JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 98, NO. Bll, PAGES 19,533-19,563, NOVEMBER 10, 1993

Petrologyand Geochemistryof the GaMpagosIslands' Portrait of a PathologicalMantle Plume

WILLIAM M. WmT•

Departmentof GeologicalSciences, Cornell University, Ithaca, New York

ALEXANDER R. McBIRNEY

Centerfor Volcanology,University of Oregon,Eugene

ROBERT A. DUNCAN

Collegeof Oceanography,Oregon State University,Corvallis We reportnew major element,trace element,isotope ratio, and geochronologicaldata on the Galfipagos Archipelago. erupted from the largewestern volcanos are generallymoderately fractionated tholeiites of uniformcomposition; those erupted on otherislands are compositionallydiverse, ranging from tholeiites to picritic basanitoids.While thesevolcanos do notform a strictlylinear age progressive chain, the agesof the oldestdated flows on anygiven volcano do form a reasonableprogression from youngest in thewest to oldestin theeast, consistent with motionof theNazca plate with respect to thefixed reference frame. lsotoperatios in theGalfipagos display a considerablerange, from values typical of mid-oceanridge on Genovesa(87Sr/86Sr: 0.70259, end: +9.4, 206pb/204pb:! 8.44), to typical oceanic island values on Floreana (87Sr/86Sr: 0.70366, œNd: +5.2, 206pb/204pb: 20.0). La/SmN rangesfrom 0.45 to 6.7; otherincompatible element abundances and ratios show comparable ranges. Isotope andincompatible element ratios define a horseshoepattern with the mostdepleted signatures in the centerof the GalfipagosArchipelago and the moreenriched signatures on theeastern, northern, and southern periphery. These isotopeand incompatible element patterns appear to reflectthermal entrainment of asthenosphereby the Galfipagos plumeas it experiencesvelocity shear in the uppermostasthenosphere. Both north-south heterogeneity within the plumeitself and regional variations in degreeand depth of meltingalso affect compositions. Rare systematicsindicate that melting beneath the Galfipagos begins in thegarnet peridotite stability field, except beneath the southernislands, where melting may occurentirely in the spinelperidotite stability field. The greatestdegree of meltingoccurs beneath the central western volcanos and decreases both to theeast and to thenorth and south. Sis. 0, FeB.0, andNaB. 0 values are generally consistent with these inferences. This suggests that interaction between the plume and surroundingasthenosphere results in significantcooling of the plume. Superimposedon thisthermal pattern producedby plume-asthenosphereinteraction is a tendencyfor meltingto be lessextensive and to occurat shallower depthsto thesouth, presumably reflecting a decrease in ambientasthenospheric temperatures away from the Galfipagos SpreadingCenter.

