seawater subduction controls the heavy noble gas composition of the mantle

© 2006 Nature Publishing Group

Seawater subduction controls the heavynoble gas composition of the mantleGreg Holland1 & Chris J. Ballentine1

The relationship between solar volatiles and those now in the Earth’s atmosphere and mantle reservoirs provides insightinto the processes controlling the acquisition of volatiles during planetary accretion and their subsequent evolution.Whereas the light noble gases (helium and neon) in the Earth’s mantle preserve a solar-like isotopic composition, heavynoble gases (argon, krypton and xenon) have an isotopic composition very similar to that of the modern atmosphere,with radiogenic and (in the case of xenon) solar contributions. Mantle noble gases in a magmatic CO2 natural gas fieldhave been previously corrected for shallow atmosphere/groundwater and crustal additions. Here we analyse new datafrom this field and show that the elemental composition of non-radiogenic heavy noble gases in the mantle is remarkablysimilar to that of sea water. We challenge the popular concept of a noble gas ‘subduction barrier’— the convectingmantle noble gas isotopic and elemental composition is explained by subduction of sediment and seawater-dominatedpore fluids. This accounts for ,100% of the non-radiogenic argon and krypton and 80% of the xenon. Approximately50% of the convecting mantle water concentration can then be explained by this mechanism. Enhanced recycling ofsubducted material to the mantle plume source region then accounts for the lower ratio of radiogenic to non-radiogenicheavy noble gas isotopes and higher water content of plume-derived basalts.

Owing to their low concentrations and unreactive nature, noble gasesare key tracers of both mantle evolution and the origin of relatedmajor volatiles such as H2O and CO2. The present day complementof noble gases in the terrestrial mantle is a function of severalprocesses. Observed He, Ne and Xe isotopes1–5 in the mantle requirethat the Earth trapped solar-like volatiles during the accretionaryprocess ,4.55 Gyr ago. Superimposed on this primordial signatureare radiogenic/nucleogenic noble gases as well as heavy noble gases(Ar, Kr and Xe) isotopically identical to the atmosphere. Never-theless, early models of mantle noble gas behaviour explicitly requirea subduction barrier for noble gases6, while others attribute volatile-rich ocean island basalts (OIBs) to assimilation of sea water duringeruption rather than as an integral part of the mantle source7,8. Also,no current model can account for the modified noble gas isotopiccomposition of the terrestrial atmosphere without early outgassingfrom a mantle with a solar-like isotopic composition9. Thus far,however, no reproducible deviations from air values in non-radiogenic Ar and Kr isotopes in mantle samples have beenobserved10,11. In mantle steady-state models12,13, air recycling ofheavy noble gases into the convecting mantle is a parameter thatdepends on the composition of an isolated deep mantle.

The issue is compounded by the ubiquitous problem of local aircontamination in all terrestrial basalt samples14,15, which occursby crustal assimilation or during/after eruption on exposure to seawater/atmosphere. The lack of information regarding the extent ofrecycling and the mass balance between accretionary and recycledvolatiles hampers interpretation of data from plume-related ‘hot-spots’ with high 3He/4He ratios and the convecting mantle sampledat mid-ocean ridges. Previous work from CO2 well gases has nowidentified the character and origin of the light noble gases in theconvecting mantle4. In this work, we extend these high precisionanalyses to Xe, which, critically, is alone among the heavy noble gasesin showing a non-air isotopic composition in the non-radiogenic

isotopes. The nature of the sample suite uniquely permits un-ambiguous removal of all local air-derived (groundwater) addition,and now allows us to quantitatively address the issue of volatilerecycling into the mantle.

Xe isotopic composition of the Bravo Dome gas field

The Bravo Dome CO2 gas field in New Mexico, USA, was the focus ofa recent mantle noble gas study4 and the detailed geology andbackground references are given there. The samples from this earlierstudy were unavailable for further analysis and new samples werecollected from producing wells in this field in September 2004, withemphasis on collection from the western section of the gas field,which is rich in mantle-derived noble gases. One sample, WBD-4,was from a well that had been non-producing for 7 days. 3He/4He,20Ne/21Ne/22Ne, 36Ar/38Ar/40Ar, 84Kr and Xe abundance and isotopiccomposition were determined in the Manchester University noblegas laboratory (Tables 1–3). In addition, 15 samples were collectedfrom Sheep Mountain, Colorado, and 5 samples were collected fromMcElmo Dome, Colorado. Xe isotopic data only are presented forthese two fields (see Supplementary Information).

