oxygen-isotope evidence for recycled crust in the sources of mid-ocean-ridge basalts

letters to nature

530 NATURE | VOL 403 | 3 FEBRUARY 2000 | www.nature.com

conduits into adjacent peridotite could affect the entire meltingregion beneath a mid-ocean ridge. If peridotite adjacent to a meltconduit were exposed to alkali diffusive in®ltration for ,7 Myr(assuming a solid upwelling rate of centimetres per year), thediffusive lengthscaleÐgiven by l � 2�Dft�0:5, where D is diffusioncoef®cient and f is melt porosityÐwould be ,230 m: this assumesthat f = 1% and that D for Na (DNa) is 6 ´ 10-5 cm2 s-1. Thus, thisprocess could affect melting of all peridotite located between meltconduits spaced less than 0.5 km apart.

Abyssal peridotites near to hotspots have the largest trace-element depletions and the highest olivine/orthopyroxene ratiosof all peridotites, this is commonly ascribed to higher degrees ofmelting and higher mantle temperatures6. An alternative explana-tion implied by the present results is that alkali-rich melts from thehotspot source have ¯uxed through shallow peridotite. Evidence forthe diffusive in®ltration process comes from two observationsconcerning Na2O. First, although most incompatible-element con-tents in abyssal peridotites require a fractional melting process6,Na2O contents are higher than expected from such a process29.Second, Na2O shows much less variability in MORB suites thanelements with similar bulk partition coef®cients30, which is con-sistent with the diffusive in®ltration process acting to homogenizedifferences in Na2O.

The experiments I report here show that diffusive in®ltration ofsodium will rapidly occur whenever alkali-rich, silica-poor magmascontact peridotite at shallow depth and magmatic temperatures.This process can strongly affect the melting relations of peridotite: itcould provide an explanation for the genesis of high-silica melts inspinel lherzolite xenoliths, and could also explain the preferentialmelting of orthopyroxene in the shallow mantle beneath mid-oceanridges. M

Received 14 April; accepted 9 December 1999.

1. Toomey, D. R., Wilcock, W. S. D., Solomon, S. C., Hammond, W. C. & Orcutt, J. A. Mantle seismic

structure beneath the MELTregion of the East Paci®c Rise from P and S wave tomography. Science 280,

1224±1227 (1998).

2. McKenzie, D. The generation and compaction of partially molten rock. J. Petrol. 25, 713±765 (1984).

3. Waff, H. S. & Bulau, J. R. Equilibrium ¯uid distribution in an ultrama®c partial melt under

hydrostatic stress conditions. J. Geophys. Res. 84, 6109±6114 (1979).

4. Klein, E. & Langmuir, C. H. Global correlations of ocean ridge basalt chemistry with axial depth and

crustal thickness. J. Geophys. Res. 92, 8089±8115 (1987).

5. Kinzler, R. J. & Grove, T. Primary magmas of mid-ocean ridge basalts; 2, Applications. J. Geophys. Res.

97, 6907±6926 (1992).

6. Johnson, K. T. M., Dick, H. J. B. & Shimizu, N. Melting in the oceanic upper mantle: an ion

microprobe study of diopsides in abyssal peridotites. J. Geophys. Res. 95, 2661±2678 (1990).

7. Hirose, K. & Kushiro, I. The effect of melt segregation on polybaric mantle melting: Estimation from

the incremental melting experiments. Phys. Earth Planet. Inter. 107, 111±118 (1998).

8. Kelemen, P. B., Shimizu, N. & Salters, V. J. M. Extraction of mid-ocean-ridge basalt from the

upwelling mantle by focused ¯ow of melt in dunite channels. Nature 375, 747±753 (1995).

9. Kelemen, P. B., Hirth, G., Shimizu, N., Spiegelman, M. & Dick, H. J. B. Review of melt migration

processes in the adiabatically upwelling mantle beneath oceanic spreading ridges. Phil. Trans. R. Soc.

Lond. 355, 283±318 (1997).

10. Stolper, E. M. A phase diagram for mid-ocean ridge basalts: Preliminary results and implications for

petrogenesis. Contrib. Mineral. Petrol. 74, 13±27 (1980).

11. Hirose, K. & Kushiro, I. Partial melting of dry peridotites at high pressures; determination of

compositions of melts segregated from peridotite using aggregates of diamond. Earth Planet. Sci. Lett.

114, 477±489 (1993).

12. Takahashi, E. Melting of a dry peridotite KLB-1 up to 14 GPa: implication on the origin of peridotitic

upper mantle. J. Geophys. Res. 91, 9367±9382 (1986).

13. Lundstrom, C. C., Gill, J. & Williams, Q. A geochemically consistent hypothesis for MORB generation.

Chem. Geol. 162, 105±126 (2000).

14. Watson, E. B. Basalt contamination by continental crust: some experiments and models. Contrib.

Mineral. Petrol. 80, 73±87 (1982).

15. Watson, E. B. & Jurewicz, S. R. Behavior of alkalies during diffusive interaction of granitic xenoliths

with basaltic magma. J. Geol. 92, 121±131 (1984).

16. Lesher, C. E. Kinetics of Sr and Nd exchange in silicate liquidsÐtheory, experiments, and applications

to uphill diffusion, isotopic equilibration, and irreversible mixing of magmas. J. Geophys. Res. 99,

9585±9604 (1994).