INTRODUCTION emergent volcanos in the Galfipagos,nine have been active historically and four othershave eruptedin the Holocene and Productionof basaltic magmasover stationary,long-lived should be considered still active. By comparison,only six hotspotsin oceanbasins is now generallythought to be a mani- Hawaiian volcanosare active,and theseare alignedin two chains festationof rising convectiveplumes in the mantle[Morgan, with a clear age progression. The GalfipagosIslands are also 1971]. The longevityof suchhotspots, their stationary nature, the exceptionalgeochemically: while mostbasalts have typical OIB simpleage progression of volcanismin the islandchains, and the incompatibleelement concentrationsand isotoperatios, geochemicaldistinctions between oceanic island basalts (OIB) that are essentially indistinguishable from MORB in their and mid-ocean ridge basalts (MORB) are among the most geochemistryare also commonproducts of Galfipagosvolcanos convincingevidence in supportof theconvective plume theory. [Baitis and Swanson,1976; Whiteand Hofmann,1978]. Not all oceanicislands conform to this simplepattern. Among We report the resultsof a petrologicaland geochemicalre- the most exceptionalisland groups is the Galfipagos,which connaissanceof the GalfipagosIslands, including Sr, Nd, and Pb consistsof 10 majorvolcanic islands and a numberof lesserrem- isotopic analysesof 124 samples,as well as trace and major nant volcanosthat emerge from a broad, shallow submarine elementanalyses, and 38 K-Ar age determinations.Our results platform(Figure 1). While two aseismicridges, the Cocosand suggestthat the uniquefeatures of Galfipagosvolcanism reflect Carnegieridges, extend from the Galfipagos Archipelago in thedi- the local plate tectonicenvironment, and thermal and dynamic rectionof Cocosand motions,the emergentvolcanos interaction between the asthenosphereand a mantle plume do not form a linear chain. Nor is there any simple geographic undergoingvelocity shear. Tectonicfactors control the location, patternto the agesof the volcanos,although the oldestvolcanos size, and styleof volcanism. Thermalinteraction between plume tend to occur in the southeastarea of the archipelago. Of the 21 and asthenosphereresults in asthenospherebeing entrainedinto the center of the plume. This, and the consequentplume- asthenospheremixing, producesmuch of the observedgeochemi- Copyright1993 by theAmerican Geophysical Union. cal variation. Superimposedon this mixing are variationsin depth Papernumber 93JB02018. and degreeof meltingwhich alsoresult from thermalinteraction 0148-227/93/93JB-02018505.00 betweenplume and asthenosphere.

19,533 19,534 WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OF THE GAL•PAGOS

93 ø W 92 ø W 91 ø W 90 ø W 89 ø W 88 ø W

DARWINOLFI. I. I CocosPlaTe

PINTA Q) reDONDA '•ENOvESA •ecuA•,,wøt••••'":•'•ARCHENA 0 0o

• ¾'""•"t''" '• '"'"'••""•'•••••'•:•--;'::-'•-:'::•:.:':'-•SEYMOUR •'"'•'••'••'••••••;. '" rABIDA BALtrA ß )•...:..::•:•:.:•:..::::-•'•u;::•::•.... :::.?• sterrA :'•'"••••••}•. '/.:-'•g4 '"R' :-:'-::-:;----:: ...... ::::?-•:.:':.': ' '• ';'...... :'" ...... ß...... : ' ••••'•2.¾;'::•::'::/.':•:::'::'::::;::.?:'.::..:'.:..... SANTA '[E ...... •••'• ...... '••'• '••'•'"-"'-"-'•1

FL ?...

m m m m m WHITMER m m m m m SMT. I I

93 ø W 92 ø W 91 ø W 90 ø W 89 ø W 88 ø W

Fig. 1. Map of the GalfipagosArchipelago. GSC is theGalfipagos Spreading Center. Boldline showsthe lithospheric faultsystem that separates thin, weak lithosphere to thenorth and east from stronger and thicker lithosphere to the south andwest inferred by M. A. Feighnerand M. A. Richards(submitted manuscript, 1993). Insetshows the regional tectonic setting.