Air and crustal-radiogenic Ne in the gas field are mixed in almostconstant proportion, before variably mixing with the magmatic CO2

(ref. 4). This generates an (airþcrust)–mantle mixing line, whichintersects with the air–mid-ocean-ridge basalt (MORB) mixing lineto uniquely define the mantle Ne isotopic composition. The Xeisotope ratios of Bravo Dome samples show the same systematics,with a pre-mixed (airþcrust) Xe component variably mixingwith mantle Xe. The intersection of the (airþcrust)–mantle Xemixing line with the air–MORB Xe mixing line enables directdetermination of the 129Xe excess in the convecting mantle for thefirst time, where 129Xe/130Xe(mantle) ¼ 7.90 ^ 0.14 (Fig. 1). Excessesin 124-126-128Xe/130Xe relative to air are similar to the single samplemeasured previously5, although with all shielded isotopes plotting on

ARTICLES

1School of Earth, Atmosphere and Environmental Sciences, The University of Manchester, Oxford Road, Manchester M13 9PL, UK.

Vol 441|11 May 2006|doi:10.1038/nature04761

186


the air–solar mixing line (Fig. 2). The least air-contaminated samplerequires a 10 ^ 4% contribution by a solar Xe component. FromFig. 1, it can be clearly seen that even the least air-contaminatedsamples still contain ,50% (airþcrust) Xe. Correcting for this localairþcrust addition allows us to now define the proportion of solar Xe(20 ^ 7%) to Xe with an air-like ratio (80%) in the convectingmantle.

In a three-dimensional plot of i/22Ne versus 21Ne/22Ne and20Ne/22Ne, where i is any noble gas isotope, a data set that is amixture of the same three end-member components will fall on aplane bounded by the crustal-radiogenic, mantle and air end-members. This is shown for i ¼ 130Xe in Fig. 3. If the mantle Neisotopic composition is known, this enables the mantle i/22Necomposition for this sample suite to be determined4. Using thenew Ne isotope data determined in this study, with more samplesfrom the mantle-rich section of the field, we are able to refine theNe isotopic estimate of the mantle end-member to be20Ne/22Ne ¼ 12.49 ^ 0.04 and 21Ne/22Ne ¼ 0.0578 ^ 0.0003 (Sup-plementary Information). We use these new values to calculate theBravo Dome mantle i/22Ne composition, where i ¼ 36Ar, 84Kr and130Xe. In comparing these data to other Solar System reservoirs, were-normalize to 36Ar and plot (i/36Ar)sample/(i/

36Ar)solar (Fig. 4).Measured elemental and isotopic ratios are also presented inTables 1–3.

Evidence for seawater recycling of noble gases

Deriving convecting mantle elemental ratios from MORB and OIBmeasurements is complicated by eruption-related fractionation. Theissue is further confused, because in addition to ubiquitous unfrac-tionated modern-air contamination14, elementally fractionated com-ponents with an air-like isotopic composition are also observed15.Nevertheless, eruption-related fractionation is probably due topartitioning between vesicles and melt and subsequent vesicle lossat elevated pressure16. In this context, highly vesicular MORB glasses

may represent the least elementally fractionated basalt samples17. Thehighly vesicular 2PD43 reference ‘popping rock’ has minimal aircontamination, evidenced by preserving noble gas isotopic compo-sitions indistinguishable from the upper mantle values defined by thewell gas data set4,18,19 (Table 4). We observe that the mantle supplyingthe Bravo Dome has an elemental composition very similar to theleast-contaminated gases released from the 2PD43 popping rock(Fig. 4). It is highly unlikely that two mantle volatile systems, formedin very different ways, would be elementally fractionated to the sameextent. We argue that the similarity in elemental and isotopiccomposition of the two systems is strong prima-facie evidence thatboth the 2PD43 popping rock and the mantle source supplying theBravo Dome well gases are elementally unfractionated samples fromthe same mantle source.

We can now show that the convecting mantle 84Kr/36Ar ratio isdistinct from modern air and that the 130Xe/36Ar ratio is an order ofmagnitude higher than air (Fig. 4). In contrast, the elementalabundance pattern of the heavy noble gases is remarkably close tothat of sea water. Given that subduction zones lie under severalkilometres of sea water this is, perhaps, no surprise. In detail, theconvecting mantle 84Kr/36Ar ratio is slightly enriched in Kr relative tosea water, and the 130Xe/36Ar ratio is within a factor of 2 of sea water(Fig. 4) after correcting the convecting mantle for the,20% solar Xe.This relatively small excess of Xe in the convecting mantle can bereadily attributed to addition of small amounts of oceanic sedimentthat is enriched in the heavy noble gases relative to air and sea water20.We calculate that the amount of sedimentary Xe required to explainelevated 36Ar/130Xe will generate 84Kr/36Ar ¼ 0.053. This is indis-tinguishable from the observed ratio, and accounts for the smallrelative Kr excess relative to seawater values. We argue that seawaterrecycling into the mantle with a small sedimentary componentaccounts for both the elemental and isotopic composition of thenon-radiogenic heavy noble gases.

