An evaluation of the C/N ratio of the mantle from natural CO2-rich gas analysis: Geochemical and cosmochemical implications

HAL Id: hal-03012907

https://hal.archives-ouvertes.fr/hal-03012907

Submitted on 18 Nov 2020

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Distributed under a Creative Commons Attribution - NonCommercial| 4.0 International

License

An evaluation of the C/N ratio of the mantle from

natural CO2-rich gas analysis: Geochemical and

cosmochemical implications

Bernard Marty, Matthieu Almayrac, Peter Barry, David Bekaert, Michael

Broadley, David Byrne, Christopher Ballentine, Antonio Caracausi

To cite this version:

Bernard Marty, Matthieu Almayrac, Peter Barry, David Bekaert, Michael Broadley, et al..

An

evaluation of the C/N ratio of the mantle from natural CO2-rich gas analysis: Geochemical and

cosmochemical implications. Earth and Planetary Science Letters, Elsevier, 2020, 551, pp.116574.

�10.1016/j.epsl.2020.116574�. �hal-03012907�

Contents lists available atScienceDirect

Earth

and

Planetary

Science

Letters

www.elsevier.com/locate/epsl

An

evaluation

of

the

C/N

ratio

of

the

mantle

from

natural

CO

₂

-rich

gas

analysis:

Geochemical

and

cosmochemical

implications

Bernard Marty

a

,

∗

,

Matthieu Almayrac

a

,

Peter H. Barry

b

,

David V. Bekaert

a

,

b

,

Michael W. Broadley

a

,

David J. Byrne

a

,

Christopher J. Ballentine

c

,

Antonio Caracausi

d

a_{UniversitédeLorraine,CNRS,CRPG,F-54000Nancy,France}

b_{MarineChemistryandGeochemistryDept.,WoodsHoleOceanographicInstitution,WoodsHole,MA02543,USA} c_{DepartmentofEarthSciences,UniversityofOxford,OX13ANOxford,UnitedKingdom}

d_{InstitutoNazionalediGeofisicaeVulcanologia,SezionediPalermo,90146Palermo,Italy}

a

r

t

i

c

l

e

i

n

f

o

a

b

s

t

r

a

c

t

Articlehistory:

Received22June2020

Receivedinrevisedform3September2020 Accepted9September2020

Availableonlinexxxx Editor:F.Moynier

Keywords:

carbonnitrogenEarthmantlegases

The terrestrialcarbon to nitrogenratio is akey geochemicalparameter that can provideinformation on the nature of Earth’s precursors, accretion/differentiation processes of our planet, as well as on the volatile budget of Earth. In principle, thisratio can be determined from the analysis of volatile elementstrappedinmantle-derivedrockslikemid-oceanridgebasalts(MORB),correctedforfractional degassing during eruption. However, this correction is critical and previous attempts have adopted differentapproacheswhichledtocontrastingC /N estimatesforthebulksilicateEarth(BSE)(Martyand Zimmermann,1999;Berginetal.,2015).HereweconsidertheanalysisofCO2-richgasesworldwidefor

whichamantleoriginhasbeendeterminedusingnoblegasisotopesinordertoevaluatetheC /N ratioof themantlesourceregions.Thesegasesexperiencedlittlefractionationduetodegassing,asindicatedby radiogenic4He/40Ar∗values(where4Heand40Ar*areproducedbythedecayofU+Th,and40Kisotopes, respectively)close tothe mantleproduction/accumulationvalues.The C /N and C/3_{He ratiosofgases}

investigatedherearewithintherangeofvaluespreviouslyobservedinoceanicbasalts.Theypointtoan elevatedmantleC /N ratio(∼350-470,molar)higherthanthoseofpotentialcosmochemicalaccretionary endmembers.Forexample,theBSE C /N and36_{Ar/N ratios(160-220and 75×10}−7_{,respectively) are}

higherthanthoseofCM-CIchondritesbutwithinthe rangeofCV-COgroups. Thissimilaritysuggests thattheEarthaccretedfromevolvedplanetaryprecursors depletedinvolatileandmoderately volatile elements.HencethehighC/N compositionoftheBSEmaybeaninheritedfeatureratherthantheresult ofterrestrialdifferentiation. TheC/N and36_{Ar/N ratios ofthesurface(atmospherepluscrust)and of}

themantlecannotbeeasilylinkedtoanyknownchondriticcomposition.However,thesecompositions areconsistentwithearlysequestrationofcarbonintothemantle(butnotNandnoblegases),permitting theestablishmentofclementtemperaturesatthesurfaceofourplanet.

©2020TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Carbonandnitrogenareessentialingredients tobuilda habit-ableEarth-likeplanet(Alexanderetal., 2012;Marty,2012;Bergin etal.,2015;JohnsonandGoldblatt,2015;Hirschmann,2018; Gre-wal et al., 2020). Together with water, these elements were de-liveredtoEarthbyaccretionarymaterials mainlyoriginatingfrom theinnersolarsystem,whoseremnantscanbefoundinprimitive meteorites(chondrites).Stableisotoperatiospointtoan asteroidal-likeoriginforterrestrialwater,carbonandnitrogen(Marty,2012;

*

Correspondingauthor.

E-mailaddress:[email protected](B. Marty).

Alexander et al., 2012), whereas other potential cosmochemical sourcessuchascometarymaterialwerelikelytobeminor(Bekaert et al., 2020). Of particular interest is C, one of the main green-houseforcing elements inthe atmosphere. Its present-day abun-danceatEarth’s surface is reasonablywell known fromthe geo-logicalrecord (27

±

5 ppmw,normalizedto themass oftheBSE, 4.0

×

1027 g; Hirschmann, 2018), but the abundance of carbon in the other major terrestrial reservoirs (e.g., the core and the mantle) is not well-constrained and somewhat model-dependent (Dalou et al., 2017; Hirschmann, 2018; Grewal et al., 2020). For example,C isoften advocated to be an important trace element in the core, but quantitative estimates of its concentration de-pendon several poorly constrainedparameters such aspressure, composition-dependentmetal-silicatepartitioningcoefficientsand https://doi.org/10.1016/j.epsl.2020.116574

0012-821X/©2020TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

B. Marty, M. Almayrac, P.H. Barry et al. Earth and Planetary Science Letters 551 (2020) 116574

chronology of core formation (e.g., Dalou et al., 2017; Li et al., 2020).As both carbonandnitrogenare incompatibleduring par-tialmeltingunderpresent-daymantleoxygenfugacityconditions, theircontentsinthemantle-derivedlavascanthus,inprinciple,be used toestimate thecomposition ofthemantlesource.However, CO2,themaincarbon-bearingspeciesinpresent-daymagmas,has

alowsolubilityinbasalticmeltsandlavasreachingEarth’ssurface are readily degassed, includingthose erupted at highhydrostatic pressure andrapidly quenchedon the ocean floor. Some authors have considered oceanic basalt glasses with minimally-degassed basalts (Hirschmann, 2018) to estimate the ratio of CO2 to

non-volatileelementswithknowngeochemicalbehaviorssuchasBaor Nd (Rosenthalet al., 2015), for whichthe mantle abundanceare known. Assuming that the CO2/Baratio (100

±

20 ppmw/ppmw;

(Hirschmann,2018)isrepresentativeofthemantlesourcesof ana-lyzedlavas,andtakingaBaconcentrationoftheconvectivemantle to be 4.0

±

0.4 ppmw (Palme and O’Neill, 2013), the concentra-tion ofC in themantle was henceestimated at110

±

40 ppmw forparentreservoirs(Hirschmann,2018).VolcanismattheEarth’s surface is volumetrically dominated by mid ocean ridge (MOR) magmatism (

∼

16 km3/yr;Le Voyeretal., 2019) whichissourced directly fromthe convectingdepleted MORB mantle(DMM).The DMM carbonconcentrationcan be estimatedindependentlyfrom the total flux of CO2 from ridges, the magma production rate,

andthe average MORmelting rate(

∼

12%). MOR CO2 flux values

computedfollowingdifferentapproachesyieldcomparable values:

[

1.3

±

0.8

]

×

1012_{mol/yrfromMORdatacompilation(LeVoyeret}

al., 2019),

[

2.2

±

0.9

]

×

1012 mol/yr (MartyandTolstikhin, 1998) and

[

1.6

±

0.2

]

