in the North Pacific An intermediate-depth source of hydrothermal He and dissolved iron Earth and Planetary Science Letters

Contents lists available atScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

An intermediate-depth source of hydrothermal ³ He and dissolved iron in the North Pacific

W.J. Jenkins

^a,∗

, M. Hatta

^b

, J.N. Fitzsimmons

^c

, R. Schlitzer

^d

, N.T. Lanning

^c

, A. Shiller

^e

, N.R. Buckley

^f

, C.R. German

^a

, D.E. Lott III

^a

, G. Weiss

^b

, L. Whitmore

^e

, K. Casciotti

^g

, P.J. Lam

^h

, G.A. Cutter

^f

, K.L. Cahill

^a

aWoodsHoleOceanographicInstitution,WoodsHole,MA02543,USA bDepartmentofOceanography,UniversityofHawaiiatManoa,HI96822,USA cDepartmentofOceanography,TexasA&MUniversity,CollegeStation,TX77843,USA dAlfredWegenerInstitute,27568Bremerhaven,Germany

eCenterforTraceAnalysis,UniversityofSouthernMississippi,StennisSpaceCenter,MS39529,USA

fDepartmentofOcean,Earth,andAtmosphericSciences,OldDominionUniversity,Norfolk,VA23529-0276,USA gDepartmentofEarthSystemScience,StanfordUniversity,Stanford,CA94305,USA

hDepartmentofOceanSciences,UniversityofCalifornia,SantaCruz,SantaCruz,CA95064,USA

a r t i c l e i n f o a b s t ra c t

Articlehistory:

Received11December2019

Receivedinrevisedform10March2020 Accepted11March2020

Availableonlinexxxx Editor:L.Robinson

Keywords:

submarinehydrothermalvents seawaterdissolvedironandmanganese heliumisotopes

marineprimaryproductivity NorthPacificcirculation

We observed large water columnanomaliesin heliumisotopes and tracemetalconcentrationsabove the Loihi Seamount. The ³He/⁴Heofthe addedhelium was 27.3 timesthe atmosphericratio, clearly marking itsorigintoaprimitivemantleplume. Thedissolvedironto³Heratio(dFe:³He)exported to surrounding waters was 9.3±⁰.3×^{106. We observed the Loihi 3He and dFe “signal” atadepth of} 1100matseveralstationswithin∼^{100–1000kmofLoihi,whichexhibitedadistaldFe:3Heratioof}

∼⁴×^{106,abouthalfthe proximalratio.Theseratioswereremarkably similarto thoseobservedover} and near the Southern East Pacific Rise (SEPR) despite greatly contrasting geochemicaland volcanic- tectonic origins.In contrast,the proximal and distal dMn:³Heratios wereboth ∼¹×^{106,less than} half ofthat observed atthe SEPR. Dissolvedmethanewas minimally enrichedinwaters above Loihi Seamountandwasdistallyabsent.Usinganidealizedregional-scalemodelwereplicatedthehistorically observedregional³Hedistribution,requiringahydrothermal³HesourcefromLoihiof10.4±^{4.2mol a−1,}

∼^{2%oftheglobal abyssal}hydrothermal³Heflux.Fromthiswecomputeacorresponding dFe fluxof

∼^{40Mmol a−1.Global}circulationmodelsimulationssuggestthattheLoihi-influencedwaterseventually upwellalongthewestcoastofNorthAmerica,alsoextendingintotheshallownorthwestPacific,making itapossiblyimportantdeterminantofmarineprimaryproductioninthesubpolarNorthPacific.

©^{2020ElsevierB.V.Allrightsreserved.}

1. Introduction

It has long been recognized that the availability of dissolved iron is a limiting factor for oceanic biological production (Mar- tinand Gordon,1988). Eoliansources of thismicronutrient were regarded asdetermining factors for globalmarine production on glacial-interglacialtimescales (Berger andWefer, 1991). However, theimportance ofhydrothermal iron hasonly recentlybeenrec- ognized(Tagliabueet al., 2010, 2017), and ithas onlynow been

*

Correspondingauthor.

E-mailaddress:wjenkins@whoi.edu(W.J. Jenkins).

quantified by the use of hydrothermal ³He as a “flux gauge”

(Resingetal.,2015).AspartoftheInternationalGEOTRACESpro- gram, the U.S. Pacific Meridional Transect along ∼^{152◦W (GP15)} was occupied during the autumn of 2018. We present here re- sultsfromstationsinthevicinityoftheLoihiSeamount(∼^18.9◦N, 154.2^◦W)showingthe relationshipbetween dissolvediron, man- ganese, methane, and hydrothermal ³He over the seamount and distallyforneighboringstationsnearthe1100minjectiondepth.

We use contemporaneous dissolved Fe, Mn, and ³He measure- ments at nearby stations to unambiguously quantify the helium isotope-metalrelationshipsandfluxes,andtodeterminetheirrole inregionalbiogeochemicalcycles.Thelocationofthestationsthat informthisstudyareshownasredstarsinFig.1,whichisachart https://doi.org/10.1016/j.epsl.2020.116223

0012-821X/©^{2020ElsevierB.V.Allrightsreserved.}

2 W.J. Jenkins et al. / Earth and Planetary Science Letters 539 (2020) 116223

Fig. 1.a,LocationmapoftheGEOTRACESGP15stations(redstars,blacklabels)superimposedonamapofδ³He(in%)interpolatedto1100mdepth.Thisissimilartothat originallypresentedbyLupton(1996).Graydotsindicatehistoricalstationsusedformapping(seeSection1).Otherstationsnearbybutoutsideofthedisplayedareawere alsousedinthemapping.Theδ³Hecontour(whitelines)intervalis2%,orabouttentimessamplemeasurementerrors.b,Insetshowingneighboringstationsclosetothe LoihiSeamount,closetotheIslandofHawaiiwithbottomtopographyingrayscale,contourintervalis1000m,islandsareblackandc,meridionalsectionof1100mdepth δ³He (in%)along∼^{152◦W.RedstarsarefromtheGP15(2018)sectionandblackdotsarehistoricaldatadatingfrom1985through2015(Jenkinsetal.,2019a).Errorbars} represent1standarddeviationuncertaintyduetomeasurementandinterpolationerrors.(Forinterpretationofthecolorsinthefigure(s),thereaderisreferredtotheweb versionofthisarticle.)

of thehelium isotope ratio anomaly (δ³He, defined inEq. (1) in Section 2) interpolated to 1100 m depth in the central to east- ernNorthPacific. Historicaldata(publiclyavailable,Jenkinsetal., 2019a;Lupton, 1998), indicated bygray dots, extendback tothe 1980s and are used to enhance spatial coverage. The contoured δ³He fieldshowsthelargelyeastwardinfluence ofthemid-depth volcanic ³He tongue first described and documented by Lupton (1996).Thelocationofthenearestfourstations,includingthehy- drocast directly over the Loihi Seamount are shown inthe inset (Fig. 1b).The 1100m δ³He signature forstations along ∼^152◦W (Fig.1c)showsthatthefeaturehasawell-definedmeridional ex- tentthatithaspersistedformanydecades.