GEOLOGICAL BACKGROUND AND PREVIOUS WORK concurrentlyto the productionof the Cocosand Carnegieridges, on the separatingCocos and Nazca plates, much as the Iceland TectonicSetting hotspotgenerates paired volcanic ridges on the Europeanand The GalfipagosIslands rise from a shallowsubmarine volcanic North Americanplates. With the northwardmigration of the ridge platform that is the westernterminus of the east-westtrending 5-6 m.y. ago, magmasupply to the CocosRidge was essentially CarnegieRidge. The GalfipagosSpreading Center (GSC), which severedand the output of the hotspotfocused on the western separatesthe Cocosplate from the Nazca plate andjoins the East CarnegieRidge. Dredgingand SeaBeam studiesof seamountson Pacific Rise in a triple junction some 1000 km west of the the CarnegieRidge revealedthat at least somewere onceislands GalfipagosIslands, bisects the region. The GSC is a moderately [Christie et al., 1992], extendingthe time availablefor evolution fast spreading ridge (•-6 cm/yr) with significantasymmetric of the uniqueGalfipagos biota back to at least9 Ma. The Coiba spreading over the last 5 m.y. along the section nearest the and Malpelo ridges are smaller blocks, dismemberedfrom the Galfipagos(85" to 91øW) with more rapid spreadingto the north eastern end of the Carnegie Ridge and transportednorthward [Hey, 1977]. Spreadingelsewhere along this plateboundary has during Nazca-SouthAmerican plate collision [Malfait and generallybeen symmetrical. North of lsabelaIsland, an ill-defined Dinkerman, 1972; van Andel et al., 1971]. Basementvolcanic ages left-lateraltransform fault at approximately90ø50'W offsetsridge from theseridges support the hotspotmodel [van Andel et al., segmentsby 150-km, and is the boundarybetween symmetric 1973]. spreading ridge segments to the north and west, and the M. A. Feighnerand M. A. Richards(Lithospheric structure and asymmeti'icsegment to the southand east. Thisresults in a 3-5 compensationmechaism of the GalfipagosArchipelago, submitted m.y. age differencein the lithosphereunderlying the westernand to Journal of GeophysicalResearch, 1993) haveanalyzed gravity easternhalves of the archipelago. data from the GalfipagosPlatform and find that the centerof the The GSC has migratednorthward away from the hotspotin the archipelagois underlain by weak lithosphere,with an elastic last5-6 Ma, judgingfrom the volcanic traces of thehotspot, the thicknessof 6 km or less,and is closeto Airy compensation.The Cocos and Carnegie ridges [Hey, 1977]. Before that time, the western and southernportions of the platform are flexurally hotspotlay directly beneaththe GSC and contributedmagmas supportedby a lithospherewith an effectiveelastic thickness of WHITE ET AL.: PETROLOGYAND GEOCHEMISTRY OF THE GAL•PAGOS 19,535

