope composition of metamorphosed ancient rocks.

[17] Samples of whole rocks and partial mineral separates were analyzed by laser fluorination using cryogenically purified BrF5 as fluorinating agent [Clayton and Mayeda, 1963] and a Synrad 25 W CO2 infrared laser to heat the samples [Sharp, 1990]. Approximately 2 mg of each sample was loaded in duplicate in a nickel-metal holder with a capacity for 12 samples. Four aliquots of USNM 107144 Gore Mountain garnet were loaded as working isotope ratio reference material for each batch of eight unknowns. The reaction chamber was fluorinated at room temperature repeatedly, including overnight, until no trace of oxygen released by spontaneous fluorination could be measured. Once the reaction chamber and loaded samples had been passivated by room-temperature fluorination, 30 torr of BrF5 was expanded into the chamber, and each sample was slowly and separately heated to incandescence with a defocused laser beam 100 µm in diameter, taking care to avoid sample scattering. The evolution of oxygen was monitored by reading reaction chamber pressure with a capacitance manometer. Once fluorination was completed, oxygen was separated from remaining BrF5, SiF4, HF, and other undesired reaction products over liquid nitrogen and pumped by a single-stage, in-line mercury diffusion pump onto molecular sieve 5A frozen with liquid nitrogen. Sample oxygen was transferred on molecular sieve 5A to the dual inlet of a Thermo-Fisher MAT 252 mass spectrometer where it was analyzed for 16O-17O-18O. Every oxygen sample was tested for NF3 and CF4 contamination by scanning the mass range from 40 to 75 amu with a Faraday cup whose pre-amplifier resistor had a value of 3 × 1011 Ω. Aliquots of Gore Mountain garnet USNM 107144 were analyzed each day in duplicate during the analysis of samples. These results were used to correct for daily and long-term analytical drift. The reference garnet, USNM 107144, gives a δ18OVSMOW value of 6.0‰ in comparison to UWG-2 [Rumble et al., 1997; Valley et al., 1995]. The 2 sigma standard deviations of Gore Mountain garnet analyses measured during the course of this study were as follows: Δ17O ±0.03, δ17O ±0.09, and δ18O ±0.17‰. The value used for the slope of the terrestrial fractionation line, 0.526 (±0.001), used in this laboratory to calculate Δ17O was determined by interlaboratory comparison of a suite of garnet analyses [Rumble et al., 2007].

5 Data Analysis

[18] The present method of data analysis employs least squares regression of measured δ17O and δ18O values to derive the slope, intercept, and statistical uncertainties of mass fractionation lines from Hadean and Archean outcrops (Tables 1-3). Measured δ17O and δ18O values were converted from conventional delta notation to a logarithmic form defined by the equations [Miller, 2002, equation 4, p. 1882]:

display math
Table 3. Oxygen Isotope Systematics of Proterozoic-Archean-Hadean Rocksa
AreaProtolith AgeSlope δ18O versus δ17OIntercept (‰)MSWDeNo. Data PointsRange in δ18O (‰)
  1. a

    Data on δ18O for coesite- and diamond-bearing eclogites from China and Kazakhstan previously published but slopes and intercepts have not been published.

  2. b

    Rumble and Yui [1998].

  3. c

    Zhang et al. [2005].

  4. d

    Masago et al. [2003].

  5. e

    The MSWD parameter “mean square of weighted deviates,” quoted in Figure 1 and Table 3, is calculated by ISOPLOT for each regression [Ludwig, 2003]. According to Ludwig, MSWD is a rough measure of the “ratio of the observed scatter of the points (from the best-fit line) to the expected scatter (from the assigned errors and error correlations). The MSWD parameter cannot be compared to the classical R2 parameter, and is not a measure of how highly correlated the X- and Y-values are. If the assigned errors are the only cause of scatter, the MSWD will tend to be near unity. MSWD values much greater than unity generally indicate either underestimated analytical errors, or the presence of non-analytical scatter. MSWD values much less than unity generally indicate either overestimated analytical errors or unrecognized error-correlations.” [Ludwig, 2003, footnote 3, p. 22].