2. Methods

Water samples were taken for helium isotope and noble gas measurements using a 36 place, ∼^{10 liter Niskin bottle rosette.}

The samples were drawn within a few minutes after the bottle was ventedviaTYGONtubing toa 76cmlength of1.6cmouter diameterrefrigerationgradecoppertubingthat wasisolated with pinch clamps and immediately crimp-sealed (Young and Lupton, 1983) for shore-based analysis. The samples were quantitatively gas-extractedandmass-spectrometricallyanalyzedforheliumiso- toperatios andnoblegasabundances referenced toa marine at- mosphericstandard(Jenkinsetal.,2019b).Wereportisotoperatio anomaliesrelativetotheatmosphereaccordingto

δ

³He

=

¹⁰⁰

×

₃

He

/

⁴He

₃ X

He

/

⁴He

A

−

¹

% (1)

wherethe subscript X refers to the sample, andA toair. Uncer- tainties depend on a combination of ion-counting statistics and

instrumentstability.Theyrangefrom∼^{0.15%forsamplescloseto} theatmosphericratio,upto∼^{0.4%forthehighestratiosobserved.}

Dissolved helium and neon concentrations were measured us- ing quadrupolemassspectrometricpeak-heightmanometrictech- niques(Jenkinsetal.,2019b;Stanleyetal.,2009)andarereported assaturationanomalies,defined(e.g.,forhelium)as

He

=

¹⁰⁰

×

C

(

He

)

CS

(

He

) −

¹

% (2)

whereC andCS arethemeasuredandsolubilityconcentrationsof helium (or neon). The solubilityconcentrations are a function of temperature and salinity (Jenkins etal., 2019b). The other noble gases (Ar, Kr,andXe) werealso measuredon thesesamples asa qualitycheck;butarenotreportedhere.Heliumandneonconcen- trations were determined to an accuracy of approximately0.15%.

ForboththeLoihiSeamountandPunaRidgestations,mosthelium isotope and noble gas samples were measured in replicate and average values are reported.In all cases, reproducibilitybetween replicateswasclosetoexpectederrors.Pooledstandarddeviations for δ³He, He,andNe replicateswere 0.22%, 0.29%,and0.28%

respectively,whichrepresentstheconvolutionofallsamplingand analytical errors.It shouldbe mentionedthat the δ³He andHe statistics are impacted (increased) due to the anomalously large hydrothermalsignals.

To determine the isotopic nature of the non-atmospheric he- lium,weaccountforthetwoprimaryatmosphericcomponentsof helium: that portion dissolved during air-sea gas exchange, and thatfractionintroducedbyhydrostaticcompressionofbubblesin- troducedbywaveactionattheseasurface.Theformercanbeac- countedforusingheliumisotopesolubilityfractionationmeasure- ments (Bensonand Krause, 1980) and noble gas solubility func- tions (Jenkins etal.,2019b).The latterisestimatedusingan em-

pirical relationship with near-surface neon saturation anomalies, since neon is similarly affected by air injection processes. The amounts of excess helium isotopes are calculated (Jenkins et al., 2018)using

CX S

4He

=

^CM

4He

−

^CS

4He

−

⁰

.

22

(

CM

(

Ne

) −

^CS

(

Ne

))

(3) and

C_{X S}

3He

=

^RA

1

+ δ

³He

/

100 C_M

4He

− α

EC_S

4He

−

⁰

.

22

(

C_M

(

Ne

) −

^CS

(

Ne

))

(4)

where the subscripts S and M signify solubility equilibrium andmeasured values respectively, RA is the atmospherichelium (³He/⁴He)isotope ratio (Clarke etal., 1976), and

α

E is thesolu- bilityhelium isotope effect (Benson andKrause, 1980). They are furthercorrectedfora slight(∼^0.2and∼^{0.3% forHe andNere-} spectively) enhancement in saturation anomaly at these depths resultingfromdiapycnalmixingeffects(Jenkinsetal.,2018).

FortheLoihi Seamount station,trace metalswere determined shipboardonwaterfromthesameNiskin bottlesusingFIA(Flow InjectionAnalysis) in order to avoidquestions of colocation due toshorttermtimeandspacescalevariabilitybetweenhydrocasts.

Thesampleswerefilteredthrough0.8/0.45 μmPallACROPAK500 capsulesthathadbeenpre-cleanedbysoakingthreedayswith10%

hydrochloricacid(1.2MHCl)andflushedwith3-5LofMilli-Qwa- ter,andstored refrigeratedbetween casts.The capsule filterwas pre-flushedwith0.5-1 Lofunacidifiedseawater.Thefilteredsea- watersamples were acidified to pH 2 using sub-boiling distilled hydrochloric acid. Althoughthe samples were drawn froma non trace-metalcleanNiskin rosette,the Fe andMn valuesare much largerthanmeasuredcontamination levelsforthissystem,asde- duced from contemporaneous comparison with the trace-metal- cleansystemelsewhere.

Due to much lower metal concentrations elsewhere, the U.S.

GEOTRACES trace-metal clean CTD carousel (GTC) was used for trace metal sampling for the other stations. This sampler has a plastic-coated aluminum frame, titanium pressure housings for electronics and sensors, no sacrificial zinc anodes, and 24 ×¹² liter General Oceanics GO-FLO bottles modified for trace metal sampling(CutterandBruland, 2012). Before deploymentandim- mediately upon recovery, the tops and bottoms of the GO-FLO bottleswerecoveredwithpolyethyleneshower-caps,andthebot- tleswere removedfromtheframe andcarriedintotheU.S. GEO- TRACESclean container laboratory forsub-sampling. TheGO-FLO bottles were pressurized to 6 psi using 0.2 μm HEPA-filtered compressed air, and samples were passed through 0.2 μm Pall ACROPAK SUPOR filter capsules that had been pre-cleaned by soaking overnight and flushed with 5 L of unacidified seawater.