17. Kushiro, I. On the nature of silicate melt and its signi®cance in magma genesis: regularities in the shift

of the liquidus boundaries involving olivine, pyroxene and silica minerals. Am. J. Sci. 275, 411±431

(1975).

18. Ryerson, F. J. Oxide solution mechanisms in silicate meltsÐsystematic variations in the activity

coef®cient of silica. Geochim. Cosmochim. Acta 49, 637±649 (1985).

19. Hirschmann, M. M., Baker, M. B. & Stolper, E. M. The effect of alkalies on the silica content of mantle-

derived melts. Geochim. Cosmochim. Acta. 62, 883±902 (1998).

20. Walter , M. J. & Presnall, D. C. Melting behavior of simpli®ed lherzolite in the system CaO-MgO-

Al2O3-SiO2-Na2O from 7 to 35 kbar. J. Petrol. 35, 329±359 (1994).

21. Baker, M. B., Hirschmann, M. M., Ghiorso, M. S. & Stolper, E. M. Compositions of near-solidus

peridotite melts from experiments and thermodynamic calculation. Nature 375, 308±311 (1995).

22. Robinson, J. A. C., Wood, B. J., & Blundy, J. D. The beginning of melting of fertile and depleted

peridotite at 1.5 GPa. Earth Planet. Sci. Lett. 155, 97±111 (1998).

23. Frey, F. A. & Green, D. H. The mineralogy, geochemistry, and origin of lherzolite inclusions in

Victorian basanites. Geochim. Cosmochim. Acta 38, 1023±1059 (1974).

24. Vannucci, R., Botazzi, P., Wulf-Pedersen, E. & Neumann, E. R. Partitioning of REE, Y, Sr, Zr and Ti

between clinopyroxene and silicate melts in the mantle under La Palma (Canary Islands): implications

for the nature of the metasomatic agents. Earth Planet. Sci. Lett. 158, 39±51 (1998).

25. Shaw, C. S. J., Thibault, Y., Edgar, A. D. & Lloyd, F. E. Mechanisms of orthopyroxene dissolution in

silica-undersaturated melts at 1 atmosphere and implications for the origin of silica-rich glass in

mantle xenoliths. Contrib. Mineral. Petrol. 132, 354±370 (1998).

26. Draper, D. S. & Green, T. H. P±T phase relations of silicic, alkaline, aluminous mantle-xenolith glasses

under anhydrous and C-O-H ¯uid-saturated conditions. J. Petrol. 38, 1187±1224 (1997).

27. Dick, H. J. B., Fisher, R. L. & Bryan, W. B. Mineralogic variability of the uppermost mantle along mid-

ocean ridges. Earth Planet. Sci. Lett. 69, 88±106 (1984).

28. Niu, Y., Langmuir, C. H. & Kinzler, R. J. The origin of abyssal peridotites: a new perspective. Earth

Planet. Sci. Lett. 152, 251±265 (1997).

29. Elthon, D. Chemical trends in abyssal peridotites; refertilization of depleted suboceanic mantle.

J. Geophys. Res. 97, 9015±9025 (1992).

30. Langmuir, C. H. & Hanson, G. An evaluation of major element heterogeneity in the mantle sources of

basalts. Phil. Trans. R. Soc. Lond. A 297, 383±407 (1980).

Acknowledgements

I thank P. Kelemen and P. Asimow for comments on the manuscript, and P. Hess,D. Forsyth, Y. Liang, Q. Williams and M. Rutherford for suggestions. I also thankM. Rutherford for use of his laboratory, J. Devine and M. Jercinovic for assistance with theelectron microprobes, and K. Hoernle, E. Takahashi and M. Per®t for sample donation.This work was supported by an SGER grant from the NSF.

Correspondence and requests for materials should be addressed to the author at theUniversity of Illinois (e-mail: [email protected]).

.................................................................Oxygen-isotope evidence forrecycled crust in the sourcesof mid-ocean-ridge basaltsJohn M. Eiler*, Pierre Schiano*²³, Nami Kitchen* & Edward M. Stolper*

* Division of Geological and Planetary Sciences, California Institute of Technology,

Pasadena, California 91125, USA² Laboratory Geochimie±Cosmochimie, IPG Paris, 4 Place Jussieu, 75252,

Paris cedex 05, France³ UniteÂ Mixte de Recherche 6524 `Magmas et Volcans', UniversiteÂ Blaise-Pascal,5 rue Kessler, 63038 Clermont-Ferrand cedex, France

..............................................................................................................................................

Mid-ocean-ridge basalts (MORBs) are the most abundant terres-trial magmas and are believed to form by partial melting of aglobally extensive reservoir of ultrama®c rocks in the uppermantle1. MORBs vary in their abundances of incompatible ele-ments (that is, those that partition into silicate liquids duringpartial melting) and in the isotopic ratios of several radiogenicisotope systems2±4. These variations de®ne a spectrum between`depleted' and `enriched' compositions, characterized by respec-tively low and high abundances of incompatible elements5,6.Compositional variations in the sources of MORBs could re¯ectrecycling of subducted crustal materials into the source reservoir7,or any of a number of processes of intramantle differentiation8±10.Variations in 18O/16O (principally sensitive to the interaction ofrocks with the Earth's hydrosphere) offer a test of thesealternatives. Here we show that 18O/16O ratios of MORBs arecorrelated with aspects of their incompatible-element chemistry.These correlations are consistent with control of the oxygen-isotope and incompatible-element geochemistry of MORBs by a

© 2000 Macmillan Magazines Ltd

letters to nature

NATURE | VOL 403 | 3 FEBRUARY 2000 | www.nature.com 531

component of recycled crust that is variably distributed through-out their upper mantle sources.