The ratios 84Kr/36Ar and 130Xe/36Ar in sea water are dominantly

Table 1 | Noble gas elemental data

Sample 4He (1025) Error (1026) 20Ne (1029) Error (10211) 40Ar (1025) Error (1027) 84Kr (10211) Error (10212) 132Xe (10211) Error (10213)

BD11-B 3.93 1.1 1.20 2.89 2.79 3.81 3.46 8.06 0.669 5.72BD20-B 4.17 5.0 1.21 2.92 2.95 4.00 3.68 5.42 0.670 2.81WBD04-B 4.01 4.0 1.09 2.59 3.07 4.16 4.23 2.51 0.687 2.79BD19-B 9.87 1.1 2.43 5.76 4.33 5.86 15.1 10.0 1.78 9.27BD15-B 5.83 3.3 1.70 4.03 3.11 4.21 5.14 9.73 1.04 4.41BD02-B 47.7 1.4 9.43 2.23 7.72 10.5 62.7 10.4 6.56 26.7BD07-B 8.22 0.95 1.93 4.64 3.48 4.71 10.5 7.09 1.31 6.64BD13-B 15.4 1.9 2.97 7.04 4.50 6.09 19.9 4.75 2.12 8.89BD05-B 28.0 1.2 5.08 1.20 6.00 8.13 3.53 14.6 4.10 17.0BD04-B 9.84 2.4 2.35 5.65 3.31 4.48 11.0 8.32 1.40 6.05WBD01-B 5.58 0.47 1.44 3.43 3.17 4.28 5.42 4.33 0.825 3.40WBD03-B 6.64 0.50 1.60 3.80 3.11 4.21 5.52 8.36 0.922 3.39

All values are corrected for full procedural blanks. Concentrations are in cm3 STP per cm3. All quoted errors are 1j.Sample nomenclature: prefix BD, Bravo Dome; WBD, West Bravo Dome.

Table 2 | Isotopic and component ratios

Sample 3He/4He Error 20Ne/22Ne Error 21Ne/22Ne Error 40Ar/36Ar Error 38Ar/36Ar Error

BD11-B 3.681 0.059 11.65 0.14 0.0568 0.0007 19,733 300 0.1878 0.0004BD20-B 3.720 0.067 11.85 0.04 0.0571 0.0003 21,401 392 0.1886 0.0009WBD04-B 3.839 0.065 11.79 0.04 0.0568 0.0003 22,548 730 0.1884 0.0008BD19-B 1.888 0.039 10.70 0.04 0.0543 0.0002 9,888 152 0.1887 0.0005BD15-B 2.424 0.043 10.88 0.07 0.0514 0.0004 13,472 224 0.1886 0.0005BD02-B 0.711 0.062 9.84 0.03 0.0500 0.0002 4,197 69 0.1880 0.0004BD07-B 2.092 0.070 10.89 0.09 0.0554 0.0005 11,205 172 0.1890 0.0004BD13-B 1.275 0.066 10.45 0.05 0.0582 0.0003 7,765 117 0.1878 0.0004BD05-B 0.884 0.016 10.04 0.03 0.0541 0.0002 5,288 81 0.1883 0.0004BD04-B 1.513 0.028 10.45 0.05 0.0517 0.0003 9,612 133 0.1884 0.0006WBD01-B 3.154 0.061 11.47 0.06 0.0565 0.0003 18,496 298 0.1879 0.0005WBD03-B 2.526 0.071 11.23 0.03 0.0556 0.0002 16,792 268 0.1881 0.0010

Errors quoted on all ratios are 1j. 3He/4He are displayed relative to the atmospheric ratio Ra (that is, as R/Ra); Ra ¼ 1.4 £ 1026.

NATURE|Vol 441|11 May 2006 ARTICLES

187


controlled by the relative partial pressure of Ar/Kr/Xe in the atmos-phere and to a minor extent by the temperature and salinity of theoceans. We make the a priori assumption that the atmosphere is notrelated to other reservoirs and serves only as a source for subductednoble gases, the relative concentrations of which have remainedunchanged since atmosphere formation9. Seawater 84Kr/36Ar and130Xe/36Ar have therefore remained nearly constant over the lifetimeof the Earth. Although subduction occurs in a seawater-dominatedfluid environment, the noble gas abundance pattern in any waterphase is highly sensitive to physical processes of fractionation.Because the relative solubilities of Ar and Xe can be an order ofmagnitude different, even a small amount of phase separation willresult in significant elemental fractionation. The absence of fraction-ation from the seawater heavy noble gas abundance pattern suggeststhat the noble gases are unlikely to have been decoupled from thehost sea water. It is reasonable to assume that if recycling of sea wateris responsible for the observed heavy noble gas pattern of theconvecting mantle, there will be a proportional volume of seawater subducted with the atmosphere-derived heavy noble gases.From Fig. 4, it is clear that seawater recycling of He and Ne intothe mantle has negligible effect on the mantle composition owingto the low concentration of He and Ne in sea water, and He/Nerecycling is not discussed further. We further note that contributionsfrom non-seawater/sediment sources to either Ar or Kr are thusnegligible.