×

1012_{mol/yr(Tuckeretal.,2018)fromcalibration}

to3_{He.ThecorrespondingMORmantleCcontentwouldbeabout}

20-30ppmw(Tuckeretal.,2018),significantlylowerthantheone obtainedfromtheBacalibration.CurrentconstraintsontheC con-tentofthemantlethereforeremainuncertain,especiallyifregions thatarenotinvolvedinmantleconvection(e.g.,theplumemantle source) are significant repositories for C, as independently sug-gestedbynoblegasisotopes(e.g.,MukhopadhyayandParai,2019). Nitrogen in the atmosphere-mantle system can be estimated onaglobalscale.N2 andradiogenic40Arabundancescorrelatein

MORBglassesandoceanislandbasalt(OIB)glassesfromdifferent localities worldwide (Marty and Zimmermann, 1999; Marty and Dauphas,2003).Consequently,theNcontentofthemantlecanbe estimatedfromthatof40_{Arproducedbythedecayof}40_Kbecause

the potassium content of the BSE is reasonably well estimated (280

±

60 ppm K,Arevalo et al., 2009). Terrestrial 40Ar has been produced by the radioactive decayof 40K inthe BSE, andabout half(42+₋32₁₈%)ofradiogenic40_Ar(noted40_{Ar*)isnowinthe}

atmo-sphereasaresultofterrestrialdegassing(Allègreetal.,1996).The surfacereservoirs(atmosphere:0.97ppmw N,sedimentsandcrust) canaccountforabout1.56

±

0.06 ppmw N(JohnsonandGoldblatt, 2015; Hirschmann, 2018; after correction in error propagation). Using a mass balance approach,the mantleN concentration was estimatedtobearound1ppmw (Marty,2012) andmorerecently tobe1.10

±

0.55 ppmw(Hirschmann,2018).Availableconstraints on the global inventory of N in the Earth’s mantle and surface constitute a solid basis for estimatingthe volatilecontent ofthe BSE(Marty,2012;JohnsonandGoldblatt,2015;Berginetal.,2015; Hirschmann,2018).

By combiningthe mantlecarbon concentration obtainedfrom calibrationtoBatogether withtheabovemantleNconcentration, Berginetal.(2015) estimateda mantleC

/

N ratioof

∼

90(in the following elemental ratios are molar and written initalics while C andN concentrations are in ppmw of the BSE having a mass of 4

×

1027 g).Using their estimated surface C

/

N ratio

(21

±

6), Bergin et al. proposed a BSE C

/

N ratio of 61

±

23. In contrast, Marty (2012) estimated a significantly higher BSE C

/

N ratio of 365

±

233 from the analysis of CO2 andN2 in MORBsand OIBs

(Marty andZimmermann, 1999; Marty andDauphas, 2003). The implicationsarenot trivial,asahighC

/

N ratiooftheBSE would implya highmantleC concentrationanddrasticvolatile fraction-ationeitherintheplanetesimalsthatformedourplanetorduring Earth’sformation.

The bulk mantle C

/

N estimate of Marty (2012) is based on theC-N-noblegases composition ofa suiteof47 MORB andOIB samplesextractedfollowingvacuumcrushingtoremovesecondary contributions within basaltic glasses. Gases trapped in oceanic glassesalways constitute a residualfraction ofthe initialvolatile content due to pre-eruption and syn-eruption degassing. Hence residual volatiles are elementally fractionated according to their solubilitiesinbasalticmelts.ThesolubilitiesofN2andArare

com-parable(KN2

∼

KAr

∼

2-3

×

10−12mol/g.hPa;Jambonetal.,1985;

Libourel et al., 2003), and their degassing fractionation is mini-mal.However, CO2 and He are

∼

3and

∼

10 times moresoluble

thanN2 orArrespectively(KCO2

=

9

×

10−12mol/g.hPa,Dixon et

al., 1995; KHe

=

2.5

×

10−11 mol/g.hPa,Jambon etal.,1985).

Ac-cordingly,He andCO2 willbe depleted (relativetoN2-Ar)inthe

firstfractionsofgasescapingfromamagmaandenrichedin resid-ualgastrappedineruptedlavas.Onewaytocorrectfordegassing fractionationistousetheradiogenic4_He/40_{Ar*ratio(where}40_Ar*

is theamount of 40_{Arcorrected for atmospheric contamination).}

Itsproduction/accumulationratiointhemantleisestimatedfrom the parent (U

+

T h

)/

K abundance ratio to be

∼

2 (range: 1.8-3.0; e.g., Marty, 2012; Mukhopadhyay and Parai, 2019). Oceanic basaltglassespresentsystematicallyhighervaluesintherange 3-96 (Marty and Zimmermann, 1999), demonstrating the effect of elementalfractionationthatcouldbemodeledfollowinga Raleigh-type distillation process. The initial

(

C

/

N

)

i ratio of the mantle

sourcecouldbecomputedfromthemeasured

(

C

/

N

)

meas ratio

ac-cordingtoEqn. (1):

(

C

/

N

)

i

= (

C

/

N

)

m

×



₄ He

/

40Ar



_m



₄ He

/

40_Ar



i



⎡ ⎣  K−1 N2−K−C O 21 K−_He1−K−_Ar1  ⎤ ⎦ (1) whereKjarethebasalticmelt/gaspartitioncoefficientsforspecies

j and suffixes i and m mean initial and measured, respectively. After correction, the mean MORB (C/N) ratio was computed to

be 535

±

224 (Marty and Zimmermann, 1999), with E-MORBs

((

C

/

N

)

=

1839

±

641) being richer in C (relative to N) than N-MORBs

((

C

/

N

)

=

273

±

106).Thisstudyindicatedthat(i)theC

/

N

ratiooftheMORBmantleismuchhigherthanthatofthesurface inventory

((

C

/

N

)

surface

=

21

±

6,computedwithabundancesgiven

inHirschmann,2018),and(ii)theCcontent ofmantlesourcesis highlyvariable,possiblyduetopreferential subductionofC com-paredtoN(MartyandZimmermann,1999).

Bergin et al. (2015) questioned this approach by noting that the extent of correction for largely variable C/N ratios in MORB complicatesthereconstructionoftheC

/

N ratioofmantlesources (Berginetal.,2015).NaturalCO2-richgasesforwhichisotopic

ra-tios of C, N and noble gases demonstrate a mantle origin may provideanindependentmeanstoinvestigatetheC

/

N ratioofthe mantle.ContinentalCO2-rich gasestap intolargecrustal volumes

wheremantle-derived gases haveaccumulated over long periods oftime (BallentineandHolland, 2008), making thempresumably morerepresentativeoftheoriginalmantlecompositionthan resid-ualgasesleft overinoceanicbasalticglasses.Atmospheric and/or crustal contamination can be problematic for natural gases, yet thenoblegases alongwithCandNcanprovidesensitive toolsto assesstheseeffectsandidentifytheoriginalmantlesource compo-sition.HerewehaveselectedCO2-richgasesfromdifferent

geody-namicenvironmentswheretheseeffectscanberobustlycorrected

for, to thoroughly estimate the C

/

N ratios of the corresponding magmasourcesandprovidenewinsightsintotheC

/

N ratioofthe BSE.

2. Samples

WehaveselectedCO2-richgasesforwhichthe40Ar

/

36Ar ratios

are markedly higherthan the atmosphericvalue of 298.6. Other noble gas constraints were used as additional sample selection criteria, namelythose with(i) mantle-derivedhelium,richin pri-mordial3_{He,(ii)mantle-derivedneonwith3-isotopecompositions}

lying along the MORB or the mantleplume correlation, and(iii) excess129_{Xefromtheradioactivedecayofextinct} 129_I(asmantle

Xeisenrichedin129XewithrespecttoatmosphericXe) (Table1). Samples having all thesefeatures are not common andasfar as weknowarerestrictedtothelistpresentedinTable1.