Samples of the 0.2 μm-filtered seawater were collected follow- ingthreebottlerinsesforshipboardanalyses andforshore-based inductively-coupledplasma massspectrometry (ICP-MS)analyses.

Theshipboardsampleswere drawnintoacidpre-washed125mL polymethylpentene bottlesand immediatelyacidified to 0.006 M hydrochloricacid(HCl)usingsub-boilingdistilledHCl.TheICP-MS samplesweredrawninto250mLlowdensitypolyethylenebottles andacidifiedto0.024MHCl(Optima,Fisher).

Forshipboarddissolvedmetalanalysis,thesamplebottleswere storedinpolyethylenebags inthe darkatroom temperaturebe- foreanalyses, usually within 24 hof collection.Prior to analysis, samplesfordissolved iron (dFe) andmanganese(dMn) were mi- crowavedingroupsof4forthreeminutesat900Wto60±^10◦C inanefforttoreleasedFefromcomplexationinthesamples.Sam- pleswere allowed to coolforatleast1 h priorto FIA.Dissolved

Fe andMn were determined subsamples usingmethods of Mea- suresetal.(1995) fordFe andResing andMottl(1992) fordMn.

Sampleswere analyzedingroupsof8,andthesamplescollected ateachstationwere generallyanalyzedtogether during thesame day.Detection limitsandprecisionswere 0.071 nM and1.9% rel- ativestandard deviation at1.4nM for Fe, and0.15nM and9.7%

relative standard deviationat0.51nM forMn. Weconverted FIA results,normallyreportedinnMunits,tonmol kg⁻¹ inthispaper by dividingby the densityofthe seawaterat20^◦C andits mea- suredsalinity.

Ironandmanganesewereextractedfromtheseawatersamples forICPMSanalysisandpre-concentrated25xusingaSeaFASTpico- automatedsystem(ElementalScientific),followingisotopespiking with⁵⁷FeandbufferingtopH∼^6.2,usinganoffline-modifiedver- sionofLagerstrometal.(2013),asreportedbyJensenetal.(2020).

ExtractswereanalyzedatmediumresolutiononaThermoElement XRICP-MS.Ironwasquantifiedusingisotopedilution,andMnus- ing matrix-matched standard curves that were set to match the concentrationrangeofthesamples.Allsamplesexceedthedetec- tionlimitsof0.036nmol kg⁻¹ and0.002nmol kg⁻¹ forFeandMn respectively.Externalaccuracywasconfirmedbyreplicateanalyses of the GSC-2 seawater solution, which were found to have con- centrationsof1.631 ± ^{0.072nmol kg−1 forFe and2.140}± ^0.129 nmol kg⁻¹ forMn(n=^{6 each),}statisticallyindistinguishablefrom consensus concentrations (1.535± ^{0.115nmol kg−1 and2.180}± 0.075nmol kg⁻¹,respectively).

AttheLoihistation,comparisonofFIAandICP-MSresultsshow goodagreement whenboth were sampledfromthe samebottles intheGTCcastandfallona1:1line(Fig.2a)withtheexception of one sample, which may have beencontaminated. Comparison ofFIAandICPMSresultsfortheneighboringstations(Fig.2b)also show goodagreement.However,atLoihi, onlyFIAmetalanalyses were done on the Niskin casts (black symbols/lines) where ³He was also collected. As discussed in Section 3.1, there was a pro- nounced difference in light transmission between the two casts below1200mthatindicatesasignificantdifferenceinhydrother- malinfluenceonthesampledwatersbetweencasts.

Dissolvedmethanewas determinedusingaheadspace equilib- ration-cavityring-down spectroscopymethod(RobertsandShiller, 2015). Samples were collected immediately after the other dis- solved gas samples, equilibrated withmethane-free zero-air, and head-space methane determined using a Picarro G2301 Green- houseGasAnalyzer.Calibration wasperformedusingbothmetha- ne-free zero air and a 5 ppmv CH₄ gas standard. Variability of dissolved methaneindeepocean samplesfromthe GP15section suggests a standard deviation of better than 0.06 nmol kg⁻¹ on samplesof ∼^{0.6 nmol kg−1,similar tothe resultsof Robertsand} Shiller(2015).

Wealsocharacterizedhydrothermalinputbydeterminingtotal dissolvedsulfides(TDS)andpH.TDSsampleswerecollectedusing the GTC,filteredthrough thesame 0.2 μmcapsules asthemetal samples,and storedin4 litercubitainers. Filteredwater samples weretransferredtoglassstrippingvessels,acidifiedtoapHof1.6 and analyzed atsea usually within 8 h of collection using cryo- genictrappingandGC/flamephotometricmethod(Radford-Knoery and Cutter, 1993) in duplicate. This method quantifies total dis- solvedsulfideasfreeionsandmetal-sulfidecomplexeswitha0.2 pM detectionlimit.TodeterminepH ateach depth,60mL offil- tered waterwere subsampled fromthe GTC using a syringe and thee-way valve. This water was hermetically transferred into an automated Ocean OpticsUV-VIS spectrophotometric system, pure m-cresol purple added,andabsorbance measured atthree wave- lengths(Carter etal.,2013;ClaytonandByrne,1993).Thesesam- pleswereusuallyanalyzedwithin2hofcollection.

4 W.J. Jenkins et al. / Earth and Planetary Science Letters 539 (2020) 116223

Fig. 2.Inter-calibrationofdissolvedFemeasurementsdrawnfromthesame(GTC)bottlesmadebyshipboardFIA(y-axis)andICPMS(x-axis)forthe(a)theLoihihydrocast andfor(b)neighboring(“distal”)stations(Stations14,16,18,19,21,andPunaRidge).Theorangelineshaveaninterceptofzeroandunitslope.