about 12 km. The transition from strong to weak lithosphere central shield volcanos with well-developed calderas,but they appearsto be abrupt and can be modeledas an arcuatefault that differ conspicuouslyfrom the familiar shields of Hawaii and runs beneathEspafiola and Floreanaislands in the south,beneath Reunion [McBirney and Williams, 1969]. Their slopessteepen easternmost Isabela in the west and intersects the Wolf Darwin abruptly betweena flat shelf closeto sea level and a calderarim, Lineament, a northwest-southeast chain of small volcanos, of giving them a distinctive"inverted soup-bowl" outline. Eruptions which only Wolf and rise abovesea level, northeast occur either from circumferential fissures around the calderas, of Santiago(Figure 1). The westernboundary between strong and which are largely independentof the caldera-relatednormal faults, weak lithosphereessentially corresponds to an extention of the or from radial vents and fissures lower on the flanks [Chadwick 90ø50'W fracture zone, so that some of the difference in and Howard, 1991]. Sirekin[ 1984] arguedthat the unusualprofile lithosphericstrength may be attributableto the lithosphereto the of the volcanos reflects the dominance of eruptions from the west being 3-5 m.y. older than that to the east. Gravity modeling circumferential fissures. Alternatively, Cullen et al. [1987] suggeststhe Wolf-Darwin Lineamentmay also be a lithospheric proposedthe profiles reflect growth of the volcanicthrough sill- fault (M. A. Feighnerand M. A. Richards,submitted manuscript, like intrusionsat 2-4 km depthbelow the summit. The differences 1993). Seismicityalso suggestsa sharpcontrast in lithospheric in size and morphologybetween these volcanos and thoseeast of strength as the western part of the archipelago, including the 90050 ' fracture zone probably reflects differences in Fernandinaand Isabela islands,is seismicallyactive but very few lithosphericthickness. Thicker and older lithospherewest of the earthquakeshave been recordedfrom the central archipelago(M. fracture zone can support larger volcanic structures (M. A. A. Feighner and M. A. Richards,submitted manuscript, 1993). Feighner and M. A. Richards,submitted manuscript, 1993) and Basedon gravity modeling,M. A. Feighnerand M. A. Richards providesgreater hydraulic head, allowing magmato ascendto (submittedmanuscript, 1993) find that crustalthickness increases greaterheight [Vogt, 1974]. from roughly 10 km on the marginsof the platformto as muchas Fernandina,at the extremewestem edge of the platform,is the 18 km beneath southeastern Isabela. largestand apparentlythe youngestvolcano in the archipelago.It Two prominentstructural lineaments control the alignmentand has the deepestcaldera (750 m) and the most frequenthistoric spacingof islandsrising from the GalfipagosPlatform [Darwin, volcanicand seismicactivity of any of the Galtipagosvolcanos. In 1860]. An east-westsystem parallels the GaltipagosSpreading 1968 the southeastportion of the calderadropped by as much as Centeraxis and the southernescarpment of the platform;it is most 350 m in trap-doorlike fashion[Sirekin and Howard, 1970]. In prominentlyexpressed in block faulting on the islandsof Baltra, 1988, a large block of the easterncaldera failed and slid into the SantaCruz, and SantaFe. A second,northwesterly system con- caldera. Oversteepeningas a resultof the 1968 calderacollapse trols alignmentof many of the Galfipagosvolcanos, particularly and injectionof a dike into the easterncaldera wall were probably the large shield volcanosof Isabela as well as the Wolf-Darwin responsiblefor this event[Chadwick et al., 1991]. Rowlandand Lineament, extending from the platform to the GSC. Although Munro [ 1992] arguethat this pattern,partial caldera floor collapse, both structural trends are seen on most central islands, east-west followed by failure of the walls due to oversteepening,is typical faults and fissuresare bestdeveloped in the southcentral islands, of the growth of Fernandinacaldera. They also concludedthat at whereasthe northwesterlytrend is most prominentwest of the times in the pastthe calderamay have been filled, or nearly so, large transform fault offseting the GSC at 90ø50'W. A with pondedlavas. The presentperiod appearsto be one of low symmetrical set of northeasttrending fractures dominatesthe magma supplyrate and calderaenlargement. regioneast of anothermajor transform fault at 86øW. The orienta- The island of Isabela is made up of five large shieldsand the tion of theseoblique fractures at oppositeends of the offsetridge truncated remains of a sixth, Volcan Ecuador. Most recent segmentof the GSC boundedby two transformfaults suggests eruptionson Isabelahave been centered on SierraNegra and Cerro that they resultfrom the sameforces responsible for the northward Azul at the southern end of the island. Volcan Wolf and Alcedo migrationof the adjacentridge segments(M. A. Feighnerand M. have eruptedhistorically and despitea lack of reliableaccounts of A. Richards,submitted manuscript, 1993). historic activity on Volcan Darwin, some of the lavas of that volcanoare so freshthat they couldwell have beenerupted within the last centuryor so. Volcan Alcedo,near the centerof the island Geologyof theIslands and directly downstreamfrom the core of the hotspot,is mantled The numberof simultaneouslyactive volcanos, their petrologic with rhyolitic pumice and the caldera containsa large flow of and morphologicaldiversity, and the apparentlack of any simple glassyrhyolite [McBirney et al., 1985; D. J. Geist, manuscriptin evolutionary pattern all distinguishthe Galfipagosfrom more preparation,1993]. Two smallerislands stand along the northern familiar examplesof hotspotvolcanism such as Hawaii. Charles and southernextension of the main axis of Isabela. The steep- Darwin madethe first geologicobservations of the islandsin 1835 sided,flat-topped island of Roca Redondais an erodedremnant of [Darwin, 1860]. The literature of the subsequent150 years a lava-filled crater with active fumaroles that is about the same containsscattered reports on variousaspects of the islands,but the distancenorth of Volcan Wolf as the spacingof the shieldson first thorough geological study of the archipelagowas that of Isabela. In the oppositedirection, and Wittmer McBirney and Williams [1969]. Subsequentstudies [Delaney et Seamountlie alongthe samealignment. al., 1973; Swanson et al., 1974; Baitis, 1976; Lindstrom, 1976; The volcanoseast of 90ø50'W havemore varied morphologies, Hall, 1983; McBirney et al., 1985; Geist et al., 1985a, b; Cullen eruptive histories, and compositions than the volcanos of and McBirney, 1987; Vicenziet al., 1990; Chadwicket al., 1991; Fernandinaand Isabela. Floreanais a roughlycircular, low shield Chadwick and Howard, 1991; Rowland and Munro, 1992; Bow whoseoutline is dominatedby numerousparasitic cinder cones. and Geist, 1992; McBirney, 1993] have addedconsiderable detail Bow [1979] divided Floreana lavas into an older Main Series, to our knowledgeof the individualvolcanos and islands. comprisingthe shield-buildinglavas, and a youngerFlank Series Volcanos west of the 90ø50'W fracture zone are distinct from relatedto the pyroclasticeruptions that producedthe cindercones the generally smaller and older eastern volcanos. The seven and the pyroclastic deposits that mantle much of the island. western volcanos of Isabela and FernandinaIslands are large, Paleomagneticpolarity measurementsindicate that a substantial 19,536 WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THEGAL•PAGOS