We includeCO2-rich gases fromcontinentalEurope.Bräueret

al.(2013) carriedoutasystematicsurveyofCO2-richgasesinthe

Eifelquaternaryvolcanicprovince(Germany),andfoundtwo loca-tions (VictoriaquelleandSchwefelquelle)withelevated40_Ar

_/

36_Ar

ratiosintherange1,500-3,000.Repeatedsamplingofthesesprings by theCRPGgroup identified evenhigher 40Ar

/

36Ar ratiosup to 8,300, 129Xe excesses up to 7% relative to atmospheric Xe and MORB-like neon (Bekaert et al., 2019). Nitrogen isotopologues of thesesamples indicate undetectable atmosphericN2 contribution

(Labidietal.,2020).WealsoincludedataforaCO2-rich gas

sam-pledatBublak,Egergraben,CzechRepublic(Bräueretal.,2004). In2018,weundertookasurveyofgeothermalgasesinthe Yel-lowstoneNationalPark(USA),whichislocatedatthepresent-day apexofahigh-3_{Hemantleplume(Craigetal.,1978).Someofthe}

gases that were collected show 40_Ar

_/

36_{Ar ratios higherthan the}

atmospheric value (Broadley et al., 2020) and their data are re-portedhereinthreegroups(Table1):(i)theMud Volcanogroup locatedinsidetheYellowstonecalderawherethehighest3He/4He

values (

≥

14 Ra, typical of mantle plume He signature) are ob-served(Craigetal.,1978);(ii)theTurbidLakegrouplocatedatthe borderofthecaldera(3_He

_/

4_He

₌

_{5.2 Ra),and(iii)theBrimstone}

Basin group outside the caldera (3_He

_/

4_He

₌

_{2.8 Ra). He isotope}

variations are consistent withdilution ofa typical mantleplume signaturebyradiogenicHefromthePrecambrianbasement. Brim-stoneBasinsamplesalsohaveXeandNeisotopesconsistentwith a crustal contributionto plume-like noblegases (Broadley etal., 2020). Nitrogen isotopologues ofYellowstone gases also point to a mantleplume originfornitrogen, withlimitedcontributions of atmosphericorcrustalN(Labidietal.,2020).

The Naftia sample fromSicily is a CO2 well gas inthe

vicin-ityofMountEtnawhichshowsan excessof129XeandMORB-like Ne (Nakaietal., 1997). At thissite,the 3He/4He ratio varies be-tween 6 and 7 Ra and correlates with the eruptive activity at Mt.Etna, suggestingthatnewinjectionsofmagmaunderthe vol-canoperiodicallycontribute freshmantlegasestothisCO2

reser-voir(Caracausietal., 2003).Volcanic gases werealso sampledat OldoinyoLengai,anactivecarbonatiticvolcanolocatedonthe east-ern branch of the East African Rift Valley (Fischer et al., 2009). Atmospheric contamination is low, asindicated by 40Ar/36Ar

ra-tiosupto948(OLD-2sample),whichisamongthehighestvalues evermeasuredforvolcanicgases. The3_He/4_{Heratio(6.9Ra)isin}

thelowrangeofMORBvalues,andneonhasaMORB-likeisotopic composition. Fischer etal.(2009) interpreted the C,N andnoble gascompositionsasreflectingdegassingofanewbatchofmagma fromaMORB-likesource.A4He/40Ar*ratioof0.4(Table 2)lower than the mantle range (1.8-3.0) is consistent with this interpre-tationbecause lesssolubleargon insilicate meltstends todegas more readilythan moresoluble heliumduring the first stagesof magmadegassing.

Gasesextractedfromtwowell-studied,gas-rich,basalticglasses are also included in this compilation for the purposes of com-parison. 2D43 is a MORB glass fromthe Mid-Atlantic Ridgeat 14◦N (Javoy and Pineau, 1991; Moreira et al., 1998). This sam-ple is considered to best represent the composition of a MORB gas phase. DICE corresponds to a subglacial glass sampled in Iceland (N64◦10’31.9”/W021◦02’43.0”) which is also a reference sample that is considered representative of unfractionated man-tleplume-sourced gas (Mukhopadhyay,2012).Both sampleshave

4_He/40_{Ar* close to, orwithin the range of}

accumulation/produc-tionmantlevalues,indicating limited, ifany,degassing fractiona-tion.

3. Datahandling

Inordertodeterminethemantlesourcecompositions,gasdata need to be corrected foratmospheric contributionand fractional degassingofmagmas,respectively,whichweexplainbelow.

3.1. Correctionforatmosphericcontamination

Thecorrection foratmospheric contamination isnegligible for CO2 givenitslow atmosphericconcentration (440ppmv)butcan

be important for N2 and Ar which are abundant in air (78.08

vol% and0.93 vol%, respectively; Table 1). Atmospheric contami-nationofnaturalgases isalways criticalfornon-radiogenicnoble gases(e.g.,36Ar)andcanoverprint theoriginal composition.This isless ofa problemfor 40_{Ar becausethis isotope ismore}

abun-dantinthemantlethanprimordial36_{Arrelativetotheatmosphere}

(40Ar/36Ar

=

298.6 in air versus

≥

10,000-40,000 in the mantle, seebelow).Correctionforatmosphericcontamination is achieved usingmeasured40_Ar/36_{Ar ratios:}

40_Ar ca

=

40_Ar meas 1

+

α

(2) with:

α

=



₃₆ Ar 40_Ar

ca

=

₃₆ Ar 40_Ar



meas

−

₃₆ Ar 40_Ar



atm

₃₆ Ar 40_Ar



atm

−

₃₆ Ar 40_Ar



mantle (3)

Wheresuffixesca,meas,atm andmantle referto40_Ar/36_{Ar values}

corrected for atmospheric contamination, measured in the sam-ple,intheatmosphereandthemantleend-member, respectively. For the latter, we chose 40,000 and 10,000 for convective man-tle andmantle plume sources, respectively (Bekaert et al., 2019; Mukhopadhyay, 2012). Taking 27,000 as the mantle source end-memberof2D43(Moreiraetal.,1998)wouldhaveaminoreffect onthecorrection.The correctionsforthe 40_{Arconcentrationsare}

inmostcases marginal(within0.1-20%),exceptfortheMud Vol-cano and Turbid Lake samples (65-71%) and for the Bublak gas sample(54%).

ForN2 whichis also lessaffectedby atmospheric

contamina-tionthan primordial noblegases ((N2

/

36_Ar

₎

₌

_1.5-3

_×

₁₀4 _{in air}

andair-saturatedwater,and

≥

106 inthemantle), thecorrection isachieved by using the N2

/

40Ar ratio,with corrected N2 being

computedusingthecorrected40_Ar

CAconcentration:



N2 40_Ar

ca

=



N2 40_Ar

meas

+

α

×



N2 40_Ar

meas

−



N2 40_Ar

atm



(4) N2,ca

=



N2 40_Ar

ca

×

40_Ar ca (5)

B. Marty ,M. Almayr ac, P.H. B arry et al. Earth a nd Plane tary Science Le tt ers 551 (2020) 116574 Table 1

Listofselectedsamples,chemicalcompositionsandgeodynamicsettings.FortheYellowstonesamples,inside,borderandoutsiderefertothelocationofthesampledsiteswithrespecttothelimitofthecaldera.DMMandPM refertothedepletedMORBmantleandtheplumemantle,respectively,asdefinedbyNeisotopesmeasuredinthesesamples.The3_He/4_{He ratiosareexpressedasmultiplesoftheatmosphericvalueofRa=1.39×10}−6_.“n.d”:

notdetermined.ppmv:partpermillioninvolume.References:1:Zartmanetal.(1961);2:Staudacher(1987);3:Bräueretal.(2013);4:Bekaertetal.(2019);5:Labidietal.(2020) forNisotopedata;6:Nakaietal.(1997);for He,NeandXeisotopedata;7:ChemicalanalysiscarriedoutatINGVPalermo.8:Fischeretal.(2009);9:Broadleyetal.(2020) forHe,NeandXeisotopedata;10:JavoyandPineau(1991);11:NeandHedatafromMoreiraet al.(1998),the40_Ar/36_{Arratioisthehighestvaluereportedbytheseauthors;12:He,NeandXeisotopedatafromMukhopadhyay(2012),the}40_Ar/36_{Ar ratioisthehighestvaluereportedbythisauthor;13:Carbonisotopesfor}

sampleMD-1fromthesamesite(Barryetal.,2014);14:Chemicalanalysis(CO2,Heisotopes)carriedoutatCRPG;15:MartyandDauphas(2003) forNisotopes;16:OzimaandPodosek(2002).

Type Country Location Sample name Excess129_{Xe Neon isotopes CO}

2 δ13C N2 δ15N He Ar 3He/4He 40Ar/36Ar Refs.