3. Resultsandmetal-³Heratios 3.1. TheLoihistation

ThestationovertheLoihiSeamountwaslocatedat18^◦54’23”N, 155^◦15’29”W,withanominaldepthof1300m,directlyabovethe centerofPele’sPit(Clagueetal.,2019),acollapsefeatureresulting froma1996seismic/eruptionsequence(Duennenbieretal.,1997).

The target samplinglocation was within ∼^{100m ofthree active} hydrothermalvents (Tower, Hiolo, andSpillway) to its southeast, and∼^{300mofashallowervent (Lohiauatabout1150mdepth)} tothenorth-northwest(Bennettetal.,2011;Germanetal.,2018).

Measured isotope and element profiles for samples drawn from the Niskin hydrocast are shownin Fig. 3 (a-f), along with prop- erties measured onthe GTC casts(g-j), aswell as thecalculated profilesofexcessheliumisotopes(Fig.3k-l)andtheirisotopicrela- tionship(Fig.3m).Exceptforneonandmethane,allelementsand isotopesfromtheNiskinhydrocastshowthesamestructurebelow 1000m,with maxima∼^{100m abovethe seafloor andelevated} concentrationsextending∼^{200mabove.TheGTChydrocastmea-} surementsexhibitthecontinuedincreaseinsignal totheseafloor (redlines/symbolsinFig.3g-h).Thetrendisalsoseeninthetotal dissolvedsulfide(Fig.3i)andmirroredbythepH(Fig.3j),further reflected inadecrease inlight transmissivitybelow1250 m (not shown).Thedifferencesbetweenthetwohydrocasts,despitepre- cise ship positioning isexpected because a modest5^◦ difference inwireanglecanresultina∼^100mlateraldisplacementpackage towardoraway fromthethree nearbyactive hydrothermalvents inPele’sPit.

Helium isotope ratio anomalies (Fig. 3a) approaching 400%

were observed, comparable to earlier observations at Loihi (Lup- ton,1996). Correspondingto thismaximumwas a ∼^{22% peak in} heliumsaturation anomalies (Fig. 3b),in contrastto neon,which showedno difference fromother stations. The small Ne excess at1100mislikelyasamplingartifact producingaminisculebias in the helium, further corrected for by the isotope calculations.

Thus,theheliumsupersaturations arenon-atmospheric.LikeLup- ton (1996), we find the isotopic ratio of the Loihi helium to be substantiallyelevatedcomparedtotypicalMidOceanRidge(MOR) sources (27.3 ± ^{0.3 R}a vs ∼^{8 R}a where Ra is atmospheric ratio 1.384×^{10−6) (Clarke etal., 1976)). Ourresult is slightly higher} thanLupton’s(1996) originallyreported23.5Ra.(Wecontendthat Lupton inappropriately includednear-surface watersasendmem- bers,whichflattenstheslopeslightly).Ourratioismoreconsistent

withobservationsofLoihisummitlavas(Kaneokaetal.,1983;Kurz etal.,1983;RisonandCraig,1983),asignatureofprimitiveman- tle volatilesassociatedwiththeleading edgeoftheHawaiianhot spot(Kurzetal.,1982,1983).

The temperature data for this station revealed no detectable thermalsignatureatthedepthsof³Heanomalies,consistentwith the veryhigh³He:heatratiosreportedfortheLoihi seamountby Lupton et al. (1.1 fmol J⁻¹) (1989). This contrasts with ratios at leasttwoordersofmagnitudesmallerformid-oceanridge(MOR) systems (see Elderfield and Schultz, 1996; Lupton et al., 1989).

Thisdifference resultsfromtheprofound geo-tectonicdifferences between amid-plate mantle-plume fueledvolcanoin the shield- buildingstageandMORspreadingcenters.WhileMORhavechar- acteristically shallow magma chambers (∼^{1 km below seafloor),} Loihi’smagmachamberis∼^{10timesdeeper,providinggreaterop-} portunity fordecoupling between primary volatiles like ³He and heat(cf.Hofmannetal.,2011).

High concentrations of dissolved methane are commonly ob- servedinhydrothermalventwatersandappeartobederivedfrom reductionofCO2toCH4underconditionsassociatedwithplutonic rocks andtemperatures of ∼^{300-400◦C (Wang et al., 2018). For} the Loihi profile, the enrichment in CH4 is modest, less than 1 nmol kg⁻¹ relative at similar depths at other stations in the re- gion (∼^{1 nmol kg−1,Fig. 3f. These values are} significantly lower than theroughly 5-15nmol kg⁻¹ reportedbyGamo etal.(1987) priortothe1996seismic/volcaniceventsandthecollapseofPele’s Vents. Incontrast,dissolvedmethanein otherhydrothermalend- membersolutionsareofteninthemicromolartomillimolarrange.

Because theprocesses offormationofhydrothermalmethane are differentfromthegenerationofhydrothermal³HeorFe,wedon’t necessarilyexpectbroadoceaniccorrelationofmethanewiththese other hydrothermal parameters. Nonetheless, the comparison is potentially instructive.Forour two most methane-enriched Loihi samples, we observe xsCH₄/C_XS(³He) molar ratios of ∼⁵×^104, much smaller than the ∼⁵×^{105 ratio reported by Gamo et al.}

(1987). Here we define XSCH4 as theamount of methane above values observed at comparable depths away from Loihi. In far greater contrast, Keir (2010) reported xsCH4/³He ratios in mid- ocean ridge hydrothermalsamples thatwere 10-1000-foldhigher than our value. Likewise, the xsCH₄/Fe ratios of our two most methane-enriched samples are much lower than xsCH₄/Fe ratios observedinmostotherventfluids(e.g.,Charlouetal.,2002).This suggeststhateithertheLoihimethanehadbeenrapidlyconsumed

Fig. 3. Loihiprofilesofa),heliumisotoperatioanomalyδ³Hein%,b),dissolvedheliumsaturationanomalyin%,c),dissolvedneonsaturationanomalyin%,d),dissolved Feinnmol kg⁻¹,e),dissolvedMninnmol kg⁻¹,f),dissolvedCH4innmol kg⁻¹,g)dissolvedFe(ICPMSforGTChdyrocastonly,andFIAforboth),h),dissolvedMn(sameas Fe),i),TDS(pM)fromtheGTCcast,j),pHfromtheGTCcastk),excess³Heinfmol kg⁻¹,l),excess⁴Heinpmol kg⁻¹,andm),correlationbetweenexcess³Heand⁴He.The dashedlinerepresentsthebottomdepthatthesamplinglocation.

bymethanotrophsduringtransportanddilutionfromtheventsto our samplinglocation, or that the Loihi systemis nearly devoid ofmethane.Indeed,basalt-hostedventsystemstendtohavelower methaneconcentrationsintheirhydrothermalfluidsthandoultra- mafic hostedsystems (Keir,2010),though their xsCH4/³He ratios stilltendtobe10-100-foldgreaterthanwhatweobserve.