Seawater-derived water in the convecting mantle

Recycling sea water back into the convecting mantle means that themantle noble gas concentrations then place a minimum constrainton the recycled water content of the convecting (MORB-source)

mantle. Using the present day 3He flux into the ocean, the averageMORB generation rate and assuming 10% partial melting gives alower limit of 1.2 £ 109 atoms 3He g21 in the convecting mantle.From the 2PD43 popping rock 3He/36Ar ratio (Table 4), the 36Arconcentration in the convecting mantle is 1.5 £ 109 atoms g21. Deepsea water contains 3.37 £ 1013 atoms of 36Ar per g. Therefore,44 p.p.m. H2O in the convecting mantle is associated with recycled36Ar, and is seawater-derived. If we assume that water behavesincompatibly during melting, then unaltered MORB glasses give aconvecting mantle source with 54–120 p.p.m. water21,22. This can betested for compatibility with the mantle D/H systematics. By defi-nition, the terrestrial oceans have a bulk dD ¼ 0‰ (all dD valuesare given here with respect to Standard Mean Ocean Water, SMOW).In contrast, the D/H ratio of mid-ocean-ridge samples and thereforethe convecting mantle is distinctly lower, with dD ¼ 271‰ to291‰. With the simplifying approximation that 50% of the mantlewater is seawater-derived, we require a primordial component withdD < 2160‰. This is compatible with the range of dD (2165‰to þ90‰) spanned by the majority of chondrite bulk analyses23.

Unambiguous noble gas evidence for recycled sea water in MORBis hard to obtain, with evidence for volatile recycling at subductionzones24–27 also ascribed to processes of shallow level or eruption/sampling-related contamination14,28. However, our new data nowenable an assessment of the efficiency of seawater recycling to theconvecting mantle. Estimates for the ratio of fluxes of water into andout of arc settings suggest the subduction flux associated withhydrated minerals and bound to sediment is a factor of 10 lower

Figure 1 | The intersection of the Bravo Dome Xe array with the MORB(mid-ocean-ridge basalt) array defines the Xe mantle composition. In ananalogous way to Ne isotope systematics (Supplementary Informationand ref. 4), Xe data from the Bravo Dome system (filled squares) show atwo-component airþcrust mixture that intersects with the MORBair–mantle mixing line to define the mantle Xe isotopic end-member to be129Xe/130Xe(mantle) ¼ 7.90 ^ 0.14. The MORB–air mixing line is defined bypopping rock data19,38. In contrast, Sheep Mountain data (open triangles)exhibit a mantle–crust mixture that subsequently mixes with air. The greyopen triangle is from awell head that had been non-producing for 7 days andis excluded from the fit to the data. McElmo Dome data are representedby open diamonds. Sheep Mountain and McElmo Dome Xe data arepresented in Supplementary Information. Vertical and horizontal dashedlines are the intersection of theMORB array with the Bravo Dome array andindicate the convecting mantle values. The dashed line through BravoDomedata is the error weighted best fit and the dark grey solid lines represent the1j error envelope. All errors are 1j. Line fitting is by the procedure of ref. 48.

Figure 2 | Non-radiogenic Xe data reveal contribution from a primordialcomponent. a, 126Xe/130Xe versus 124Xe/130Xe; b, 128Xe/130Xe versus124Xe/130Xe. Plotted data are the least air-contaminated sample from eachfield as defined by 40Ar/36Ar. Error bars show measurement uncertainty(1j). Symbols as for Fig. 1. Data from a single gas well5 at BravoDome, SheepMountain andMcElmo Dome are included for comparison (cross symbols).Mixing lines between air and a primordial Xe component are also shown: thesolid line is air–solar mixing and the dashed line is air–Q mixing, whereQ is a planetary Xe component observed in meteorites. Tick marks denote10% primordial contribution. Primordial component data in refs 9, 49.

ARTICLES NATURE|Vol 441|11 May 2006

188


than the flux out of arcs29. One explanation for this discrepancy issubduction of unbound sea water into the melting zone. If all waterfluxing out of arc-related settings is derived from sea water enteringthe subduction zone together with unfractionated noble gases, thisequates to 4.3 £ 1028 atoms 36Ar yr21 available for subduction. Usingthe 36Ar concentration calculated earlier, and approximating themass of the convecting mantle to be that of the whole mantle,the convecting mantle contains 6.0 £ 1036 atoms 36Ar. With the

assumption that subduction rates have been constant over most ofgeological time, this means that a subduction efficiency of only 3%over 4.5 Gyr can account for all the 36Ar in the mantle. The convect-ing mantle value assumes no degassing of the mantle throughoutgeological time. Taking the extent of mantle degassing to be 40%, asinferred from 40Ar in the atmosphere30, increases the subductionefficiency to a maximum of 5%. In either case, if arc-derived water inan unbound form does indeed penetrate into the subduction zone,

Figure 3 | Correlation of measured 20Ne/22Ne, 21Ne/22Ne with130Xe/22Ne. a, The plane of best fit to the data. Note the constancy of130Xe/22Ne in both the groundwater-dominated (20Ne/22Ne < 9.8) andmantle-rich (20Ne/22Ne < 11.85) samples. b, A rotation of the graph to viewthe plane edge-on. The good fit to a plane shows that the elementalabundance and isotopic data can be described by the same three end-member components for all samples. Bravo Dome data only are presented.