(%) (, PDB) (%) (, ATM) (ppmv) (ppmv) (Ra) CO2-rich gases

CO2 well gas USA Bueyeros NM Mitchell n◦4 √ DMM 99.9 −3.9 0.12 n.d. 45 28 3.2 22500 1,2 CO2 well gas USA Bueyeros NM Mitchell n◦12 √ DMM 99.8 −4.1 0.15 n.d. 47 29 3.2 34000 1,2 Bubbling water Germany Eifel Victoriaquelle √ DMM 99.77 −2.1 0.18 −1.7 32 37 4.4 4040 3,4,5 Bubbling water Germany Eifel Schwefelquelle √ DMM 99.68 −2.0 0.24 −1.8 31 62 4.5 8287 3,4,5 CO2 well gas Czech Rep. Eger graben Bublak n.d. n.d. 99.8 −2.0 0.20 −3.2 19 36 5.9 551 3 CO2 well gas Italy Etna area Naftia √ DMM 97.6 −1.0 0.53 n.d. 64 78 6.6 1445 6, 7 Volcanic gas Tanzania Oldoinyo Lengai OLD-2 n.d. DMM 99.34 −2.4 0.66 −4.1 27 90 6.9 947 8 Geothermal gas USA Yellowstone, inside Mud Volcano 2A n.d. PM 98.50 −2.6 0.26 −8.6 15 40 14.4 414 5,7,9 Geothermal gas USA Yellowstone, inside Mud Volcano 2B n.d. PM 98.10 −2.7 0.31 −1.9 18 77 13.8 450 5,7,9 Geothermal gas USA Yellowstone, border Turbid Lake 3A n.d. PM 96.34 −2.9 0.23 −1.9 34 67 5.27 433 5,7,9 Geothermal gas USA Yellowstone, border Turbid Lake 3B n.d. PM 96.55 −2.6 0.22 −0.1 38 62 5.19 433 5,7,9 Geothermal gas USA Yellowstone, outside Brimstone Basin 4A √ PM 98.15 −2.7 0.16 0.2 72 52 2.77 1536 5,7,9 Geothermal gas USA Yellowstone, outside Brimstone Basin 4B √. n.d. 98.05 −2.8 0.16 −1.4 68 52 2.85 1417 5,7,9 Basaltic glasses

MORB glass Mid-Atlantic 14◦N 2D43 √ DMM 95.0 −3.7 0.12 −3.1 33 24 8.1 24500 10.11 Subglacial glass Iceland 64◦N DICE √ PM 99.9 −3.6 0.04 0.2 14 6.2 17.9 7047 12,13,14,15

Atmosphere 0.04 78.08 0 5.24 9340 1 298.6 16

Table 2

AbundanceratiosofCO2-richgases(samenotationasinTable1).TheratiosarecorrectedforatmosphericcontaminationusingEqns.(2)-(5) (firstthreenumericalcolumns).

The4th_{columngivestheC /N ratiowheretheNcontenthasbeencorrectedforatmosphericcontamination.The5}th_{columngivesthe}4_He/40_{Ar*ratiowhere}4_Heequalsto

theHecontentand40_{Ar*isapproximatedbythetotalArcontentcorrectedforaircontamination.n.d.:notdetermined.Notethatthe}4_He/40_{Ar*ratiosarewithin,orclose}

to,themantleproduction/accumulationrangeof1.8-3.0.Theexactvalueofthisratioisnotpreciselyknownsinceitdependsontheparent(K+U)/Thabundanceratioand ontheresidencetimeofnoblegasesintherespectivemantlesources.ThelasttwocolumnsrepresenttheC /N molarratioscorrectedforatmosphericcontaminationand degassingfractionation.CorrectionforthelatterhasbeenmadeusingEqn. (1) forthetwoend-membervaluesofthemantlerange.ThemeanC /N ratiooftheCO2-rich

gasesanalyzedhereexcludingglassdata(412-617)ofwithglassdata(453-680)isconsistentwiththemeanMORB(535±224;MartyandZimmermann,1999). Location Name Type C/N N/Ar He/Ar N/Ar C/N 4_He/40_{Ar* C/N}

Meas. Meas. Meas. Corr. Atm. Corr. Atm. Corr. Atm. Corrforatm.cont &degas.fract. 1.8 3.0 CO2-rich gases Bueyeros NM Mitchell n◦4 DMM 416 86 1.59 86 417 1.60 459 688 Bueyeros NM Mitchell n◦12 DMM 333 103 1.60 103 333 1.60 366 549 Eifel Victoriaquelle DMM 277 97 0.87 101 285 0.94 479 718 Eifel Schwefelquelle DMM 208 77 0.50 79 211 0.52 565 847 Eger graben Bublak n.d. 249 100 0.53 108 559 1.15 798 1198 Etnean province Naftia DMM 92 136 0.82 161 98 1.03 152 228 Oldoinyo Lengai OLD-2 DMM 76 145 0.30 192 83 0.43 256 384 Yellowstone, inside Mud Volcano 2A PM 189 130 0.39 355 241 1.35 304 456 Yellowstone, inside Mud Volcano 2B PM 158 81 0.23 158 232 0.67 511 767 Yellowstone, border Turbid Lake 3A PM 209 69 0.50 131 343 1.57 382 573 Yellowstone, border Turbid Lake 3B PM 219 71 0.61 139 352 1.92 334 502 Yellowstone, outside Brimstone Basin 4A PM 307 62 1.38 66 344 1.67 366 548 Yellowstone, outside Brimstone Basin 4B PM 306 62 1.31 67 348 1.61 379 569

Basalt glasses

MAR 14◦N 2D43 DMM 396 100 1.36 101 396 1.37 492 738 Iceland 60◦N DICE PM 1130 142 2.26 146 1141 2.26 952 1428

CO2-rich gases, mean 412 617

All MORB 535±224

N-MORB 273±106 T-MORB 433±392 E-MORB 1839±641

where

α

is computed from Eqn. (3). The atmospheric contam-inant for geothermal and volcanic gases is generally carried by underground aquifers (Giggenbach, 1980). For the atmospheric (N2

/

40Ar)atm ratio,we assume avalue 39.2, whichisa meanfor

air-saturatedwateratequilibriumtemperaturesof10◦C(37.9)and 100◦C(40.5)(computedwiththeBunsen coefficientsgivenin Oz-imaandPodosek,2002).

3.2. Correctionforfractionaldegassing

Asforoceanicbasalts,thecorrectionforfractionaldegassingis basedon themeasurementofthe4He/40Ar*ratiousingEqn. (1), where40_{Ar*isapproximatedby}40_Ar

ca.The abundanceofhelium

doesnotneedtobecorrectedforatmosphericcontaminationgiven its low atmospheric content (5.24 ppmv) and the generally low level ofair contamination forthe gases selected here(correction for the He isotopic ratios are also negligible for these samples). The4_He/40_{Ar*ratiosofthegases(includinggasestrappedinDICE}

and 2D43 vesicles) range between 0.4 (Oldoinyo Lengai) and 2.3(DICE) (Table 2). Thesevalues are muchcloser tothe mantle production/accumulation range (1.8-3.0) than ratios measured in oceanicbasaltglasses(3-96;MartyandZimmermann,1999), mak-ing correction forfractional degassing limited. Contrary toMORB glasses which are extensively degassed, CO2-rich gases represent

a gasphaseinequilibriumwitha degassingbatchofmagmaand may also be related to new batches of rising magmas as in the caseofOldoinyoLengai(Fischer etal., 2009) andEtna(Caracausi etal.,2003).TheC /N,C

/

3He,N

/

3He and36_{Ar/N ratioscorrected}

foratmosphericcontamination andfractional degassingare given inTable2andTable3,respectively.

Fig. 1. Kerneldensitydistribution(madewithPast4@_{software)ofC /N ratiosinCO} 2

-richgases(thiswork,includinggas-richglasses) andMORBsamples(Martyand Zimmermann,1999).FortheCO2-richgases,eachdatacorrespondstothemeanof

thecorrectedratiosusing4_He/40_{Ar*of1.8or3,assigninganuncertaintyequalto}

thedifferencebetweenthetwocalculatedvalues(typically20%).