ThemaximumdFe anddMnconcentrationsmeasuredforboth hydrocastsatLoihistation(Fig.3d-eandg-h)wereroughlyhalfof thosereportedearlierby Bennett etal.(2011) forPele’sPit.This couldbeduetospatialvariations(seepreviousdiscussioninmeth- ods), butGerman et al. (2018) have described a gradual, decade time-scaledecreaseinventtemperaturesatLoihi.Nonetheless,the

maximumdFeisabout7timeslargerthanobserveddirectlyover the southern East Pacific Rise (SEPR) (Fitzsimmons et al., 2017;

Resingetal.,2015),whilethedMnisroughlytwofoldsmaller.

The overallslope ofthe“proximal”Fe:³He relationshipwithin Pele’sPitwas 7.4±⁰.5×^{106 (Fig.4a),}surprisinglysimilarto the SEPR. Since the published SEPR relationship (Resing etal., 2015) was based on preliminary shipboard dFe measurements, we use the average of shore-based ICP-MS measurements within 50 m of the coreof the ³He maximum for the station over the SEPR (Schlitzer et al., 2018) to obtain a revised proximal SEPR Fe:³He ratio of7.6×^{106, virtually identical to Loihi.This is remarkable} since theoveralldFe and³Heconcentrationsatthe twositesare

6 W.J. Jenkins et al. / Earth and Planetary Science Letters 539 (2020) 116223

Fig. 4. TheobservedcorrelationsattheLoihistationbetweenexcess³Heinfmol kg⁻¹anda,dissolvedFeandb,Mn,bothinnmol kg⁻¹,andc-d,molarratioprofilesof dissolvedFe,Mn,andemethaneto³He.Notetheshallowestplumevalues(redstar)whichlikelyrepresentratios“exported”fromthesite.

sodifferent, andtheprocesses leading totheir creationare com- pletely decoupled.Thatis,thedFe enrichmentatLoihi isderived from low-temperature leaching of basalt by hotspot-derived vol- canic carbon dioxide, which drives down the pH of Loihi fluids (Sedwick et al., 1992). In contrast, the dFe enrichment at mid- oceanridge systemssuch asSEPR isdependentonhot (>300^◦C) reactions ofacidic fluidswithFe silicates inbasalts (Von Damm, 1985).

The overall Mn:³He relationship (Fig. 4b) is 0.8±⁰.1×^106, morethan twofold smallerthan observed atthe SEPR (Resinget al.,2015).ThedepthprofilesoftheFe:³Heratio,showninFig.4c, ishighest∼^{100mabovetheplume’score.Incontrast,theMn:3He} ratioprofile,showninFig.4d,showsanincrease belowthecore, resemblingmore the CH4:³He ratio profile shown inFig. 4e. Al- thoughdivergenceoftheFe:³Heprofilebelowthecoreofthe³He maximumcouldresultfrominsituscavenging,considerationofthe detailedbathymetry and nearbyvent locationsin relationto the samplingsite(e.g.,seeFig. 1bofBennettetal.,and4bofClagueet al.Bennettetal.,2011;Clagueetal.,2019),suggeststhatthepro- filesarethecompositesignatureofmultiplehydrothermalsources at different locales anddepths in and near Pele’s Pit. Moreover, sincethesilldepthofPele’sPitis∼^{1100m,themetal:3Heratios} thatareexportedtoinfluencesurroundingopenoceanwatersare likelythevaluesatornearthe1100mdepth.Thus,weadoptthe somewhathighervaluesof9.3±⁰.3×^{106 and0}.93±⁰.04×¹⁰⁶ forthe molar Fe:³He and Mn:³He ratios that escape to feedthe regional-scale“hydrothermaltongue”.

3.2. The“distal”stations

The GEOTRACES GP15 δ³He and dFe profiles (Fig. 5) showed distinctmaximaat1100mdepth forthe stationswithin 500km ofLoihi(18,PunaRidge,and19),whilemoreremote“background”

stations(14, 16,and21),falling outsideof themainLoihi plume

(Fig.1aand1c)donotexhibitdiscernible features.Concentrations of excess³He, dFe,anddMnat allsixstations were interpolated to the 1100m depth, and decrease monotonically with distance fromLoihi(Fig.6a-c),withtheexceptionofstation14wheredFe and dMn appear elevated (and dissolved oxygen more depleted) due tothe influence ofthe NorthPacific oxygen minimum zone.

There, a secondary subsurface maximum ofreduced metalssuch asMn(II)ispresentduetotheirredoxstabilityundersuboxiccon- ditions(Johnsonetal.,1996).

The dissolvedmetal relationshipsto³HeareplottedinFig.6d and 6e. We invoked a linear dFe:³He relation because the data at the critical intermediate concentrations were insufficient in number to distinguish from a non-linear relationship. Included in the plot is an exponential regression (dashed line) that fits the data well, but has only one degree of freedom in the fit.

The linear dFe:³He regressionslope for the fivedistal stations is 4.4±⁰.9×^{106, about a factor of two lower than the proximal} ratio of 9.3±⁰.3×^{106.This decrease in slope iscomparable to} that observed over the SEPR (Fitzsimmonsetal., 2017). At Loihi, the proximalrelationship arisesfromnon-conservativedFe losses during plumetransport,whichcouldincludeabioticscavengingof Fe(II) butalsomayincludemicrobialFeutilization, giventherich Fe-oxidizing microbial communities known to exist in Loihi mi- crobial mats (Emerson and Moyer, 2002; Emerson et al., 2007).

The distal dMn:³He relationship (Fig. 6e) of0.9±⁰.1×^{106 (ex-} cluding Station 14,which is largely unaffected by Loihi)appears equaltotheproximalratio,whichisconsistentwithSEPR(Fitzsim- monsetal.,2017)andearlierstudies(JamesandElderfield,1996) thatdMnbehavesconservativelyinthenon-buoyanthydrothermal plume.