Figure 4 | Mantle noble gas isotopes relative to 36Ar normalized to thesolar abundance. The range of values derived for the Bravo Dome system isalmost identical to those of the reference mid-ocean-ridge popping rock2PD43. As both volatile systems are produced by very different fractionatingmechanisms, convergence suggests that neither system is significantlyfractionated and therefore together provide an unambiguous mantle noblegas abundance pattern. This pattern, when compared with air, suggests thatunfractionated air cannot provide the source. The convecting mantleheavy noble gases are clearly most similar to sea water (see text), and thisprovides prima-facie evidence for a seawater origin for these tracers. Errorbars include 1j error of the plane fitted to the data and the uncertainty in theNe isotopic composition of the mantle end-member (fitting software byJ. D. Gilmour, manuscript in preparation).

Table 3 | Xe isotope data

Sample 124Xe/132Xe Error 126Xe/132Xe Error 128Xe/132Xe Error 129Xe/132Xe Error

BD11-B 0.003524 027 0.003297 030 0.07073 019 1.0625 026BD20-B 0.003515 026 0.003277 049 0.07042 016 1.0618 036WBD04-B 0.003517 028 0.003271 048 0.07054 018 1.0625 018BD19-B 0.003506 017 0.003249 018 0.07060 011 1.0170 010BD15-B 0.003517 049 0.003265 053 0.07039 023 1.0333 025BD02-B 0.003535 019 0.003248 016 0.07054 011 0.9942 014BD07-B 0.003563 047 0.003310 036 0.07077 025 1.0309 031BD13-B 0.003553 049 0.003305 046 0.07061 021 1.0135 027BD05-B 0.003507 024 0.003309 027 0.07090 016 0.9993 021BD04-B 0.003522 035 0.003279 033 0.07113 017 1.0214 021WBD01-B 0.003516 022 0.003265 023 0.07061 014 1.0510 018WBD03-B 0.003542 042 0.003295 039 0.07083 021 1.0490 028Data from ref. 5 0.003511 021 0.003293 020 0.07044 010 1.0598 005

Sample 130Xe/132Xe Error 131Xe/132Xe Error 134Xe/132Xe Error 136Xe/132Xe Error

BD11-B 0.14861 038 0.77556 193 0.41162 103 0.36156 089BD20-B 0.14783 052 0.77456 291 0.41142 143 0.36195 126WBD04-B 0.14798 032 0.77568 134 0.41308 078 0.36156 066BD19-B 0.14927 016 0.78076 076 0.40379 040 0.35017 036BD15-B 0.14923 041 0.78161 190 0.40905 102 0.35936 092BD02-B 0.14984 021 0.78354 107 0.40205 055 0.34716 048BD07-B 0.14943 048 0.78079 233 0.40338 123 0.35296 108BD13-B 0.14992 042 0.78184 206 0.40535 108 0.35275 095BD05-B 0.14996 033 0.78240 163 0.40198 085 0.34898 073BD04-B 0.14919 032 0.77928 161 0.40814 086 0.35580 074WBD01-B 0.14809 028 0.77369 135 0.41271 073 0.36136 064WBD03-B 0.14892 045 0.78153 213 0.40997 113 0.36093 099Data from ref. 5 0.14838 014 0.77549 041 0.41359 027 0.36279 025

Errors are last three digits and 1j. Included for comparison is previous work5.


189


accounting for the backarc water flux, ,5% of that water is requiredto progress to the deeper convecting mantle.

Recycling of sea water into the OIB source

There is clear evidence from radiogenic isotope geochemistry thatrecycled material forms a significant component of OIB material31.Similarly, numerical models of whole mantle convection, which donot invoke a physically unobservable chemically derived deep densitycontrast, show that a significant portion of subducted materialreaches the core–mantle boundary, and from this boundary layercontributes to the upwelling plumes32. It is therefore difficult toenvisage a mechanism of seawater recycling into the convectingmantle that does not also have a major impact on the OIB source.Although early OIB noble gas isotope data gave results little differentto atmospheric Ar and Xe, these have since been shown to be theresult of atmospheric contamination33. The 40Ar/36Ar ratio ofthe mantle sourcing plumes is still ambiguous but, nevertheless,the 40Ar/36Ar upper limit appears to be ,10,000, with the Icelandmaximum being 4,530 (ref. 34), the Juan Fernandez maximum being8,000 (ref. 35), the Loihi maximum being 8,300 (ref. 34), and theReunion maximum being 7,600 (ref. 11). One exception is Samoa,with a 40Ar/36Ar maximum of 15,000 (ref. 35), but this is probablydue to a metasomatic overprint from crustal fluids36. In contrast, theconvecting mantle has a far higher proportional contribution fromthe radiogenic 40Ar isotope, with the 40Ar/36Ar ratio in the convectingmantle unequivocally resolved to be 35,000–52,500 (ref. 4) andrefined with the larger number of mantle-rich samples in thisstudy to 41,050 ^ 2,670 (Supplementary Information).