4. Discussion

4.1. DistributionofC/NvaluesinCO2-richgasesandMORBs

InFig.1wecomparethedistributionofcorrectedC /N ratiosin CO2-richgasesandinthetwo gas-richglasses(2D43andDICE)

B. Marty, M. Almayrac, P.H. Barry et al. Earth and Planetary Science Letters 551 (2020) 116574

Table 3

C/3_He,N/3_{He andN/}36_{Ar ratiosofanalyzedgases(samenotationasinTable2).}$_:36_{Ar/N ratioscorrectedforatmosphericcontaminationassuming}40_Ar/36_{Ar ratiosof}

40,000fortheconvectiveMORBmantle(Bekaertetal.,2019)and10,000fortheplumemantle(Mukhopadhyay,2012),respectively.

Location Type C/3_{He N/}3_He 36_Ar/N$ 4_He/40_{Ar* C/}3_{He N/}3_He

Meas. Corr. Atm Corr. Atm Corr. Atm Corr. degas. fract. Corr. degas. fract.

109 ₁₀6 ₁₀−7 ₁₀9 ₁₀6 1.8 3.0 1.8 3.0 CO2-rich gases Harding Bueyeros NM CM 5.12 12.3 2.91 1.60 4.99 4.50 10.9 6.54 Harding Bueyeros NM CM 4.88 14.7 2.43 1.60 4.76 4.29 13.0 7.81 Eifel CM 5.01 18.1 2.46 0.94 4.38 3.94 9.15 5.49 Eifel CM 5.06 24.3 3.18 0.52 3.91 3.52 6.93 4.16 Eger graben n.d. 8.34 21.2 2.31 1.15 7.60 6.84 9.52 5.71 Etna area CM 1.66 18.0 1.56 1.03 1.48 1.33 9.71 5.83 Oldoinyo Lengai CM 3.82 50.3 1.30 0.43 2.85 2.56 11.1 6.67 Yellowstone PM 3.29 17.4 2.82 1.35 3.10 2.79 10.2 6.12 Yellowstone PM 2.86 18.1 6.32 0.67 2.33 2.09 4.55 2.73 Yellowstone PM 1.42 6.8 7.61 1.57 1.38 1.24 3.62 2.17 Yellowstone PM 1.26 5.7 7.21 1.92 1.28 1.15 3.82 2.29 Yellowstone PM 3.87 12.6 15.1 1.67 3.80 3.42 10.4 6.25 Yellowstone PM 4.39 14.3 15.0 1.61 4.29 3.86 11.3 6.78 Mean 5.40 3.55 3.20 8.79 5.27 Median 2.90 3.80 3.42 9.71 5.83 Basalt glasses MAR 14◦N CM 2.60 6.57 2.48 1.37 2.46 2.21 5.00 3.00 N 64◦ PM 3.03 2.53 6.84 2.26 3.18 2.86 3.15 1.89

Fig. 2a. C/3_{He versusC /N forCO}

2-richgases(datafromTables2&3).Reddots:

Yellowstonegases;Greendots:volcanicgases;Browndots:Eifelgases;Bluedots: NewMexicoandEgerRiftgases.Variationscanbeunderstoodasresultingfroma 3-end-membermixingbetweenN-MORB,E-MORBandacrust-influencedcomponent (thelatterexemplifiedbytheEgerRiftBublakgas.(Forinterpretationofthecolors inthefigure(s),thereaderisreferredtothewebversionofthisarticle.)

with that of all corrected MORB data (Marty and Zimmermann, 1999). The distribution of CO2-rich gas data shows a mean C /N

around 500 whereas the distribution of MORB values is shifted

towards low N-MORB values, which were more numerous than

carbon-richE-MORBvaluesinthecompilationofMarty and Zim-mermann(1999). Hencedatafromnaturalgases confirmthat the respectivemagmasourceshaveC /N ratioshigherthanSolar, Chon-dritic(CIandCM,Fig.5)andthemantlevaluepreviouslyproposed byBerginetal.(2015).

4.2. Crustalorsubcontinentallithosphericmantle(SCLM)contribution?

AlthoughgascompositionsreportedhereindicatethatC /N and C

/

3He ratios are comparable to those of the mantle sources of oceanic basalts (Figs. 1 & 2; Tables 2 & 3), the underlying crust

Fig. 2b. Same formatasFig.2a.AttemptofcorrectionforcontributionofcrustalC andN.ThegreenarearepresentsC/3_{He andC /N ratiostrendingtowardsthe}

ori-gin(nocarbon).N-MORBandE-MORBaveragesarefromMartyandZimmermann (1999).TheblueareasrepresenttherangesofC/3_{He ratiosfortheMORBmantle}

accordingtoTuckeretal.(2018) andMartyandZimmermann(1999),respectively. ThemantlesourceC /N ratioisobtainedbyintersectingthegreenareawiththe bluearea.

and/or subcontinental lithospheric mantle may also significantly contributetothebudget ofoutgassingvolatiles.Wenote thatthe isotopiccompositions ofcarbon andnitrogenare consistent with a mantle origin(δ13C

= −

5

±

2



;

δ

15N

= −

5

±

2



for MORBs anddiamonds,

δ

15N

=

0

±

3



formantleplumes;e.g., Javoyand Pineau,1991;MartyandZimmermann,1999;MartyandDauphas, 2003) for thesegases (Table 1), although some of the valuesare intermediate between those ofcarbonates (δ13_C

₌

_{0) andof the}

atmosphere(δ15N

=

0



),respectively,suggestingcontributionsof surfaceand/orcrustalvolatiles.The3He/4He ratiosofsomeofthe CO2-richgasesconsideredhere(3.2RaforNewMexicogases,

4.4-5.9RaforcontinentalEurope gases,2.4Ra forYellowstone gases sampledoutsidethecaldera) arelower than thecanonical MORB

rangeof8

±

1Ra,andsamplesfrominsidetheYellowstonecaldera. This isalso consistentwith contributionsof radiogenic4He from theregionalcrust.

HereweinvestigatecrustalC(andbyextensionN)contribution tomantle-derived gasesby usingtheC

/

3He ratios,which consti-tute a good indicator of C provenance since the crust is highly depletedinprimordial3_{Herelativetothemantle.InaC}

_/

3_He

ver-sus C /N diagram(Fig. 2),severalofthesamplesplotcloseto the fielddefinedbyMORBsandareintermediatebetweenthemean

N-MORB andthecarbon-richE-MORB endmembers. However,some

of the data points also seem to depart from the mixing trend betweenthesetwo MORBendmembers.Thistendencyisbest ex-emplifiedbytheEgerRiftsample(Fig.2a),butitcouldalsobethe casefortheNewMexicoandEifelsamples.FortheEifelsamples, Labidietal.(2020) arguedfromNisotopesystematicsthatcrustal contamination ofthe gas is unlikely.However, theseauthors did notexcludethecontributionofancientsubductedcrusttothe sub-continentalmantleinWestern Europe.Forthe Yellowstonegases, WernerandBrantley(2003) estimatedthat

∼

50%ofCO2 isof

sed-imentary originfor the emanations outside the caldera andthat morethan

∼

70%CO2 ismagmaticinsidethecaldera.Inlinewith

this view, we note that the C O2

/

3He ratios at Mud Volcano in

the centerofthecaldera arelower (2.5

×

109_{) thanthose ofthe}

Brimstone basin gases (4

×

109_{) located outside the caldera,}

de-spitehavingcomparable

δ

13C values(Table1).Labidietal.(2020) concluded from N isotopologues that the contribution of crustal nitrogento theYellowstone gases insidethecalderais negligible. OveralltheC

/

3He ratiooftheYellowstonegasesinsidethecaldera (2.7

×

109_{) is comparable to the Icelandic DICE ratio(2.8}

_×

₁₀9₎

(Table3),wherecrustalcontaminationisminimal.

In principleit should bepossible tocorrect Cforcrustal con-tributionusingthe C O2

/

3He ratios,butthisiscomplicatedbythe

factthatE-typeMORBglassesorplumesamplesalsohaveelevated

C O2

/

3He ratios(upto6

×

109)andhighC O2

/

3He ratioscouldbe

due tomantle source heterogeneitiesand not crustal contamina-tionofmagmasorrelatedgases. InaC

/

3He versus C

/

N diagram

plot (Fig. 2b),contributions of pure crustal C (without crustal N) willproducecorrelationsanchoredatthemantleend-member val-uesandextending towardhighC

/

3He andC /N ratios. Eachdata pointiscorrectedforcrustalCcontributionbyextrapolating mea-sureddatatowardszeroandtakingtheintersectionoftheselines withtheC O2

/

3He ratio inferredforthemantle(2.2

±

0.6

×

109,

Marty andZimmermann, 1999 or 1.7

±

0.2

×

109_{, Tucker etal.,}

2018;Fig.2b).WiththesetwoC O2

/

3He values,themantlesource C

/

N ratiosareconstrainedtobewithin100-530(Fig.2b).