The similarity ofboth proximal and distal dFe:³He ratios be- tweentheLoihiandSEPRisremarkablefortworeasons.First,³He is a primary volatile that is strongly partitioned out of the par- entmagmawhiledFeisaffectedbywater-rockexchange,chemical

Fig. 5. Compositedepthprofilesatthedistalstationsofa,δ³Hein%,b,dissolvedFeinnmol kg⁻¹,c,dissolvedMninnmol kg⁻¹,andd,dissolvedoxygenin μmol kg⁻¹for themoredistantstationsshowninFig.1(notincludingtheLoihistation).Fromnorthtosouththeyarestation14(cyan),16(blue),18(magenta),PunaRidge(red),19 (green),and21(yellow).Notethepronouncedmaximainδ³He,dFe,anddMnat1100mdepthidentifiedinδ³HebyLupton(1996) astheLoihiplumeforthe“nearby”

stations(PunaRidge,18,and19).Also,notetheaberrantbehaviorforthedMn(andtoalesserextentdFe)forstation14.

Fig. 6. Thedistalexcess³He,dissolvediron,andmanganeseconcentrationvariationvs.distancefromLoihifora.excess³Heinfmol kg⁻¹,b.dissolvedFeinnmol kg⁻¹,and c.dissolvedMninnmol kg⁻¹.Correlationwithexcess³Heford.dissolvedFeandfore.dissolvedMn.Errorbars,whichcombineanalyticalandinterpolationuncertainties, are1σforbothvariables.Thesolidlinesarefromtype-IIweightedlinearregressions(excludingstation14,highlightedinpinkforFeandMn).Thedashedlineindisan exponentialfitforcomparison.

8 W.J. Jenkins et al. / Earth and Planetary Science Letters 539 (2020) 116223

Fig. 7. Themid-depthexcessheliumisotopesrelationshipsforthebackgroundandthreeneareststations.Thefilledcirclescorrespondtosamplescloseto1100mdepth.

aStations14,16,and21areconsidered“background”andtheblueline,whichfallsthroughthesepointsisforreferenceonly.b-dareforthethreecloseststations(18,19, andPunaRidge,seeFig.1a).Thebluestaristheestimatedunperturbed1100mvalueandthereddashedline’sslopecorrespondstotheisotopiccompositionofLoihiplume helium(seeFig.2i).

precipitation,microbial transformation, andthe presence/absence of binding ligands. Thus, given these very different geochemical supply mechanisms to hydrothermal fluids, it is remarkable that Fe and ³He wouldproduce similar ratios in different vents. Sec- ond, Loihi and the SEPR have vastly different geochemical, geo- logical,hydrothermal, andtectoniccharacteristics,making iteven more unlikely that Fe and ³He would have similar ratios. The SEPR is a mid-ocean ridge spreading center with a shallow (∼1 km depth) magma chamber, tholeiitic basalts, andhigh temper- ature hydrothermalvent fluids, while Loihi is a mid-plate, man- tle plume-driven shield-buildingvolcanowitha deeper(∼^10km depth)magmachamber,andlowertemperaturehydrothermalvent fluids(Germanetal.,2018;Rouxeletal.,2018).Furthermore,SEPR vents have millimolar concentrations of H2S (Von Damm et al., 2003), which should minimize dFe concentrations dueto precip- itationofironsulfides(Feelyetal.,1996,1987).Weobservedless thannanomolarTDSconcentrationsatLoihiandonlymodestacid- ification,minimizingdFeremoval.

Incontrast toboth Loihi andtheSEPR, recentobservationsin theNorth AtlanticTAGhydrothermalplume showdFe concentra- tions (Fitzsimmons et al., 2015) of ∼^{60 nmol kg−1 with a cor-} responding ³He excess (Jenkins et al., 2015) of ∼^{0.6 fmol kg−1,} leadingtoamolarFe:³Heratiocloseto100×^{106.Thisisaboutten} tofifteentimeshigherthanobservedhereandattheSEPR.Obser- vationsoverthesouthernMid-AtlanticRidgeataround13^◦Syield asimilarlyhighvalue of∼⁷⁰±³⁰×^{106 accordingto Saitoetal.}

(2013).Thus,thereappearstobeagreaterAtlantic-Pacificdispar- ityindFe:³Heratiosthanbetweenmid-oceanridgeandmid-plate

mantle plumeswithin the Pacific.Thisverifies thatthere arenu- merous factors fixing the persistent Fe:³He ratiosupplied to the ocean by any single hydrothermal vent, including tectonic con- text, depth of themagma chamber (and subsequenttemperature atventing),hostrockcomposition,microbialcomposition,andbot- tomwaterconditions(especiallyoxygenation).Thenumericalpre- diction of³He and persistentdFe fluxesfromhydrothermal vent fluidsisacomplextask.Weattempttoquantifythosefluxeshere forLoihiSeamount.

4. Estimatingthehydrothermal³Heflux

Loihi’suniquelyhigh³He/⁴Hesignatureprovidesanunambigu- ous and quantitative measure of the amount of excess ³He at- tributable to Loihi at each station, as distinct from the regional

“background” ofvolcanicandradiogenichelium emittedby more distant ventsand sediments.InFig. 7 we show therelationships between excess ³He and⁴He for the more distant “background”

stations (14, 16, and 21 in Fig. 7a) and the three stations near Loihi(18,19,andPunaRidgeinFig.7b-d).Intheneighboringsta- tions,thereisacleardeviationfromthebackgroundslopedefined bythemoredistantstationsthatallowsustoquantitativelysepa- rateouttheLoihicontribution.Inallpanels,theshallowestpoints areontheleft anddeepestpointsontherightofthegraphs.The blue linereferences thebackground trendandthe bluestar rep- resents a “best guess unperturbed” ³HeXS-⁴HeXS composition for the1100mdepth.Intheotherpanels,thereddashedlinecorre- spondstotheunique³He/⁴HetrendlineassociatedwiththeLoihi

Fig. 8. Profilesofa)excess⁴Heinnmol m⁻³,b)excess³Heinpmol m⁻³,andc)Loihi-linkedexcess³Heinpmol m⁻³forsixGP15stations.Thesolidlinesinc)aresmoothing splinefitstoalldataineachprofile.Weonlyintegratedoverthedepthrangeofintegration(900to1300 m).d)Computed³Heinventoriesinpmol m⁻²forstationsalong

∼152^◦Wvs.distancefromLoihi,includingbothGP15andtherecent2015occupationofP16N.

mantleplume signature(27.3R_a),whichdemonstratesconvincing evidence that the 1100 m features (filledcircles) arise from the Loihisource.Thisgivesusaquantitative methodfordefining and integratingthe watercolumnexcess ³Heattributableto theLoihi source.