Geochemical variations in MORBs occur in several physicalsettings. These include a global-scale isotopic signature of enrich-ment in low latitudes (the `DUPAL' anomaly centred on theIndian ocean11); enrichments spatially associated with ocean-island basalt volcanic centres12; enrichments in near-ridgeseamounts13; scattered anomalies of enriched MORB (EMORB,de®ned as having chondrite-normalized La/Sm ratios greater than1) within ridge segments dominated by normal MORB (NMORB,de®ned as having chondrite-normalized La/Sm ratios less than orequal to 1; ref. 14); and ubiquitous, minor variations in thecomposition of NMORBs (for example, ref. 15) that may re¯ecteither weak in¯uences of the other types of enrichment or a separatephenomenon.

Several aspects of the compositional heterogeneity of MORBs(including variations in abundance ratios of highly incompatibleelements and in radiogenic-isotope ratios) cannot easily beexplained as consequences of variations in extent of melting of ahomogeneous source. Instead, these aspects suggest variations inthe compositions of their mantle sources. The following mechan-isms have been proposed as causes of compositional variations inthe mantle and may contribute to heterogeneity in MORBs: (1)variable extents of depletion by prior partial melting (that is,enriched sources could be residual to relatively less melting thandepleted sources)8,13; (2) in®ltration of depleted mantle by mantle-derived silicate melts or hydrous and/or carbonic ¯uids rich inincompatible elements9; (3) mixing of depleted mantle rocks withenriched, metasomatized portions of continental lithosphericmantle detached from the roots of continents10; (4) mixing betweendepleted mantle and crustal rocks, sediments, and/or ¯uids injectedinto the mantle by subduction (for example, refs 7, 16).

The ®rst three of these hypotheses involve intramantle fractiona-tions controlled by high-temperature partitioning of elements

between minerals and silicate melts or ¯uids; it is therefore dif®cultto discriminate between them. The fourth hypothesis, however,requires material from the crust as the agent of enrichment. Crustalrocks and sediments often have distinctive properties resulting fromexposure to the hydrosphere, including increases or decreases ind18O (where d18O = ((18O/16Osample)/0.0020052 - 1)1000) relative tovalues typical of mantle peridotites; for example, altered sea-¯oorbasalts and sediments are elevated in d18O by ,5 and ,15±25½respectively, relative to average mantle17,18, whereas gabbroic rocksin the lower oceanic crust are often lower in d18O than mantle rocksby up to ,5½ (ref. 17).

Previous studies of oxygen-isotope variability of fresh MORBsmeasured values of d18O between 5.35 and 7.30½ (ref. 19; seeSupplementary Information). Variations were not found to besigni®cantly correlated with established indicators of geochemicalvariability in mantle sources of MORBs; it has therefore beengenerally suspected that oxygen-isotope ratios in fresh MORBsonly vary to small extents, because of trace alteration, magmaticdifferentiation in crustal magma chambers, or fractionations duringmelting19.

We analysed d18O values and water contents of 28 MORB glassesfrom the east-Paci®c rise, the mid-Atlantic ridge and the IndianOcean (Table 1); sampling locations and other geochemicalcharacteristics of these samples (including previous measurementsof d18O and water content) are referenced in the SupplementaryInformation. Analytical techniques used in this study are describedin the Methods. These samples exhibit ranges in the isotopiccompositions of Sr, Nd, Pb, and Os that span most of the rangeknown in MORBs (refs 5, 6, 20). All but two samples are within thecompositional range of NMORBs; the two exceptions havechondrite-normalized La/Sm ratios of 1.04 and 1.05; that is, theyare on the border between NMORB and EMORB compositions.Samples were collected from segments of the ridge system distantfrom regions clearly in¯uenced by hot-spot volcanism (such as

Table 1 Oxygen isotope data and water contents of MORB glasses

Sample d18O* 61 j 61 standard error wt% H2O Mg number² La/SmN³...................................................................................................................................................................................................................................................................................................................................................................

Atlantic OceanRidelente DR10 5.50 0.04 0.02 0.23 0.67 1.05CH77 DR05 105a 5.40 0.01 0.00 0.27 0.60 n.d.Mapco CH98 DR12 5.41 0.02 0.01 0.24 0.58 0.48Mapco CH98 DR11 5.45 0.02 0.01 0.27 0.63 0.50Mapco CH98 DR10 5.49 0.05 0.04 0.27 0.61 0.48EW 9309 25D 5.81 0.04 0.03 0.17 0.64 0.97EW 9309 1D 5.47 0.01 0.01 0.23 0.59 0.84...................................................................................................................................................................................................................................................................................................................................................................