There is howevertwo caveats with thisapproach. Firstly, the dispersionofdatacanbeaccountedforbypreferentiallossof3He whileleavingunaffectedCandN(near-constantC/Nratio),as sug-gestedbyonereviewer.Inthiscase,theobservedC/Nratiosshould berepresentativeofthemantlesourcevalue(s).Secondly,the cor-rection forcrustal C addition assumesthat onlyC, andnot N, is contributedbythecrust.Tocheckthisassumption,theC

/

3He

ra-tios are plottedasa functionof the N

/

3_{He ratios in Fig.3(data}

fromTable2).Dataseemtodefinetwodomainsthatbothoriginate

from the N-MORB composition, one encompassing the E-MORB

endmember andthe other one pointingtoward high N

/

3He

val-uesandencompassesmostofcontinentalgasdata.Thiscorrelation corresponds to a C

/

N ratio around 400-500 (Fig. 3). However, a typical crustal C

/

N ratiowould be

≤

70(with C

=

27

±

5 ppmw, Hirschmann,2018,andN

=

0.50

±

0.04 ppmw,Johnsonand Gold-blatt, 2015) and cannot therefore account for the elevated C /N

ratiooftheCO2-richgases.

The sub-continental lithospheric mantle (C/N

=

390

±

260, computedwith C

=

100

±

20 ppmw,Hirschmann,2018,andN

=

0.3

±

0.2 ppmw, Johnson and Goldblatt, 2015) could be a pos-sible endmember for some of the continental gases (Fig. 3), but

Fig. 3. C/3He ratiosasafunctionofthe N/3He ratiosforCO2-richgases.Same

symbolsasinFig. 2.TheMORBellipse(blue)encirclesN-MORBandE-MORBvalues (MartyandZimmermann,1999).TheyellowellipseencompassesCO2-richgases,

withanapproximateslopecorrespondingtoaC/N ratioof∼450,whichishigher thanthecrustalratio(≤70,seeSection4.2).Uncertaintieson the N/3_{He ratios}

arerepresentedbythedifferencebetweencorrectedvaluesusingthetwopossible mantle4_He/40_{Ar*ratios(1.8and3.0).}

other geochemicaland geophysicalevidence support an astheno-spheric, rather than lithospheric, originfor Cenozoic magmatism inthese regions. Evidence include: (i) trace element andisotope systematicsoferuptedlavas(Hoernle etal.,1995);(ii) geochemi-cal affinitieswithHIMU-typemagmatism forEifel(Bekaert etal., 2019, and refs. therein, Sicily (Mt. Etna; Caracausi et al., 2003, andrefs.therein),Tanzania(Oldoinyo Lengai;Fischeretal., 2009, andrefs. therein) and New Mexico (Fittonet al., 1988), and (iii) geophysical imaging showing upwelling of asthenospheric mate-rial below European magmatic provinces (Hoernle et al., 1995). Thus, theC-N-3_{He signatures ofrelatedCO}

2-rich gases are likely

toreflectthecompositionofasthenosphericsourcesdifferentfrom thosefeedingmid-oceanridges.EnrichmentsinCandNcompared totypicalMORBs(Fig.3)maybeduetoprimordialheterogeneities, ortherecyclingofsurfaceCandNdeepintothemantle(Busigny andBebout,2013),inlinewiththeHIMUcharacterofsomeofthe corresponding magmatic provinces. In support of thispossibility, Labidietal.(2020) arguedfromNisotopologuesthatabout30%of nitrogenintheEifelCO2-richgasescouldberecycledinorigin.

Contribution of C-rich material into the mantle sources of E-MORBsand of intraplate volcanoes could potentially account for thehighC

/

N componentsuggestedby trendsdisplayedinFig. 3. ItisnotclearhoweverifthehighC

/

N ratiooftheIcelandicDICE sample and of the (inside caldera) Yellowstone samples have a similarrecycledorigin. Labidietal.(2020) proposedthat nitrogen inYellowstone gasesdefinesaprimordialmantleplume signature having a

δ

15N value close to 0

(+

3

±

2



) and a N

/

36Ar

ra-tio of 2.4-3.9

×

106. Marty and Dauphas (2003) also noted that plume-relatedgases have

δ

15N valuescloseto0



(

−

1.5to3.3



for DICE) but these signatures were instead attributed to recy-clingof surface N intothe source ofmantle plumes (Marty and Dauphas, 2003; Halldórsson et al., 2016). We note that the C /N

ratios are different betweenYellowstone (400-500) and Icelandic (1200-1400) samples (Table 2), suggestingthat the C

/

N ratio of

B. Marty, M. Almayrac, P.H. Barry et al. Earth and Planetary Science Letters 551 (2020) 116574

Table 4

C,Nand36_{Arconcentrations(inventoriesdividedbythemassofthesilicateEarth}

of4×1027_{g)andelementalratios.ForthesurfaceoftheEarth(atmosphereand}

crust),CandNdataarefromHirschmann(2018) andrefs.thereinandthe36_Ar

con-centrationcorrespondstotheatmosphericinventory(OzimaandPodosek,2002), thecrustalreservoirbeingneglected.Forthemantle,theNconcentrationisfrom Hirschmann(2018) andtheCand36_{Arconcentrationsarecomputedfromthe}

man-tleC /N and36_{Ar/N ratiosestimatedhere(seetextandSupplementaryMaterial).}

A2andA3refertoApproach1andApproach2,respectively,asdescribedinthe maintextandintheSupplementaryMaterialsection.A2ismedianandinterquartile range(betweenbrackets,Q1=1st_{interquartileandQ3=3}rd_{interquartile,}

respec-tively)ofgas dataonlyandA3ismedianand interquartilerangeofgas+MORB data.

Surface Mantle BSE N, ppmw 1.56±0.06 1.10±0.55

C, ppmw 27±5 A2:479Q 1Q 3::323646

A3:337Q 1Q 3::211504 36_{Ar, 10}−6_{ppmw 50 1.29}Q 1:0.82

Q 3:1.89

C/N mol/mol 21±6 A2:474_{Q 3}Q 1_::399₆₁₈ A2:219Q 1_{Q 3}:_:174₂₅₉ A3:352Q 1Q 3::167606 A3:158 Q 1:112 Q 3:215 36_{Ar/N, 10}−7_{mol/mol 124±5 2.90}Q 1:2.35 Q 3:7.40 74.3 Q 1:65.7 Q 3:85.2

thedeepmantleisheterogeneousasa possibleresultofdifferent mixingsbetweenprimordialandrecyclednitrogen.

4.3. TheC/Nratioofthemantle

We attempted to estimate the C /N ratio of the BSE using a mass balance involving nitrogen andmantle C /N ratios (Supple-mentaryMaterial).Asinprevious work,thisapproachisbasedon (i)thebulkKcontentoftheBSE(Arevaloetal.,2009)andK-Ar-N systematicsofMORBandOIB-relatedsamples(Marty,1995;Marty and Dauphas, 2003; Johnson and Goldblatt, 2015; Bergin et al., 2015; Hirschmann, 2018), which are usedto estimate the BSE N content,and(ii)theC/NratiosmeasuredinMORBandOIBbasalt glasses (Marty, 1995; Marty and Dauphas, 2003) and in mantle-derived CO2-rich gases (thisstudy). Indoing so, we assume that

theC/Nratiosmeasuredinsurfacebasaltsarerepresentativeofthe mantlesourcecomposition,aftercorrectionforfractionaldegassing andatmosphericcontamination(seeIntroduction).Becausewe se-lected samples from both the convective mantle (MORB glasses, CO2-richgaseswithMORB-likenoblegascomposition,e.g.,Ne

iso-topes) andfromOIB-like mantledomainsincludinghigh-3_He/4_He

plumes(Iceland,Yellowstone),wefurtherpositthatthesesamples are representative of not only the convective mantle responsible for feeding mid-ocean ridges, butalso encompass mantle plume domains.However,thisapproachreliesontheassumptionthatthe C/N ratio measured in mantle-derived basalts and gases is rep-resentative of thewhole mantle composition from its shallowest regions down to the core-mantle interface, which isnot guaran-teed. There are some suggestions from diamondstudies that the C/Nratioofmantlefluidsfromwhichdiamondsformedmayvary withdepth,althoughsuchvariations couldresultfromprocessing during diamond formation ratherthan from compositional varia-tionofthemantleitself(e.g.,MikhailandHowell,2016,andrefs. therein).