Coesite eclogitesb, c750 to 800 Ma0.529 ± 0.001+0.00 (±0.03)0.88250−11 to +6
Coesite eclogitesd1100 to 1300 Ma0.528 ± 0.001−0.02 (±0.05)1.3256−8 to +14
Diamond eclogitesd1100 to 1300 Ma0.526 ± 0.005−0.05 (±0.04)0.9679+6 to +13
Isua, Greenland3800 Ma0.525 ± 0.003+0.04 (±0.03)0.4760+2 to +14
Acasta, Canada4000 Ma0.527 ± 0.011−0.05 (±0.08)0.5232+6 to +8

[19] Logarithmic data were regressed with Ludwig's ISOPLOT computer program [Ludwig, 2003]. Miller (personal communication) has emphasized the necessity of avoiding artifacts introduced unwittingly in the course of data normalization. Accordingly, in this study, two types of data were analyzed: (1) “raw” data, that is, the delta value of the difference in composition between the mass spectrometer O2 gas working standard and the O2 gas derived by laser fluorination from unknown rocks and minerals; and (2) delta values corrected to the nominal Vienna standard mean ocean water (VSMOW) value of the working standard O2 gas, including a small correction for the interference of 17O17O on the abundance of 16O18O. No significant difference in slope was observed between regressions of the two types of data. The “raw” data, however, reveal a positive y-axis (δ17O) intercept of 0.3‰ which records the fact that the commercially produced tank O2 used in this laboratory as a working standard has a deficiency in 17O/16O, as observed for atmospheric O2 by Luz et al. [1999]. The regression of “corrected” delta values yields y-axis intercepts of zero, within analytical uncertainty. The data subjected to linear regression have not been normalized to a δ18O value of 5.8‰ for UWG-2 garnet [Valley et al., 1995] in order to avoid the problem of artifacts as noted by Miller. The values of δ18O reported in Tables 1 and 2 may be normalized to UWG-2, approximately, by adding 0.4‰.

6 Results

[20] Regression of the δ17O′ versus δ18O′ values of 3.8 Ga Isua rocks and minerals gives a slope of 0.525 (±0.003) and a y-axis intercept of 0.04 (±0.03; mean square of weighted deviates (MSWD) = 0.47). Acasta's ca. 4.0 Ga rocks yield statistically indistinguishable values of 0.527 (±0.011) and −0.05 (±0.08; MSWD = 0.52) for slope and intercept, respectively (Figure 1 and Table 3, definition of MSWD). Regression of Acasta data is hampered by the smaller range of δ18O values measured, 2‰, compared to a range of 12‰ for Isua samples. From the perspective of linear regression, a short straight line is less well defined than a longer one. Analyzed Acasta rocks are all orthogneisses: because they crystallized at magmatic temperatures, their range of oxygen isotope fractionations is small. Isua rocks, however, include banded iron formations, sediments deposited under low temperature, surface conditions which, correspondingly, exhibit a wider range in δ18O values that facilitate more precise regression analysis.

Figure 1.

Plot of δ18O versus δ17O for analyzed Isua and Acasta samples, calibrated to VSMOW.

[21] Analyses of quartz, actinolite, grunerite, and magnetite (Table 2, samples 56-A and B, 193-A and B, 230-A, 232 A and B, AL10-3, AL15-B, and AL35-6) separated from banded iron formations (BIF) plot at the high and low ends of Isua's mass fractionation line. These minerals crystallized in a metamorphic, thermal event subsequent to deposition of BIF sediments. Their analyses are concordant with those of whole rock samples indicating their growth in an oxygen reservoir inherited from protoliths.

7 How Old Are the Mass Fractionation Lines?

[22] Oxygen, an element with only stable isotopes, cannot be used for geochronology by direct analysis. The crystallization age of oxygen-bearing minerals, however, may be analyzed by measuring the abundances of radioactive parents and their daughter isotopes bound in the minerals. Inasmuch as the crystallization of a mineral such as zircon, for example, from magma presupposes the availability of sufficient oxygen, silicon, and zirconium to form a correct stoichiometric formula, it is evident that the oxygen of such a mineral must have been present when crystallization began. The age of the oxygen reservoir sampled by a zircon crystal is at least as old as its crystallization age, as determined by analysis of U-Pb isotopes. If a different mineral, say pyroxene, nucleated at the same time as the zircon and if 18O/17O/16O are fractionated between the two growing crystals, both crystals are sampling the same local oxygen reservoir. In this case, the age of the mass fractionation line defined by zircon and pyroxene would be the same as the crystallization age.