East Paci®c RiseNaudure ND 51 5.47 0.02 0.02 0.20 0.66 0.60Naudure DR21-4 5.41 0.04 0.03 0.21 0.61 0.57Searise 1 DR5-102 5.43 0.02 0.02 0.26 0.60 0.56Searise 1 Dr04 5.61 0.02 0.02 0.30 0.55 0.70CHEPR 60 5.43 0.06 0.03 0.17 0.62 0.54CY 82 27-1 5.51 0.03 0.02 0.30 0.62 0.75Clipperton DR01-01a 5.51 0.09 0.04 0.22 0.64 n.d.Venture number 32 5.67 0.03 0.02 0.28 0.53 n.d.Searise 2 DR07 5.49 0.08 0.04 0.39 0.47 0.53CYP 78 18-62 5.45 0.03 0.02 0.20 0.65 0.69CYP 78 18-65 5.54 0.03 0.02 0.24 0.61 n.d.CYP 78 04-06 5.37 0.01 0.00 0.18 0.60 0.62CYP 78 06-10 5.42 0.03 0.02 0.16 0.65 0.55...................................................................................................................................................................................................................................................................................................................................................................

Indian OceanMD57 D 13-7 5.41 0.04 0.03 0.47 0.59 0.64MD 57 D910-1 5.46 0.03 0.02 0.43 n.d. 0.85MD 57 D910-4 5.61 0.01 0.00 0.29 0.63 1.04MD 57 D8-1 5.57 0.06 0.04 0.22 n.d. 0.59MD 23 site 4 5.49 0.05 0.04 0.24 0.71 0.68MD 37 07-04-D1 5.45 0.05 0.04 0.23 0.61 0.75MD 34 D3 5.47 0.01 0.01 0.36 n.d. 0.37MD 34 D4 5.64 0.01 0.00 0.36 0.58 0.85...................................................................................................................................................................................................................................................................................................................................................................

* d18O � ��18O=16Osample=18O=16OSMOW�2 1� 3 1000, where 18O=16OSMOW � 0:0020052. SMOW, standard mean ocean water.

² Mg number based on data from ref. 20 and the relationship: Mg# = Mg/(0.9 ´ Fe + Mg) where Mg and Fe are molar abundances.³ Chondrite-normalized La/Sm ratio.n.d., not determined; j, standard deviation.


letters to nature


Iceland), although evidence has been used to infer an in¯uence of`plumes' on otherwise normal ridge segments21 and our suite mightsample such effects.

We measured a range in d18O for fresh MORB glass of 5.37±5.81½ (average of 5.50½). This range is signi®cantly smaller thanthat observed in previous studies of MORBs19, although the mini-mum d18O observed here is indistinguishable from that observed inprevious studies. The differences in range, average and maximumd18O between this and previous studies could re¯ect differences inanalytical precision (see Table 1 and Methods), subtle but systematicdifferences in the extent of sample alteration, or heterogeneity infresh MORBs not sampled by our suite.

Values of d18O observed in this study are correlated with absoluteand relative abundances of incompatible elements (Fig. 1), a featurenot present in previous data sets. The best de®ned of these correla-tions are between d18O and the concentration of K2O and the K2O/TiO2, K/Sr, K2O/H2O, and La/Ti ratios; signi®cant but weakercorrelations are observed with the La/Sm, K/U and Ba/Th ratios.Relationships between d18O and radiogenic-isotope ratios will notbe considered in detail here (see ref. 20 for radiogenic-isotope data);brie¯y, relatively high d18O samples are consistently radiogenic intheir 87Sr/86Sr and 187Os/186Os ratios and have relatively high valuesof the D207Pb index11, which is consistent with the association ofhigh d18O with `enriched' radiogenic-isotope compositions. How-ever, some lower-d18O samples also have enriched radiogenic-isotope compositions, and thus the relationship of oxygen isotopesto radiogenic-isotope enrichment is more complex than that toelemental enrichment (Fig. 1).

Values of d18O for samples examined in this study are notcorrelated with the abundances of incompatible elements that aresensitive to the extent of melting but are not usually considered tore¯ect enriched geochemical signatures in MORBs (for example,Na2O(8.0) and TiO2; ref. 22, see ref. 20 for major element composi-tions). In addition, most of the parameters plotted in Fig. 1 (that is,those other than K2O, K2O/TiO2 and La/Ti) are expected to besensitive only to variations in source composition, provided thetotal extents of melting of the sources are more than about 1±2%(ref. 23). Although a component of very low degree melt appears toin¯uence the trace-element geochemistry of some MORBs (ref. 24),such melts are believed to be only minor components of eruptedlavas. Therefore the small (less than ,0.1½) differences in melt-residue oxygen-isotope fractionations expected between high- andlow-degree melts of a homogeneous source25 are unlikely to con-tribute signi®cantly to the observed variations in d18O of eruptedlavas. We therefore conclude that the trends in Fig. 1 are unlikely tore¯ect variable extents of melting of a homogenous source. How-ever, there are few experimental constraints on oxygen-isotopepartitioning during such processes, and thus this hypothesisshould be re-examined when such data become available. Neither

d18O nor other variables illustrated in Fig. 1 are well correlated withMg number (the principal index of differentiation in basaltic lavas;Table 1). Furthermore, most elemental indices correlated with d18Oin Fig. 1 are insensitive to fractional crystallization. We thereforealso conclude that the variations in d18O and relationships of d18Oto chemistry in the MORBs studied here do not re¯ect magmaticdifferentiation alone. Finally, d18O values are not correlated withH2O abundances (Table 1), suggesting that they do not re¯ect theassimilation of water-rich rocks or sediments in the young oceaniccrust through which they were erupted or 18O-enrichment duringsubsolidus hydration of the sampled glasses17.