Withthesecaveats inmind,weconsidered two datasets; the previously publishedMORBdata(MartyandZimmermann,1999), and the volcanic/geothermal gas data presentedin this study. In order toestimate themantle andBSE C /N ratios,we carriedout MonteCarlosimulationsfollowingthreeapproachesasdetailedin Supplementary Material. We also estimated the 36Ar/N ratios of thesurfaceandmantlereservoirsinordertofurtherconstrainthe compositionoftheBSE(Table4).

Approach1 uses gas data only (without gas-rich glasses) for the mantle. The corresponding samples tap large magmatic gas

reservoirswhicharemorepronetohomogenizethechemical com-positions, and for which the effects of fractional degassing are limited.InthisapproachweusethemeanC

/

N ratioof541andits standarddeviationof199(computedwithCO2-richgasdatafrom

Table2;foreachsample wetooktheaverageofthetwonumbers obtainedwith4He/40Ar*valuesofeither1.8or3.0).

Approach2 alsoreliesonthegasdataonly,usingamedianC /N

value of474,andan interquartile range ((IQR

=

75th percentile -25th percentile

=

Q3 - Q1

=

618-399

=

219).The fact that the meanandthemedianvaluesarecloseis consistentwiththe ob-servednear-Gaussiandistributionofgasdata(Fig.1andFig. S1).

Approach3 considers all available data for MORB and gases. Whentakingthisapproach,thedistributionbecomesnon-Gaussian andwethereforeusemedianandIQRvalues.

In the followingwe considered only Approaches2and 3 (me-dian andquartiles)given thenon-Gaussian distribution of MORB

C

/

N datathatprohibitsusingmeansandstandarddeviations.For thesakeofhomogeneity,wealsoconsideredmedianandquartiles forthe36Ar

/

N valuesofthemantleandoftheBSE.Theoutcomes ofthesecalculationsaresummarized inTable4 anddisplayedin Fig.4.Ournewrangeof bestestimatesforthe mantleC /N ratio

is between 350-420, which is significantly higher than the ratio of

∼

90obtainedfromtheCO2/Bacalibration(Berginetal.,2015;

Hirschmann, 2018), although marginally compatiblewhen taking intoaccount the large range ofuncertainties (Q1: 167, G3:618). We nevertheless assertthat ourestimate is representative ofthe mantlevalue(s)becauselowmantleC /N valuespredictedfromthe CO2/BaapproacharerarelyobservedinMORBsorCO2-richgases.

Doingso,weestimatedamantleCcontentaround330-450ppmw (Table 4), which is higherthan values previously suggested(e.g., 110

±

40 ppmw;Hirschmann,2018),althoughtherangeof uncer-taintiesisalsolarge(Q1:157,Q3:583).

4.4.OriginoftheBSEC/Nratio

FromournewlycomputedmantleC /N ratiowe estimatedBSE

C /N ratioto be 158-219(range ofIQs:112-259; Table4), higher than the previously proposed value of

∼

60(Bergin et al., 2015), further emphasizing the non-canonical composition of terrestrial volatilespointed out by Berginetal. (2015). The highC

/

N ratio

ofthemantlecould haveresultedfrompreferential sequestration ofnitrogeninplanetary corescompared tocarbon during Earth’s formationeventsand/or during differentiationofparent planetes-imals(Marty, 2012; Roskosz etal., 2013). However, experimental evidence doesnot support a higher metal-silicate partitioningof Ncompared to C (Dalouetal., 2017), although therole of other elements likeH andpressure needto be assessed (Grewal etal., 2020).Onepossibilityisthatthemetal-silicatepartitioning behav-ior ofC/N couldbe dependent on pressure,whichremains to be experimentallyconstrained”

Preferentialdegassingofnitrogenduringaccretionaryprocesses andits subsequent loss to space, possibly combinedwith metal-silicatepartitioning, was proposedasan alternative(Berginetal., 2015). It is not clear if magma degassing during Earth’s accre-tionaryevents could havefractionated enough Nrelative to C to account for theBSE C/N ratio,which signature might havebeen inherited from planetary bodies that accreted to form the Earth (Hirschmann,2016). Alternatively, Liu et al.(2019) proposed the formationofnitrogen-richimmiscibleliquidsthat wouldhave re-moved Nfrom theforming mantlematerial during metal-silicate interaction.

Here we add a supplementary dimension to this problem by addingthenoblegas/nitrogenratios.InFig.theC /N ratiosof me-teorites and terrestrial reservoirs are plotted against the 36Ar/N

ratios (Supplementary Material). The different meteorite groups appear to define a cosmochemical trend fromCI-CM, EC and CR

Fig. 4. Graphic summary of the estimates presented in this work (data from Tables3and4). Abundance ratios are molar.

groups towards CV and CO groups (and possibly ureilites) hav-ing elevated C /N and 36Ar/N ratios. The CV (and CO)chondrites aredepletedinvolatileelements(includingC,Nandnoblegases) and moderately volatile elements compared to CM-CIs (Palme andO’Neill,2013).CVs,whichare metamorphizedtovariable ex-tent, are depleted in volatile elements and thought to represent the non-differentiated upper layers of a differentiated planetary body (Elkins-Tanton etal., 2011). The silicateEarthalso presents a monotonous depletion of moderately volatile elements (com-pared to CI-CM chondrites) that, although more pronounced, re-semblesthat ofCVs (PalmeandO’Neill,2013;Braukmülleretal., 2019). This trend cannot be due solely to nebular condensation and is instead consistent with volatile loss during highly ener-getic events (e.g., impacts) and/or from metamorphism on plan-etary parentbodies.Hence thecosmochemicaltrendofmeteorite families(Fig.4)mayrepresentthecombinedeffectsofparentbody metarmophismandpartialdifferentiation.Theseprocessesresulted in(i)thedepletion ofnitrogenrelativetoCand36_{Ar,and(ii)the}

moderately volatile element depletion trend observed in evolved planetarybodiesincludingtheEarth.

BecauseArisnotasdepletedasN(Fig.4),thisdifferential be-havior of these two elements may be due to variable retention ratesoftheirrespectivehostphases.Indeed,meteoriticnoblegases aretrappedinarefractoryphase(thepoorly-definedQphase asso-ciatedwithinsolubleorganicmatter- IOM)whereasaconsiderable fractionof nitrogenistrapped inlessrefractory organicslike the soluble organic matter (SOM). Due to the much higher thermal sensitivity of the SOM relative to the IOM (Remusat, 2015), the preferentiallossofNovernoblegasescouldbeaccountedforbya preferentialdestructionoflabile,N-bearingorganicphasesrelative tothemorerefractory,noblegas-bearingones.Analternative pos-sibility wouldbe the formationof N-rich liquidsand subsequent degassing/loss to space as proposed by Lui et al. (2019), which wouldhavedepletedNonly,andnotCnorAr,fromforming plan-etesimalsorproto-Earth.

ThefactthattheC-N-ArsignatureoftheBSEiswithinthefield of CV-CO data does not imply that the building blocks of Earth weremadeofthismaterial.ConverselyweproposethatEarthwas madeof an assemblage ofprecursors thatcondensed and evapo-ratedatvarious temperatures,whicharebest representedfroma volatileelementperspectiveby theCVchondrites.Thispossibility is independently supported withthe monotonous volatile deple-tion trend of the silicate Earth that parallels that of CVs (Palme andO’Neill,2013;Braukmülleretal.,2019).

Fig. 5. C /N versus36_{Ar/N formeteoritesandterrestrialreservoirs(notethe}

log-arithmic scale). See Table4 for terrestrialestimates, SupplementaryTableS1 for

meteoriticdata,andSupplementaryFig.S5fordiscussionofmeteoriticdata.