8 Discussion

[23] The results of this investigation demonstrate that the slopes and intercepts of oxygen isotope mass fractionation lines have been the same in Earth rocks for more than 4.0 Ga, from the Hadean to Archean to Proterozoic rocks and to present-day meteoric water (Table 3) [Li and Meijer, 1998]. The presence of a 4,203 Ma zircon xenocryst in Acasta sample AC 012 (Table 1) as well as Hf-isotope data demonstrates that Acasta magmas formed by remelting or assimilation of still older crustal material [Amelin et al., 2000; Iizuka et al., 2006; Iizuka et al., 2009]. In previous work, Robert et al. [1992] demonstrated that the mass-dependent fractionation lines of modern cherts with a slope 0.522 (±0.009) and an intercept of 0.05 (±0.30), Precambrian cherts, 0.515 (±0.002) and 0.02 (±0.03), and mantle-derived rocks ranging in age from recent to 3.5 Ga to be statistically indistinguishable. The present results are consistent with the earlier work but benefit from improved age dating methods and laser fluorination oxygen isotope techniques. Recent determinations of the terrestrial fractionation line give slopes of 0.526 (±0.001) by Spicuzza et al. [2007] and 0.525 (±0.001) by Wiechert et al. [2001]. An interlaboratory comparison of measurements of the terrestrial mass fractionation line between the Geophysical Laboratory and the Planetary and Space Sciences Research Institute gave 0.524 (±0.002) for quartz samples from a variety of low-temperature environments and 0.526 (±0.001) for garnets from ultrahigh-pressure metamorphic rocks [Rumble et al., 2007].

[24] The age of formation of a terrestrial homogeneous oxygen isotope reservoir may be deduced by considering the origin of the Moon. Earth and Moon have identical oxygen isotope mass fractionation lines and similar estimated bulk δ18O values of 5.5–5.7‰ [Wiechert et al., 2001; Spicuzza et al., 2007]. Measured Earth-Moon isotopic similarities include not only oxygen but also chromium [Lugmair and Shukolyukov, 1998], titanium [Zhang et al., 2012], and tungsten [Touboul et al., 2007]. These isotopic similarities impose strong constraints on formation of the Moon and, until recently, presented an obstacle to full acceptance of the giant impact model of lunar origin. The latest impact models, however, are consistent with the condensation of the Moon from a debris disk derived primarily from the mantle of proto-Earth [Canup, 2012; Cuk and Stewart, 2012], thus accounting for observed isotopic similarities. With regard to the existence of a terrestrial magma ocean, computer simulations show Earth to have a temperature greater than 6440 K following a giant impact, more than hot enough to insure a molten surface [Canup, 2012]. The age of formation of a terrestrial magma ocean as a consequence of the giant Moon-forming impact may be estimated as approximately 60 Myr after formation of the solar system [Touboul et al., 2007; Amelin et al., 2010; Connelly et al., 2012]

9 Conclusion

[25] Prior to the crystallization of Acasta orthogneisses and the melting of their protoliths, at least as long ago as 4.0 Ga and probably before 4.2 Ga, there must have been a global homogenization of the oxygen isotopic composition of the Earth. Such a mixing of oxygen isotopes was realized in a global magma ocean. Whether the heat needed to melt proto-Earth was acquired through the energy released by bolide impacts during accretion of proto-Earth, or during a giant, Moon-forming impact, or by the decay of radioactive isotopes, or the segregation of core from mantle, there were ample opportunities to melt and mix early Earth. The oxygen isotope data presented herein do not permit a distinction to be made between alternative models for melting Earth but they are consistent with the existence of a terrestrial magma ocean.


[26] Rumble acknowledges support from NASA's Cosmochemistry program, grant NNX07AI48G. He is grateful to the Bureau de la Recherche et de l'Innovation of the Marie de Paris for support of a sabbatical at l'Institut de Physique du Globe de Paris (l'IPGP). The hospitality of Claude Jaupart, Pierre Cartigny, and l'Equipe de Physico-chimie des Fluids Geologiques contributed materially to the present study. N. Assayag (l'IPGP) analyzed samples AL10-3, quartz and tremolite, and AL 35-6, quartz and magnetite (Table 2). The authors would like to acknowledge the Museum of Evolving Earth at the Tokyo Institute of Technology, Tokyo, Japan, for access to rock samples. M. T. Rosing was supported by the Danish National Research Foundation grant DNRF53.Constructive reviews by A. Shahar, L. Elkins-Tanton, K. Muehlenbachs, and an anonymous reviewer improved the manuscript.