We conclude from the discussion above that the variations ind18O observed in the MORB glasses included in this study princi-pally re¯ect variations in the d18O of their mantle sources. Given theassociation of elevated d18O with enrichments in elements that areconcentrated in high-d18O oceanic crustal rocks and sediments (forexample, K, Ba), the trends in Fig. 1 could re¯ect the presence ofvariable amounts of 18O-enriched oceanic crustal materials in thesources of the studied lavas. This hypothesis can also explain the lackof a relationship between d18O and H2O content (Table 1). That is,crustal materials are believed to be strongly dehydrated duringsubduction26, so their addition to the upper mantle could enrich itin elements that are concentrated in the oceanic crust but notef®ciently removed during subduction without strong enrichmentin H2O. We examined this hypothesis by comparing our data with amodel in which mantle peridotite that is normal in d18O andrelatively poor in incompatible elements mixes with subducted,dehydrated oceanic upper-crustal rocks and sediments. Accordingto the model, this mixed source then melts by a constant amount togenerate MORBs (solid curves, Fig. 1; the dashed curves in Fig. 1eand f show the results of an alternative model in which enrichmentsare introduced as separate domains of eclogite; see Methods fordetails). We note that mixing between mantle peridotites andoceanic, lower-crustal gabbros would probably produce trends ofdecreasing d18O with increasing enrichment and would therefore beinconsistent with our observations. However, the contrasts in thevalues of d18O and of most incompatible-element concentrationsbetween mantle peridotites and oceanic lower-crustal gabbros17,27

are subtle, and therefore the trends in Fig. 1 would also be consistentwith models in which the recycled materials contain both upper-oceanic crust and a component of gabbro from the lower-oceaniccrust.

The observed trends in Fig. 1 are well described by the crustalrecycling model described above and are consistent with thepresence of 0±7 wt% of subducted upper-crustal material in thesources of MORB magmas ranging from NMORB to transitionalEMORB. We conclude that recycling of subducted oceanic crustplausibly controls variations in d18O and correlated aspects of theincompatible-element geochemistry of these MORBs. The slope of

Table 2 Compositions of components for crustal-recycling model

Depleted mantle* Recycled sediment² Recycled, altered MORB² Bulk recycled upper crust³ Dk (solid/melt) peridotite D (solid/melt) eclogite...................................................................................................................................................................................................................................................................................................................................................................

wt%K2O 0.005 2.04 0.40 0.48 0.0053 0.0041TiO2 0.199 0.62 1.35 1.31 0.086 0.44H2O 0.026 0.2 (7.3) 0.2 (2.0) 0.2 0.012 0.012p.p.m.Ba 0.56 776 30 67 0.0003 0.00059U 0.004 1.68 0.14 0.22 0.0007 0.00195Th 0.011 6.91 0.19 0.53 0.0077 0.0015La 0.15 28.8 3.0 4.3 0.0061 0.028Sm 0.24 5.78 2.6 2.8 0.0287 0.312Sr 12 327 80 92 0.0183 0.08d18O§ 5.4 20 10 10.5...................................................................................................................................................................................................................................................................................................................................................................

* De®ned such that the composition of a 10% batch partial melt will have a composition near the low-d18O end of the trends in Fig. 1.² Water content after dehydration assumed based on K/H versus La/Sm systematics in MORB from ref. 15. Water content before subduction given in italics.³ Assuming 5% sediment, 95% altered oceanic crust.§ d18O of listed source component, estimated for crustal components using data from refs 17, 18, and references supplied in the Supplementary Information.kD is the distribution coef®cient de®ned as the ratio of concentrations (by weight) between two coexisting phases (for example, solid/melt) at equilibrium.All other elemental abundances and distribution coef®cients are based on data in references provided in the Supplementary Information and ref. 15.


letters to nature

NATURE | VOL 403 | 3 FEBRUARY 2000 | www.nature.com 533

the trend for our data in Fig. 1e (and perhaps Fig. 1f) agrees moreclosely with the model in which the enriched material is present asma®c domains (dashed curves) than with the model that assumesenrichment to be a compositional component of a mineralogicallyhomogeneous peridotite (solid curves), consistent with previoussuggestions of such a distribution of recycled crust in the sources ofMORBs (see ref. 13). We note that even the depleted ends of thetrends in Fig. 1 need not be entirely free of recycled crust; if not,the quantitative modelling we have presented would underestimatethe importance of recycled materials in the incompatible-elementand oxygen-isotope geochemistry of the upper-mantle sources ofMORBs.