A similar difference between OIB and MORB is observed in Xeisotopes. Xe isotope ratios in excesses of air in OIB are only a recentdiscovery, owing to the high precision required to resolve such asmall excess, with 129Xe/130Xe ,7 (refs 34, 37) compared with the airvalue of 6.49. The convecting mantle value defined by this data set(129Xe/130Xe ¼ 7.90 ^ 0.14) compares with the highest measuredvalues in MORB popping rock (129Xe/130Xe ¼ 7.5 (ref. 38) and129Xe/130Xe ¼ 7.73 (ref. 19)) and from the South Mid-AtlanticRidge (129Xe/130Xe ¼ 7.78 (ref. 27)). These observations of low40Ar/36Ar and 129Xe/130Xe isotopic ratios in OIB relative to MORB,combined with the association of OIB with high 3He/4He ratios, havebeen used to argue that the OIB source is undegassed and therefore avolatile-rich reservoir, preserved since accretion30. We argue thatrecycling of unfractionated sea water into the OIB source is respon-sible for the reduction in 40Ar/36Ar and 129Xe/130Xe ratios towardsair-like values. Deuterium measurements from hotspots are variable,with dD values higher than MORB ranging from 236‰ over theSalas y Gomez plume39 to 250‰ near Iceland40. In contrast, lowerthan MORB values of dD ¼ 2118‰ are observed at Loihi, Hawaii41.In the context of recycling, data with dD lower than MORB reflect amore primitive hydrogen signature present in the lower mantle, anddD higher than MORB is a signature of a proportionally largerrecycled component. A more significant recycled seawater contri-bution to the Iceland source is completely consistent with the lowermaximum 40Ar/36Ar ratios at Iceland (,4,000) compared to Hawaii(,8,000) and a relatively dry source for the Hawaiian plume42.

Recycling dominates the complement of heavy noble gases in theconvecting mantle. This clearly precludes 36Ar from being used as a

simple constraint on primitive volatile outgassing from the mantle30.This does not, however, account for the high 3He/4He ratio andhigher solar Ne content of many OIBs. It has been shown by severalworkers that these primitive volatile signatures in OIB can be simplyaccounted for by the addition of a small volatile-rich component43,44.We speculate that this is associated with OIB during the recyclingprocess by sampling a volatile-rich reservoir preserved at the core–mantle boundary, perhaps at the D 00 layer45. Alternatively, it has beensuggested that the recycling process is not efficient at degassingmantle volatiles, and that the recycled material itself may preservehigher 3He/4He ratios46. What is clear is that our explanation for thedichotomy between OIB and MORB heavy noble gas isotope signa-tures now provides a quantitative constraint on numerical simu-lations of mantle convection47, which in turn may provide a solutionto the mechanism of preservation of ancient accretionary volatilesignatures in our planet.

Received 11 October 2005; accepted 23 March 2006.

1. Lupton, J. E. & Craig, H. Excess 3He in oceanic basalts, evidence for terrestrialprimordial helium. Earth Planet. Sci. Lett. 26, 133–-139 (1975).

2. Honda, M., Reynolds, J. H., Roedder, E. & Epstein, S. Noble gases in diamonds:occurrences of solar like helium and neon. J. Geophys. Res. 92, 12507–-12521(1987).

3. Sarda, P., Staudacher, T. & Allegre, C. J. Neon isotopes in submarine basalts.Earth Planet. Sci. Lett. 91, 73–-88 (1988).

4. Ballentine, C. J., Sherwood Lollar, B., Marty, B. & Cassidy, M. Neon isotopesconstrain convection and volatile origin in the Earth’s mantle. Nature 433,33–-38 (2005).

5. Caffee, M. W. et al. Primordial noble gases from Earth’s mantle: identificationof a primitive volatile component. Science 285, 2115–-2118 (1999).

6. Staudacher, T. & Allegre, C. J. Recycling of oceanic crust and sediments: thenoble gas subduction barrier. Earth Planet. Sci. Lett. 89, 173–-183 (1988).

7. Kent, A. J. R. et al. Widespread assimilation of a sea-water-derived componentat Loihi Seamount, Hawaii. Geochim. Cosmochim. Acta 63, 2749–-2761 (1999).

8. Kent, A. J. R., Norman, M. D., Hutcheon, I. D. & Stolper, E. M. Assimilation ofseawater-derived components in an oceanic volcano: evidence from matrixglasses and glass inclusions from Loihi seamount, Hawaii. Chem. Geol. 156,299–-319 (1999).