part of the present shield existed prior to 0.78 Ma [Cox and Baltraand Seymour, to thenorth, and the Las Plazasislets, just off Dalrymple, 1966]. Main Series lavas are primarily aphyric or the eastcoast, are simplefault blocks. sparselyolivine-phyric alkali basaltsand basanitoids,often con- Bow [1979] divided lavas of Santa Cruz into a Platform and tainingwehrlite, gabbro, and dunitexenoliths. Flank Serieslavas, Shield Series. The Platform Seriesis the older and occursmainly eruptedfrom the parasiticcones that now decoratethe island,are in the northeastof the island;it constitutesall of Baltra, Seymour, olivine- and plagioclase-phyricbasanitoids that are somewhat and Las Plazas. The lowermost units of this series have mor- poorer in alkalis and other incompatible elementsthan Main phologiesindicative of submarineeruption and are intercalcated Serieslavas [Bow, 1979; Bow and Geist, 1992]. CaptainPorter of with shallow (< 100 m) marine carbonates.Windows of reversely the U.S.S. Essexreported witnessing an eruptionof Floreanain polarizedflows exposedat 300 m elevationon the northernslope 1813, but it is possiblehe mistook Sierra Negra for Floreana of the shield are compositionallysimilar to flows exposedalong [Simkinet al., 1981]. As the Flank Seriesconformably overlies the coastand are includedin the Platform Series. Plagioclasewith the Main Series, there is no evidencefor a significanttime gap subordinate olivine dominates the phenocryst assemblagein betweenthese two series. Becauseerosion rates in the Galfipagos Platform Series lavas, which are predominantly, though not are generallylow, however,such a gap is not precluded. exclusively,tholeiitic. A few of the plagioclase-ultraphyriclavas Santiagois a WNW elongateshield reaching an elevationof resemblethe "abingtonites"of the northernislands [Cullen et al., just over 900 m. The oldestlavas were eruptedfrom summitvents 1989]. Clinopyroxeneis a phenocrystonly in the differentiated in the central northwesthighlands in an area now occupiedby a ferrobasaltsof Baltra and SeymourIslands. faintc•ldera-like depression [Swanson et al., 1974]. Theyrange SantaCruz ShieldSeries lavas range from basanitoidto olivine in compositionfrom picritic alkali basaltsthrough siliceous tra- tholeiite,but transitionalalkali basaltspredominate [Bow, 1979]. chyte[McBirney and Williams,1969; Baitis, 1976; Swansonet al., They are petrographicallydiverse and include olivine-only and 1974]. Somewhatyounger lavas, again alkali basaltsand their plagioclase-dominantassemblages as well as aphyriclavas. The differentiation products,were erupted from east-westtrending youngerlavas includedhawaiites. The youngestlavas, erupted fissuresin the central and easternhighlands. Olivine and plagio- from ventson the north slopeas well as the summitfissure, are claseare the dominantphenocrysts in the basaltsand arejoined by distinctlypoor in K20 andother incompatible elements. In places, augitein the more differentiatedrocks [Baitis, 1976]. K-Ar ages these lavas uncomformablyoverlie the older platform series. reportedby Swansonet al. [1974] rangefrom 0.77 Ma to 0.14 Ma, Accordingto Bow [1979], someflows appearto be no morethan a and all lavas are normallypolarized, implying agesless than 0.78 few thousandyears old judgingfrom theirmorphology and lack of Ma. Recent activity has been concentratedat the northwestern vegetation. and southeastern ends of the island and the southern flank and has As the islands drifted eastward from the hotspot, they con- produced a number of young tuff cones and islands, such as tinued to have intermittent eruptionsfor at least 2 m.y. before Bartholom6,just off the southeastend of the Santiago. These finally becomingextinct. The island of San Cristobalat the far young flows, which are barren of vegetationand have perfectly