Given estimates of the average elemental composition ofNMORB (ref. 28), our model suggests that their upper-mantlesources contain an average of ,2% recycled upper-crustal oceanicrocks and sediments. If this estimate represents a steady-statecondition5, and if we assume that the MORB reservoir correspondsto the upper 660 km of the mantle5 and that residence times ofincompatible elements within it are ,1 Ga (ref. 5), then our resultwould require that the long-term rate of addition of upper-crustalrocks and sediments to the sources of MORB is approximately 6 km3

per year. This estimate is close to the current rate at which upperoceanic crust is both produced and subducted (,1/3 of the rate oftotal oceanic crust production and consumption, given the typicalstratigraphy of the oceanic crust29), suggesting that the quantities ofupper-crustal material that we call upon as an agent of chemicalheterogeneity in the sources of MORBs are reasonable. The abun-dances of incompatible elements used in our models for thedepleted mantle and recycled upper-crustal components (Table 2)and the composition of average NMORB (ref. 28) suggest that, onaverage, approximately half of the total K, Ba and U, and 10±30% ofthe Sr and rare-earth elements in the upper mantle sources ofNMORB, are introduced by subducted oceanic crust. Our inter-pretation suggests that subduction of oceanic crust into the sourcesof MORBs has shifted the average d18O of those sources by ,0.1½with respect to depleted upper-mantle peridotites. This resultconstrains models of the oxygen-isotope budget of the upper

mantle±crust±hydrosphere system, although detailed explorationof such models is beyond our scope here.

The fact that the data from the Atlantic, Indian and Paci®c MORBsamples studied here all follow the same trends in Fig. 1 suggests thatdistinctions among the major ocean basins in other geochemicalcharacteristics (see ref. 11) do not apply to variations in d18O andincompatible-element abundances among the sources of MORB.Instead, our results suggest that compositional heterogeneitiescontributing to the variability of oxygen-isotope ratios and corre-lated incompatible-element characteristics in the sources of MORBsare similar in origin and magnitude over a global length scale andmay be a background heterogeneity onto which other types ofenrichment are superimposed. This interpretation supportsprevious conclusions that sources of MORBs contain variableamounts of globally disseminated, enriched components andsuggests that such components are principally derived fromrecycling of the oceanic crust7,16 rather than from intramantledifferentiation8±10. M

MethodsMeasurements of d18O were made by laser ¯uorination at the California Institute ofTechnology using a 50-watt CO2 laser, BrF5 as the reagent, and a gas-handling apparatussimilar to previous designs (see Supplementary Information); all gases were analysed on aFinnigan 251 gas-source isotope-ratio mass spectrometer. Data were standardized usingseveral silicate standards (SCO-1 and GMT-2, see Supplementary Information; andAH95-22, a basalt glass standard used at Caltech); data gathered on separate days werenormalized to a common value for these standards (the average correction for any givenday was 0.06). After normalization, replicate analyses of AH95-22 glass standard had astandard deviation of 60.03 (n = 27); the standard deviations for replicate analyses ofunknowns also averaged 60.03½ and are listed for individual samples in Table 1. Theuncertainty in the mean for each sample (that is, 1j/n1/2) averaged 60.02½ (also listed inTable 1 for unknowns). Measurements of water content were made by Fourier transforminfrared spectroscopy at Caltech, with methods referenced in the SupplementaryInformation.

The crustal mixing model (curves in Fig. 1) assumes a depleted mantle source (de®nedto have a composition consistent with generating lavas at the low-d18O ends of the trendsin Fig. 1; see Table 2) into which oceanic upper-crustal material (assumed to be a mixtureof 5% average sea-¯oor sediment and 95% average altered sea-¯oor basalt; Table 2) ismixed. Note that oceanic sediments and altered crustal rocks are assumed to besigni®cantly dehydrated but otherwise compositionally similar to existing sea¯oormaterials. This source then undergoes 10% batch partial melting using bulk partition

5%

2%

g

5.4

5.5

5.6

5.7

5.8

5.4

5.5

5.6

5.7

5.8

0.1 0.2 0.3 0.10 0.20

0.5 1.0

0.5 1.0

5%

2%

d5 10 15

10,000 15,000

a b c

e

2%

5%

2%

5%5%

2%

2%

5%

f

50 100 150 200

h

51 2 3 4 6

5%

2%

0.75

5%

2%δ18 O

(‰)

δ18 O

(‰)

wt% K2O K2O/TiO2 K2O/H2OK/Sr

La/Sm(chondrite-normalized)

La/Ti x 104 K/U Ba/Th

±1σ

Figure 1 d18O values versus minor- and trace-element contents and ratios in MORB

glasses. a±h, Data are from ref. 20 and references therein or in the Supplementary

Information; water contents (d) from this study. All abundance ratios are on a weight basis.

Solid curves are calculated compositions of basaltic melts formed by 10% batch melting

of depleted mantle containing a variable component of recycled upper-oceanic crust and

sediment. White circles mark the locations of melts of model depleted mantle alone; black

circles show the locations of partial melts of depleted mantle containing variable amounts

of recycled crust in increments of 1%; compositions of melts derived from sources with

2% and 5% recycled crust are labelled. Dashed curves in e and f are calculated

compositions of basaltic melts formed by mixing of 10% batch partial melts of depleted

peridotite with 10% batch partial melts of eclogite having the composition of recycled

upper-oceanic crust. Dashed curves are omitted from panels other than e and f because

of their close correspondence to the solid curves. See Table 2, Methods, and text for

details and discussion. Symbols discriminate samples by major ocean basin: squares,

Paci®c; diamonds, Atlantic; triangles, Indian. The illustrative error bar is the typical

standard deviation (61j) of replicate measurements of unknowns (Table 1).