4.5.Terrestrialfractionation

Theterrestrialreservoirs (surfaceandmantle) clearlyplot out-sidethe meteoritictrend(Fig. 5).The surface ofEarthhasa C /N

ratio compatible with the range of CI-CMs but its 36Ar/N ratio

is in the range of CV-CO values, whereas the terrestrial mantle has an elevated C /N ratio akin of CV-COs, but its ratio 36_Ar/N

ratioiswithintherangeofCI-CM-ECs.Hencenoneofthese reser-voirs can be directly linked to known chondritic compositions. However,summinguptheC

,

N36_{Ar concentrationsofthese}

reser-voirs,theobtainedBSE compositionfallsonthecosmictrendand plotswithin the field ofCV-COs, asexplained inSection 4.4.We thereforeproposethatsurface andthemantlecompositionswere established by chemical fractionationof theinitial BSE, involving preferentialincorporation,orretention,ofCintothesilicateEarth

B. Marty, M. Almayrac, P.H. Barry et al. Earth and Planetary Science Letters 551 (2020) 116574

comparedtovolatileNandnoblegases,anditscorresponding de-pletionattheEarth’ssurfacerelativetoNandnoblegases.

TheC/N ratio of the mantle, asestimated in this study (350-420), is in line with that outgassed atmid ocean ridges(385

±

287),ascalculatedfromtheCO2 (MartyandTolstikhin,1998)and

N fluxes (5.7

±

3.6

×

1012 mol/yr; Hirschmann, 2018 and refer-ences therein).The similaritybetweenthemantle C /N estimated

in this study and that degassed at MOR suggests that C and N arenot significantlyfractionatedduringdegassingunderthe rela-tively oxidizingconditions thatprevail inthepresent-day mantle (Hirschmann,2016), aswell as furthersupporting the validity of ourmantle C /N estimates.Sinceboth CandNare highly incom-patible during mantle meltingthisis not unexpectedbut itdoes highlight that the long-term degassing from the mantle cannot account for the difference in C /N between the mantle and the surface. Therefore to explain the difference in the C/N between themantleandthesurface requiresthatChasbeenpreferentially retainedinthesilicateEarth,orthatvolatileNhasbeen preferen-tiallylostoftheEarth’ssurface.

Impacts during terrestrial accretion might have severely de-gassedimpacting planetesimals, whichthemselves could have al-ready been differentiated between a core, a silicate mantle and a proto-atmosphereandcould thereforeprovideamechanismfor fractionationCandN.IfNwaspreferentiallyreleasedtotheearly atmosphere, then during periods of impact driven atmospheric loss, nitrogen wouldbe preferentially lost to space and the C /N

oftheEarth’ssurfacewouldbeloweredtowardsitspresentvalue. Thismechanismhasbeenpreviouslysuggestedtoaccountforthe Earth’ssubchondriticC /H andCl/F ratios(Tuckerand Mukhopad-hyay, 2014). For thismechanism tosatisfactorily explain the low C/NoftheEarth’ssurfacerequiresthatincompatibilityofCandN duringmantlemeltingwasdifferentthanatpresent,withNbeing preferentially degassedrelativetoC.Thismayhavebeenthecase during the early Earth when the mantle was most likely domi-natedbyreducingconditions(AulbachandStagno,2016).However, recentexperimentalstudiesshowthatevenunderreducing condi-tionsCandNmaybehavesimilarlywithbothlikelybeingretained inthesilicatemantle(Grewaletal.,2020).Itthereforeremains un-clearwhetherNwaspreferentially degassedto,andsubsequently lost from, the Earth’s early atmosphere. Further constraints can also be gained by comparing the 36Ar/N of the mantle and the Earth’s surface reservoir(Fig. 5). Since Aris a noblegas,it’s sol-ubility in mantlemelts is not subjectto differences in oxidation state andthereforeeven duringreducing conditions,Arwouldbe concentrated in the atmosphere relative to the mantle. Episodes ofatmosphericloss shouldtherefore lowerthe 36_{Ar/N inthe}

at-mosphere relative tothe mantle,which is not thecase. The role of atmosphericlosson the C/Nof thesurface reservoir therefore remainsunclearandwarrantsfurtherinvestigation.

The fractionation of C/N betweenthe Earth’s surface and the mantle maybe theresult of thelong-term preferential recycling ofCinthecrustorthemantle.Lietal.(2019) proposedthat the alteredoceaniccrustisthekeycarrierofCrecyclingintothe man-tle. Anintermediate reservoir could be thearccrust (Barry etal. (2019) orthelithosphere(KelemenandManning,2015),although thisview has beenchallenged(Plank andManning,2019). These processesmighthavebeenoperativesincetheonsetofplate tec-tonicsbutlittleisknownaboutthetectonicregimeofEarlyEarth. Recycling/sequestrationofcarbonshouldhavebeenspecifictothis element with regard to nitrogen in order to yield the elevated C/N ratio ofthe mantle.Furthermore,recycling ofC should have takenplaceearlyinEarth’shistoryinordertoavoidaVenus-like greenhouse effect. Sleep et al. (2001) proposed an elegant solu-tion tothisproblem. Theseauthorspostulated thatC mighthave beenreintroducedintothemantleearlyinEarth’shistorythrough the foundering of crustal blocks that formed the lid of magma

oceans (Hadean vertical tectonics). The carbonate formed in the crust wouldhaveformeddirectlyfromthe massiveCO2 early

at-mosphere.IfsuchamechanismcouldtransportCfromthesurface tothemantle,withoutventingCO2 backto theatmosphere,then

the fractionation of C and N between the mantle and the sur-facecould have beenan early feature. Theearly sequestration of Cinto themantleisindependently requiredtoallow forclement temperatureconditionsandthe formation ofliquidwateron the Earth’ssurface,whichisknowntohaveexistedearlyontheEarth’s surface. Inorder to achieve thisandavoid thedevelopment of a CO2 dominated atmospheres such as that exists on Venus, large

amountsof CO2 must havebeentransportedfromthe surface to

the mantleand sequestrated there over geological timescales.To accountforthedifferentC-N-36Arcompositionsofthesurfaceand the mantle,

∼

90% of surface C should have been being seques-tratedintosilicatesduringtheseearlyepisodes.Followingthe on-setofplatetectonicssubduction couldhavefurther augmentedC sequestrationinthemantlebypreferentialrecyclingCrelativetoN (BusignyandBebout,2013) andnoble gases(Mukhopadhyayand Parai, 2019, andrefs. therein). Carbon sequestration in the deep mantleeitherviaearly crustalfounderingorsubductiontherefore not onlyplays a significant role infractionatingthe Cand N be-tween the Earth’ssurface and interiorbutmay play an essential role inmodulatingthe Earth’sclimateanddevelopingthe condi-tions rightforthe emergence of life onthe early Earth. The fact thattheisotopiccompositionofmantlecarboniscomparablewith thatofsurfacecarbon(MackenzieandLerman,2006)suggeststhat carbon sequestration did not fractionate its isotopes. In contrast, Nandnoblegases werelessefficientlyhomogenizedbetweenthe surface andthe mantle, which permitted the survival ofisotopic heterogeneitiesfortheseelements.

5. Conclusions

AllsampleconsideredinthisstudyshowhighC /N and36Ar/N

ratios relative to the surface (atmosphere plus crust) inventory, pointing towards a mantle that has a unique volatile composi-tionwhencomparedtocosmochemicalreservoirs(Fig.5).TheBSE composition resembles that of chondritic groups (CV-CO) which arevolatile-depletedandappeartohaveexperienced variable ex-tent ofmetamorphism andplanetary differentiation. We propose thatEarthformedfromarangeofalreadyevolvedplanetary pre-cursorsthatweredepletedinvolatileandmoderatelyvolatile ele-mentschemically(butnotisotopicallyaffected)akintoCVandCO groups. Hence the high C /N character ofthe BSE may be inher-itedratherthanuniquelyduetoEarth’sformingprocesses.Inthis scenario,most C (

∼

90%)was preferentially sequestrated intothe mantlerelativetovolatileelementslikeNandnoblegasesearlyin Earth’s history, permitting habitable environmental conditions to developatthesurfaceofourplanet.

CRediTauthorshipcontributionstatement

B.M.conceptualizedthestudy.B.M.,P.H.B.,D.V.B.,M.W.B.D.J.B. A.CandC.J.B.contributedtofieldsampling.A.C.,B.M.,M.W.B.,D.J.B. andM.A.performedtheanalyses.Allauthorsprovidedinputtothe dataanalysisandmanuscriptpreparation.

Declarationofcompetinginterest

Theauthorsdeclarethattheyhavenoknowncompeting finan-cialinterestsorpersonalrelationshipsthatcouldhaveappearedto influencetheworkreportedinthispaper.