Fig.8a and8b show theexcessheliumisotopeprofilesforthe GP15stations, calculatedusingequations (3) and(4), andFig. 8c showsthe Loihi-sourced ³He profile as estimatedby subtracting theaverageof“background” stationprofilesbetween5and10^◦N (station21) andbetween30 and35^◦N(stations 14and16). The heliumisotopeprofilesnearestLoihiexhibit asignificant³Hesig- nature between 900 and 1300 m depth with a maximum near 1100 m.Thelargeverticalspreadmaybedueinparttothetopo- graphicdistributionofthemanyhydrothermalventsites(Bennett etal.,2011;Clagueetal.,2019;Rouxeletal.,2018)atLoihi.Inad- dition,althoughdiapycnalmixingintheopenoceanisweak(Led- well etal., 1998, 1993), seamounts are known sitesof enhanced vertical mixing(Lueck andMudge, 1997; Toole et al., 1997). We thusarguethattheverticalspreadlikelyoccursasaresultofboth verticallydistributedsourcesandenhancedverticalmixinginthe vicinityoftheinjectionsites.

Wemodeltheregionaldistributionoftheverticallyintegrated 3Heanomaly(theinventory)between900and1300mdepthsus- ing the observations reportedhere along withselected historical profiles. Uncertainties are assigned as a combination of analyti- cal precisions and interpolation errors. The latter, estimated by

assuming aprofile shapeintermediate betweena box-car(acon- stantvalueovertherange)andatriangulardistributiondominates.

The inventories, which include results from a recent occupation along thismeridian (P16Nin2015)showa smoothtrendvs.dis- tance(seeFig.8d),supporting theideathatthefeaturedescribed is persistent.Usinghistorical data,we can furtherdevelop a pic- ture of the distribution ofthe vertically integratedLoihi-sourced 3Heanomaly.Forthoseprofileswithcontemporaneousneoncon- centrationmeasurements,similarcalculationscanbedone,butfor those without neondata (notably the more distant WOCE P17N section at135^◦W)a much moreapproximate scheme(Jenkins et al.,2018)mustbeused,withsignificantlylessaccuracy.

This model rests on the assumption that the existence of a persistent regional ³He feature emanating from Loihi requires a sustainedhydrothermalfluxofthisisotope.Themagnitudeofthe required flux depends on the regional scale velocity and turbu- lent diffusionresponsible forits dispersion. We usean idealized advection-diffusion model to set approximate constraints on this flux. We constructed a two-dimensional rectangular slab-channel modelforthisdepthrange,spanningfromthedatelineto120^◦W and between 10^◦ and 30^◦N. We impose a uniform zonal veloc- ityanda spatially variablebut isotropichorizontalturbulent dif- fusivity. Horizontal resolution is 25 km, and a standard, mass- conserving,upwind-differencing algorithm(Glover etal., 2011) is usedtointegratethemodeltosteady-state.Choiceofthezonalve- locityiscriticalinthattheresultswillscalealmostlinearlywithit.

10 W.J. Jenkins et al. / Earth and Planetary Science Letters 539 (2020) 116223

Weuseavelocityof0.0040±^{0.0016m s−1obtainedforthisloca-} tionanddepthfromaglobalmulti-tracerdataassimilationmodel calibratedby radiocarbon andCFCs(Schlitzer, 2004, 2007) which hasbeen used successfullyto model regional/globaldistributions of radiocarbon and ³He (Schlitzer, 2007, 2016). This velocity is consistent witha recent analysis ofcirculation at 1000m depth fromfloatdatabyOllitraultandDeVerdiere(2014,their Fig. 12), andweconservativelyestimateanuncertaintyof40%tothisvalue.

ThepresenceoftheHawaiianIslandchaininthepathofthiszonal currentmaylead tospatialenhancement oflateraldiffusion(e.g., Chenetal., 2015). Thus,we imposea simplespatial structureon thelateral diffusivityasthesumof aconstant backgrounddiffu- sivityplus a halfsine function centered on 20^◦N,150^◦W whose valueiszeroat10^◦ and30^◦N,andat180^◦ and120^◦W.Themag- nitudesofthebackgroundandamplitudediffusivitiesweretreated asvariablestobeoptimizedagainstobservations.

Toconstrain themodel,we compute integrated watercolumn inventories between 800 and 1300 meters depth of the Loihi- attributed³He(based onthe³He/⁴He anomaliesandcalculations described above) using only the more recentlyoccupied stations (2015and2018)thathavenoblegasmeasurementsalongthe152

◦W section. Forthe 135^◦W section,due tothe lack ofnoblegas data,weusealessprecisecalculation (Jenkinsetal.,2018) using thehistoricalheliumisotopemeasurements largelymadebyLup- ton (1996,1998).Forweightingpurposes,weestimateduncertain- tiesintheseinventoriesdueto thecombinationofmeasurement andinterpolation/integrationerrors.

We embedded the advection-diffusion model within a non- linearoptimizationalgorithmthatvariedtheLoihisourcestrength (released at thecell nearest theseamount’s location) along with boththebackgroundandpeakhorizontalturbulentdiffusivitiesto minimize the model mismatch to the observed distribution. We characterizethemodel-datamisfitbycomputingthereducedchi- squared

χ

_ν²

=

¹ N

−

3

i

(

Mi

−

^Oi

)

²

σ

_i^{2 (5)}

whichisthesumofthesquareddifferencesbetweenmodel(Mi) andobservation(O_i)points normalizedbytheuncertainty(

σ

i) of the observationsdivided by the numberof constraining observa- tionsminusthenumberofadjustableparameters(e.g.see(Glover et al., 2011)). Since it is difficult to present the distribution of thispropertyinthree-dimensionalparameterspace,wepresentin Fig. 9 various “slices” in2D (Fig. 9a)and 1D (Fig. 9b-d) formats along with the optimal modelLoihi ³He inventory spatial struc- turewiththeconstrainingdata(Fig.9e).