easternend of the archipelagohas the longesthistory of activity, preservedflow tops,cover about 25% of the island'ssurface. One havinghad at leastfive recognizableepisodes of eruptions[Geist flow in the James Bay area at the northwest end contains et al., 1985a]. The southwesternhalf of the island is dominated inclusionsof quincemarmalade jars stashedthere by marinersand by an older (2.35-0.6 Ma) .The northeasternhalf thereforemust have been eruptedwithin the 200 yearspreceding consists primarily of relatively young basalts erupted from Darwin's mentionof it duringhis 1835 visit. Severalsubsequent northeast-southwestfissures, though there are sparseoutcrops of historic eruptionsof olivine and plagioclasephyric basaltshave lavas erupted contemporaneouslywith the constructionof the occurred,the most recentin 1906. southwesternshield. Judgingfrom the well-preservedflow top Swansonet al. [1974] reportedages ranging from 0.92 to 1.09 morphologyand the sparsevegetation, the youngestof theselavas Ma for Rfibida and Pinz6n. Rfibida, near the south coast of are certainlyno more than a few thousandyears old [Geistet al., Santiago,measures only 2.5 by 3 km and consistsmainly of a 1985a]. Petrologically,they are dominatedby sparselyolivine- cluster of steep-sidedviscous domes, with the remains of two and plagioclase-phyric,relatively primitive (Mg# 60-70; Mg# is cinder cones at its northernbase. Lavas range in composition theatomic ratio of Mg toMg + Fe2+, expressed aspercent) alkali from olivine tholeiitethrough ferrobasalt to dacite[Swanson et al., basalts;tholeiites are much lesscommon. Clinopyroxenepheno- 1974]. Pinz6n is a small shield with two coalescingcalderas. crystsoccur only in the more evolvedlavas, many of whichcon- The high seacliffson the southwesternside of the islandexpose at tain gabbroic xenoliths. Though the lavas are compositionally leastseven eruptive cycles associated with the older,southern vol- heterogeneous,there is no systematicvariation with age or cano [Swanson et al., 1974]. Each begins with the explosive location. eruptionof daciteor ferrobasalt(icelandite) followed by lavasof Santa Fe and Espafiola are the only two major Galfipagos aphanitic ferrobasaltand finally plagioclase-richtholeiite. The islandsthat do not have volcanicmorphologies. On the basisof lavas in each sequencecan be related by shallow fractional brief visits, McBirney and Williams [ 1969] concludedboth these crystallizationof the observedphenocrysts and may result from islandswere uplifted submarineplatforms, but moredetailed sub- tapping successivelydeeper levels of a zoned magma chamber sequentwork showsthat both consistpredominantly of subareally [Baitisand Lindstrom,1980]. Activity subsequentlyshifted to the eruptedlavas [Hall, 1983; Geist et al., 1985b]. Both have been northerncaldera, which eruptedlavas and pyroclasticsof similar disruptedby extensiveeast-west faulting. Though a numberof composition. small cinder cones, some related to the east-westfaulting, have Santa Cruz is a gently sloping,elliptical shield rising 950 m been identified on each, variousindicators suggest that the main above sea level. A seriesof youthful pyroclasticcones and pit vent for SantaFe lies offshoreto the east [Geist et al., 1985b] and craters is aligned along a WNW trending axial fissure system that for Espafiolalies offshoreto the south[Hall, 1983]. Thus definingthe summitof the volcano. The islandis cut by a number both are probably remnantsof somewhatlarger shields. The of east-westtrending faults that are conspicuousalong the north- limited investigationsto date suggestthat Espafiola lavas are eastern, eastern, and southerncoasts. The adjacent islands of relatively uniform and typically are primitive, olivine- and WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THEGAL•PAGOS 19,537