letters to nature


coef®cients listed in Table 2; resulting melt compositions are shown in Fig. 1 as circlesjoined by solid curves (the white circle is the model melt of the assumed depleted source;the black circles are for melts of sources to which 1±7 wt% of the crustal component hasbeen added in 1% increments). Sediments and altered crustal rocks vary considerably intheir composition (particularly in their Ba/Th ratios; Fig. 1h) and therefore our modelcurves are best regarded as averages about which the range of melts of natural mantle±crust mixtures might scatter. The calculated solid curves in Fig. 1 assume that partitioncoef®cients for trace elements and oxygen isotope fractionation between melt and residueduring melting are independent of enrichment. This approximation is likely to beincorrect for the La/Sm and La/Ti ratios if enriched material is present as ma®c domainsbecause Sm and Ti are relatively compatible in garnet (a major constituent of high-pressure ma®c rocks). We have therefore shown an alternative model in Fig. 1e and f inwhich MORBs are mixtures of 10% batch partial melts of depleted peridotite and 10%batch partial melts of eclogite having a composition equal to our recycled-crust modelcomponent; eclogite is assumed to be 50% garnet and 50% clinopyroxene. Eclogiteprobably melts to a different extent from peridotite at a given set of conditions. Thedirection and extent of these differences are expected to vary with temperature, pressureand water fugacity; rather than specifying these variables we have adopted the simplifyingassumption that eclogite and peridotite components melt to comparable extents. Theslopes of curves calculated by this model are not signi®cantly changed by adopting otherreasonable assumptions about the melting behaviour of eclogite, although for somemodels the amount of recycled material in the source may be signi®cantly different fromthe fraction of melt derived from that material. This alternative differs little from the ®rstmodel for indices other than La/Sm and La/Ti as a result of the subtle differences betweenmelt±peridotite and melt±eclogite distribution coef®cients for other elements plotted inFig. 1. We therefore omit the results of this alternative calculation from panels other thanFig. 1e and f. We have assumed that oxygen-isotope fractionations during melting areconstant (that is, variations in the d18O of melts are directly proportional to variations intheir sources). Data sources for our model calculations are referenced in the Supple-mentary Information.

Received 12 July; accepted 9 December 1999.

1. Wood, J. A. et al. in Basaltic Volcanism on the Terrestrial Planets 631±699 (Pergamon, New York, 1981).

2. Tatsumoto, M. Genetic relations of oceanic basalts as indicated by lead isotopes. Science 153, 1094±

1101 (1966).

3. Kay, R., Hubbard, N. J. & Gast, P. W. Chemical characteristics and origin of oceanic ridge volcanic

rocks. J. Geophys. Res. 75, 1585±1614 (1970).

4. O'Nions, R. K., Hamilton, P. J. & Evenson, A. M. Variations in 143Nd/144Nd and 87Sr/86Sr ratios in

oceanic basalts. Earth Planet. Sci. Lett. 34, 13±22 (1977).

5. AlleÁgre, C. J. Isotope geodynamics. Earth Planet. Sci. Lett. 86, 175±203 (1987).

6. Zindler, A. & Hart, S. Chemical geodynamics. Annu. Rev. Earth Planet. Sci. 14, 493±571 (1986).

7. Armstrong, R. L. A model for the evolution of strontium and lead isotopes in a dynamic Earth. Rev.

Geophys. 6, 175 (1968).

8. DePaolo, D. J. & Wasserburg, G. J. Inferences about magma sources and mantle structure from

variations of 143Nd/144Nd. Geophys. Res. Lett. 3, 743±746 (1976).

9. Green, D. H. Composition of basaltic magmas as indicators of conditions of origin: application to

oceanic volcanism. Phil. Trans. R. Soc. Lond. A 268, 707±725 (1971).

10. McKenzie, D. & O'Nions, K. Mantle reservoirs and ocean island basalts. Nature 301, 229±231 (1983).

11. Hart, S. R. A large-scale isotope anomaly in the southern hemisphere mantle. Nature 309, 753±757

(1984).

12. Schilling, J. G. Iceland mantle plume: geochemical study of Reykjanes Ridge. Nature 242, 565±571

(1973).

13. Zindler, A., Staudigel, H. & Batiza, R. Isotope and trace element geochemistry of young Paci®c

seamounts: Implications for the scale of upper mantle heterogeneity. Earth Planet. Sci. Lett. 70, 175±

195 (1984).

14. Sun, S. S., Nesbitt, R. W. & Sharaskin, A. U. Geochemical characteristics of mid-ocean ridge basalts.

Earth Planet. Sci. Lett. 44, 119±138 (1979).

15. Michael, P. Regionally distinctive sources of depleted MORB: Evidence from trace elements and H2O.

Earth Planet. Sci. Lett. 131, 301±320 (1995).

16. Hofmann, A. W. Mantle plumes from ancient oceanic crust. Earth Planet. Sci. Lett. 57, 421±436

(1982).

17. Muehlenbachs, K. in Stable Isotopes in High Temperature Geological Processes (eds Valley, J. W., Taylor,

H. P. & O'Neil, J. R.) Reviews in Mineralogy Vol. 16, 425±444 (Mineralogical Society of America,

Washington, DC, 1986).

18. Arthur, M. A., Anderson, T. F. & Kaplan, I. R. in Stable Isotopes in Sedimentary Geology, SEPM Short

Course Vol. 10, 1±151 (SEPM, Tulsa, Oklahoma, 1983).