9. Pepin, R. O. & Porcelli, D. Origin of noble gases in the terrestrial planets. Rev.Mineral. Geochem. 47, 191–-246 (2002).

10. Kunz, J. Is there solar argon in the Earth’s mantle? Nature 399, 649–-650 (1999).11. Trieloff, M., Kunz, J. & Allegre, C. J. Noble gas systematics of the Reunion

mantle plume source and the origin of primordial noble gases in Earth’s mantle.Earth Planet. Sci. Lett. 200, 297–-313 (2002).

12. Porcelli, D. & Wasserburg, G. J. Mass transfer of xenon through a steady-stateupper mantle. Geochim. Cosmochim. Acta 59, 1991–-2007 (1995).

13. Porcelli, D. & Wasserburg, G. J. Mass transfer of helium, neon, argon, andxenon through a steady-state upper mantle. Geochim. Cosmochim. Acta 59,4921–-4937 (1995).

14. Ballentine, C. J. & Barfod, D. N. The origin of air-like noble gases in MORB andOIB. Earth Planet. Sci. Lett. 180, 39–-48 (2000).

15. Harrison, D., Burnard, P. G., Trieloff, M. & Turner, G. Resolving atmosphericcontaminants in mantle noble gas analyses. Geochem. Geophys. Geosyst. 4, art.no. 1023, doi:10.1029/2002GC000325 (2003).

16. Sarda, P. & Guillot, B. Breaking of Henry’s law for noble gas and CO2 solubilityin silicate melt under pressure. Nature 436, 95–-98 (2005).

17. Sarda, P. & Graham, D. Mid-ocean ridge popping rocks: implications fordegassing at ridge crests. Earth Planet. Sci. Lett. 97, 268–-289 (1990).

18. Burnard, P., Graham, D. & Turner, G. Vesicle specific noble gas analyses ofpopping rock: implications for primordial noble gases in Earth. Science 276,568–-571 (1997).

19. Moreira, M., Kunz, J. & Allegre, C. J. Rare gas systematics in popping rock:isotopic and elemental compositions in the upper mantle. Science 279,1178–-1181 (1998).

Table 4 | Resolved mantle elemental ratios3He/36Ar 22Ne/36Ar 84Kr/36Ar 130Xe/36Ar

Bravo Dome 0.302 (0.024) 0.112 (0.028) 0.0556 (0.0088) 0.001045 (0.00009)Solar wind 11.92 3.846 0.0005 0.0000125CI chondrite 0.021 0.0153 0.00700Air 0.053 0.0207 0.00011Sea water 0.015 0.0410 0.000563Popping rock 0.79 0.161 0.0564 0.00103

Popping rock data from ref. 19. Errors in parentheses are 1j.

ARTICLES NATURE|Vol 441|11 May 2006

190


20. Matsuda, J. & Nagao, K. Noble gas abundances in a deep-sea sediment corefrom eastern equatorial Pacific. Geochem. J. 20, 71–-80 (1986).

21. Thompson, A. B. Water in the Earth’s upper mantle. Nature 358, 295–-302(1992).

22. Symons, K., Dixon, J., Schilling, J. G., Kingsley, R. & Poreda, R. Volatiles inbasaltic glasses from the Easter–-Salas y Gomez Seamount Chain and EasterMicroplate: implications for geochemical cycling of volatile elements. Geochem.Geophys. Geosyst. 3, doi:10.1029/2001GC000173 (2002).

23. Robert, F., Gautier, D. & Dubrulle, B. The solar system D/H ratio: observationsand theories. Space Sci. Rev. 92, 201–-224 (2000).

24. Matsumoto, T., Chen, Y. L. & Matsuda, J. Concomitant occurrence ofprimordial and recycled noble gases in the Earth. Earth Planet. Sci. Lett. 188,211–-219 (2001).

25. Sarda, P., Moreira, M. & Staudacher, T. Argon–-lead isotopic correlation inMid-Atlantic ridge basalts. Science 283, 666–-668 (1999).

26. Bach, W. & Niedermann, S. Atmospheric noble gases in volcanic glasses fromthe southern Lau basin: origin from the subducting slab? Earth Planet. Sci. Lett.160, 297–-309 (1998).

27. Sarda, P., Moreira, M., Staudacher, T., Schilling, J. G. & Allegre, C. J. Rare gassystematics on the southernmost Mid-Atlantic Ridge: constraints on the lowermantle and the Dupal source. J. Geophys. Res. Solid Earth 105, 5973–-5996(2000).

28. Hilton, D. R., Hammerschmidt, K., Loock, G. & Friedrichsen, H. Helium andargon isotope systematics of the central Lau Basin and Valu Fa ridge: evidenceof crust–-mantle interactions in a back-arc basin. Geochim. Cosmochim. Acta 57,2819–-2841 (1993).

29. Hilton, D. R., Fischer, T. P. & Marty, B. Noble gases and volatile recycling atsubduction zones. Rev. Mineral. Geochem. 47, 319–-370 (2002).