The

χ

_ν² distributionasafunction ofsourcestrengthandmax- imum (peak) diffusivity is shown in Fig. 9a for a globally opti- mal background diffusivity of 405 m²s⁻¹ (see Fig. 9b). The

χ

_ν²

contours in Fig.9a show a sloping ellipticalminimum indicating a broad correlation betweensource flux anddiffusivity: increas- inglateraldiffusivitynecessary to“spread”theeastward oriented tonguemeridionallyrequiresanincreaseinsourcefluxrequiredto support the overall inventory. The global minimum

χ

_ν² was 1.5, indicating good agreement with observations, especially consid- ering the simplicity of the model. The optimal case peak diffu- sivity was 645 m²s⁻¹ (see Fig. 9c) and the optimal source flux (Fig.9d)was 10.43mol a⁻¹.Theoptimalsourcestrengthdepends mostontheaggregatehorizontaldiffusivity:inspectionofFigs.9b and 9d reveals that required source strength is relatively insen- sitive tothe value ofbackgrounddiffusivity, changingonly about

±^6%fora±^{50%changeinbackgroundmixingrate.Asonereduces} thebackgroundmixingrate,theamplitudeoftheislandwakeen- hancement(characterizedbythesinusoidamplitude)increasesto compensate,maintainingaroughly constantaverage diffusivityin

theplume.Infact,choosingaspatiallyconstanthorizontalmixing ratehasonlyamodestinfluenceontheoverallresult.

Fig. 9e shows model ³He inventory distribution for the op- timal case, along with the constraining observations. The model parameters, ³Hesource strength (10.4 mol a⁻¹) alongwith back- ground(405m²s⁻¹)andpeak(645m²s⁻¹)horizontaldiffusivities scale linearlywiththehorizontalvelocity, soassuming anuncer- tainty ±^{40%inthevelocitytranslatestoa}corresponding±^40%in thoseparameters.Thusweestimatethe³Hefluxtobe10.4±^4.2 mol a⁻¹,apeakhorizontaldiffusivityof645±^{258m2s−1,anda} backgrounddiffusivityof405 ± ^{162m2s−1. Therangeofproba-} blehorizontaldiffusivities agreeswell withmanyother estimates of intermediate depth open-ocean values on regional scales (for example,Jenkins,1991; Ledwelletal.,1998;Robbinsetal.,2000;

SundermeyerandPrice, 1998).Moreover, elevated isopycnalmix- ing should result from the expected enhancement ofdeep eddy energetics around the Hawaiian Island and seamountwakes. We thereforeregard the predictedfluxesandhorizontaldiffusivityas consistentwithcurrentunderstandingofthoseprocesses.

As a slab model, these calculations are relatively immune to the effects of vertical (diapycnal)mixing becausethe inventories used toconstrainthe modelarevertically integratedoverseveral hundred meters. In thisway itcompensates forthe initial verti- cal dispersion caused either by enhanced vertical mixing in the vicinity of rough topography (e.g.,Ledwell et al., 2000; Polzin et al., 1997) orby multiplehydrothermalinjectionsites atdifferent depths. We can, nonetheless,place an upperlimit on the poten- tial impact of diapycnal mixing by combining our knowledge of the verticaldistributionofthe Loihi³He withareasonablerange ofdiffusivities.Estimatesforopenoceandeepwatersaretypically oftheorderof1×^{10−4 m2s−1 (Munk,1966),while}tracer-based estimates in the main pycnocline are of order 2×^{10−5 m2s−1} (Ledwell etal.,1998,1993;Rooth andOstlund, 1972). Weexpect anactualdiffusivitysomewhereinbetweenthesetwoatinterme- diate depths.Using anaveragevertical ³Hegradient of2×¹⁰⁻¹⁶ mol m⁻⁴ inthe “core” ofthe horizontal Loihi heliumplume and anapproximatearealextentof1×^{1012m2,wearriveatafluxof} 3−¹⁵×^{10−9 mol s−1 or 0.1– 0.5 mol a−1. This} corresponds to lessthan5%oftheestimatedflux,smallcomparedtootheruncer- tainties.

The mostprobable³He sourceflux of10.4mol a⁻¹ isremark- able, beingof order of2% ofthe 548 ± ^{31 mol a−1 globaldeep} hydrothermal ³He flux (Bianchi et al., 2010; DeVries andHolzer, 2019;Schlitzer,2016).Whileseeminglylarge,netplumebuoyancy fluxes indicate that the Hawaiian hot spot contributes ∼^{15% of} theglobalmantleplumebuoyancyflux(Sleep,1990) andsinceat somelevelinthemantleheatand³Heshouldbecorrelated,asim- ilarly large fractionof themantle plume ³He flux. Becausesome portion ofthe Hawaiian plume volatile release will be subaerial (e.g.,throughKilauea, whichhasa³He/⁴Heratiointermediatebe- tweenLoihiandasthenosphericvalues),ourresultsuggestsalower boundglobalplume³Hefluxof70mol a⁻¹,about13%ofthemid- ocean ridge flux (Bianchi et al., 2010; DeVriesand Holzer,2019;

Schlitzer, 2016). Givenan estimate ofthe currentmagma supply rateof2.5×^{107 m3a−1 forLoihi(Garciaetal.,2006)andassum-} ing virtually completevolatile loss,we estimate the un-degassed magmatic ³He concentration of4×^{10−7 mol m−3 (which corre-} spondsto∼3.5×10⁻⁹cc-STPg⁻¹).Thisisapproximately100-fold more thanthe largestreportedconcentrations observedinglassy rims of extruded flows on Loihi (e.g., Kurz et al., 1983), but it is consistent withthe magmaticdegassing concepts (e.g.,Chavrit et al., 2012). Such a disparity can be expected, both due to ev- idence atLoihi ofextensivedegassing prior toeruption (Byers et al.,1985),andduetotheverylarge³He:heatratioobservedforthe Loihi hydrothermalsystems(Elderfield andSchultz, 1996;Lupton etal.,1989).Inthelattercase,theroughly100-foldenhancement