19. Harmon, R. S. & Hoefs, J. Oxygen isotope heterogeneity of the mantle deduced from global 18O

systematics of basalts from different geotectonic settings. Contrib. Mineral. Petrol. 120, 95±114 (1995).

20. Schiano, P., Birck, J. -L. & AlleÁgre, C. J. Osmium-strontium-neodymium-lead isotopic covariations in

mid-ocean ridge basalt glasses and the heterogeneity of the upper mantle. Earth Planet. Sci. Lett. 150,

363±379 (1997).

21. Mahoney, J., le Roex, A. P., Peng, Z., Fisher, R. L. & Natland, J. H. Southwestern limits of Indian Ocean

ridge mantle and the origin of low 206Pb/204Pb mid-ocean ridge basalt; isotope systematics of the

central Southwest Indian Ridge (17 degrees - 50 degrees E). J. Geophys. Res. B 97, 19771±19790 (1992).

22. Langmuir, C. H., Klein, E. M. & Plank, T. in Mantle Flow and Melt Generation at Mid-Ocean Ridges

(eds Morgan, J. P., Blackman, D. K. & Sinton, J. M.) 183±280 (American Geophysical Union,

Washington, DC, 1992).

23. Shaw, D. M. Trace element behaviour during anatexis. Geochim. Cosmochim. Acta 34, 237±243 (1970).

24. McKenzie, D. & O'Nions, R. K. Partial melt distributions from inversion of rare earth element

concentrations. J. Petrol. 32, 1021±1091 (1991).

25. Matthews, A., Stolper, E. M., Eiler, J. M. & Epstein, S. Oxygen isotope fractionation among melts,

minerals, and rocks. Min. Mag. 62A 971±972 (1998).

26. Poli, S. & Schmidt, M. W. H2O transport and release in subduction zones: Experimental constraints on

basaltic and andesitic systems. J. Geophys. Res. 100, 22299±22314 (1995).

27. Pallister, J. S. & Knight, R. J. Rare-earth element geochemistry of the Samail ophiolite near Ibra,

Oman. J. Geophys. Res. 86, 2673±2697 (1981).

28. Sun, S. S. & McDonough, W. F. in Magmatism in the Ocean Basins (eds Saunders, A. D. & Norry, M. J.)

313±345 (Geological Society of London Special Publication 2, 1989).

29. Crisp, J. A. Rates of magma emplacement and volcanic activity. J. Volcanol. Geotherm. Res. 20, 177±

211 (1984).

Supplementary information is available on Nature's World-Wide Web site(http://www.nature.com) or as paper copy from the London editorial of®ces of Nature.

Acknowledgements

We thank P. Michael; we also thank D. Anderson, P. Asimow, A. Halliday and C. Langmuirfor comments on the manuscript. We thank the Chevron Corporation for donation of themass spectrometer used for isotopic measurements.

Correspondence and requests for materials should be addressed to J.M.E.(e-mail: [email protected]).

.................................................................Quality of the fossil recordthrough timeM. J. Benton*, M. A. Wills*² & R. Hitchin*

* Department of Earth Sciences, University of Bristol, Bristol BS8 1RJ, UK² Oxford University Museum of Natural History, Parks Road, Oxford OX1 3PW,

UK

..............................................................................................................................................

Does the fossil record present a true picture of the history oflife1±3, or should it be viewed with caution4±6? Raup5 argued thatplots of the diversi®cation of life2 were an illustration of bias: theolder the rocks, the less we know. The debate was partiallyresolved by the observation7 that different data sets gave similarpatterns of rising diversity through time. Here we show that newassessment methods, in which the order of fossils in the rocks(stratigraphy) is compared with the order inherent in evolution-ary trees (phylogeny), provide a more convincing analytical tool:stratigraphy and phylogeny offer independent data on history.Assessments of congruence between stratigraphy and phylogenyfor a sample of 1,000 published phylogenies show no evidence ofdiminution of quality backwards in time. Ancient rocks clearlypreserve less information, on average, than more recent rocks.However, if scaled to the stratigraphic level of the stage and thetaxonomic level of the family, the past 540 million years of thefossil record provide uniformly good documentation of the life ofthe past.

The reduction in quality of the fossil record backwards in timeseems self-evident. Fossils in ancient rocks are more likely to havebeen eroded, crushed, melted, subducted, not collected or mis-understood than younger fossils5. The demonstration7 that similarpatterns of diversi®cation were found from different data setsseemed to indicate that, at least when viewed on a broad scale, thefossil record did correctly document the history of life. Indeed,palaeobiologists subsequently used the fossil record as a literalsource of data on the history of the diversi®cation of life8,9 and ofmass extinctions10,11 without applying any correction factors forpossible time-related bias. However, the different data setscompared7 were all subject to the same geological biasing factors,and that study did not use independent data to demonstrate thatolder parts of the fossil record could be trusted. This lingering doubtabout the long-term quality of the fossil record has resurfaced indebates about the timing of origin of major groups of organisms:molecular studies suggest that the Metazoa (animals)12 and modernbird and mammal orders13,14 apparently originated much earlier


oxygen-isotope evidence for recycled crust in the sources of mid-ocean-ridge basalts

Documents