30. Allegre, C. J., Staudacher, T. & Sarda, P. Rare gas systematics: formation of theatmosphere, evolution and structure of the Earth’s mantle. Earth Planet. Sci.Lett. 81, 127–-150 (1987).

31. Hofmann, A. W. Mantle geochemistry: the message from oceanic volcanism.Nature 385, 219–-229 (1997).

32. Van Keken, P. E., Ballentine, C. J. & Hauri, E. in Treatise on Geochemistry: TheMantle and Core (ed. Carlson, R. W.) 471–-491 (Elsevier Pergamon, Oxford, 2003).

33. Patterson, D. B., Honda, M. & McDougall, I. Atmospheric contamination—apossible source for heavy noble gases in basalts from Loihi Seamount, Hawaii.Geophys. Res. Lett. 17, 705–-708 (1990).

34. Trieloff, M., Kunz, J., Clague, D. A., Harrison, D. & Allegre, C. J. The nature ofpristine noble gases in mantle plumes. Science 288, 1036–-1038 (2000).

35. Farley, K. A. & Craig, H. Atmospheric argon contamination of ocean islandbasalt olivine phenocrysts. Geochim. Cosmochim. Acta 58, 2509–-2517 (1994).

36. Burnard, P. G., Farley, K. A. & Turner, G. Multiple fluid pulses in a Samoanharzburgite. Chem. Geol. 147, 99–-114 (1998).

37. Harrison, D., Burnard, P. & Turner, G. Noble gas behaviour and composition inthe mantle: constraints from the Iceland Plume. Earth Planet. Sci. Lett. 171,199–-207 (1999).

38. Kunz, J., Staudacher, T. & Allegre, C. J. Plutonium-fission xenon found in Earth’smantle. Science 280, 877–-880 (1998).

39. Kingsley, R. H. et al. D/H ratios in basalt glasses from the Salas y Gomezmantle plume interacting with the East Pacific Rise: water from old D-richrecycled crust or primordial water from the lower mantle? Geochem. Geophys.Geosyst. 3, doi:10.1029/2001GC000199 (2002).

40. Poreda, R., Schilling, J. G. & Craig, H. Helium and hydrogen isotopes in ocean-ridge basalts north and south of Iceland. Earth Planet. Sci. Lett. 78, 1–-17 (1986).

41. Hauri, E. SIMS analysis of volatiles in silicate glasses, 2: isotopes andabundances in Hawaiian melt inclusions. Chem. Geol. 183, 115–-141 (2002).

42. Dixon, J. E. & Clague, D. A. Volatiles in basaltic glasses from Loihi Seamount,Hawaii: evidence for a relatively dry plume component. J. Petrol. 42, 627–-654(2001).

43. Ballentine, C. J., Van Keken, P. E., Porcelli, D. & Hauri, E. H. Numerical models,geochemistry and the zero-paradox noble-gas mantle. Phil. Trans. R. Soc. Lond.A 360, 2611–-2631 (2002).

44. Ellam, R. M. & Stuart, F. M. Coherent He-Nd-Sr isotope trends in high 3He/4Hebasalts: implications for a common reservoir, mantle heterogeneity andconvection. Earth Planet. Sci. Lett. 228, 511–-523 (2004).

45. Tolstikhin, I. & Hofmann, A. W. Early crust on top of the Earth’s core. Phys.Earth. Planet. Inter. 148, 109–-130 (2004).

46. Class, C. & Goldstein, S. L. Evolution of helium isotopes in the Earth’s mantle.Nature 436, 1107–-1112 (2005).

47. Ballentine, C. J., Van Keken, P. E., Holland, G., Hauri, E. H. & Brandenburg, J. P.Recycling volatiles and attaining a geochemical and fluid dynamicallyconsistent model of mantle convection. Eos 86 (52), Fall Meet. Suppl., Abstr.V53F-01 (2005); at khttp://www.agu.org/meetings/fm05/fm05-sessions/fm05_V53F.htmll (2005).

48. York, D. Least squares fitting of a straight line with correlated errors. EarthPlanet. Sci. Lett. 5, 320–-324 (1969).

49. Ott, U. Noble gases in meteorites—trapped components. Rev. Mineral.Geochem. 47, 71–-100 (2002).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank Oxy for permission to sample the Bravo DomeField, Amerada Hess for permission to sample West Bravo Dome, andM. Cassidy for logistics and field assistance. We thank D. Blagburn andB. Clementson for laboratory support. We also thank J. Gilmour for provision ofthe error weighted plane fitting software, S. Gilfillan for air calibration work andD. Hilton for comments on the manuscript. This work was funded by NERC.

Author Information Reprints and permissions information is available atnpg.nature.com/reprintsandpermissions. The authors declare no competingfinancial interests. Correspondence and requests for materials should beaddressed to G.H. ([email protected]).


191

seawater subduction controls the heavy noble gas composition of the mantle

Documents