plagioclase-phyricalkali basalts[McBirney and Williams, 1969; laboratories at the University of Oregon for major and trace Hall, 1983]. Lavas from Santa Fe are typically transitional elements. Ferrous-ferric ratios were measuredby titration and between alkali basalt and tholeiite and tend to be more evolved water contentsby a DuPont moistureanalyzer. Standardsof the than those of Espafiola [Geist et al., 1985b]. Phenocrystsof U.S. GeologicalSurvey and the GeologicalSurvey of Japanwere olivine and plagioclasein the more mafic lavas are joined by usedas controlsthroughout this period. augitein the more evolvedrocks. Becausethe quality of the early XRF analyseswas found to be The northern islands form a distinct group. The two north- inferiorto that of laterones performed by atomicabsorption, many ernmost islands, Wolf and Darwin, are remnants of volcanos that of the former were repeatedby the latter method, and selected rise 2000 m abovethe seafloor. Pinta, which last eruptedin 1928 sampleswere analyzedby XRF by J. M. Rhodesat the University [Simkin et al., 1981], is an elongatedshield, boundedby a steep of Massachusetts. The same is true of instrumental neutron fault on its westernside. Cullen and McBirney [ 1987] recognized activation analysis (INAA), which also becamemore preciseas an older shield, which emerged about 0.7 Ma, and a younger improvedmethods were introduced. Thus all samplescited in the fissurestage. The compositionaldistinctions between these stages accompanyingtables have been examined and revised to a are subtle, both consistingof light-rare-earth-enrichedolivine uniform level of quality. tholeiites. Marchena is a circular shield with a flooded caldera Isotopicdata were acquiredover a periodof 15 yearsin three that last eruptedin 1991. Vicenziet al. [1990] recognizedtwo laboratories:the Departmentof TerrestrialMagnetism of Carnegie lava series,separated by calderaformation and explosive activity. Institiution of Washington (DTM), the Max-Planck-Institut far Both seriesconsist predominantly of aphyric,moderately evolved Chemie (MPI), and the Departmentof Geological Sciencesat tholeiitesand alkali basalts,with little compositionaldistinction Cornell University. Consequently,analytical procedures and the between the precalderaand postcalderaseries. Genovesais a quality of the data vary somewhat. Chemicalseparation and mass smaller shield with a breachedcaldera, which forms Darwin Bay spectrometrytechniques used at DTM for Sr isotopeand Nd iso- on the southerncoast. The rockshave yet to be studiedin detail, toperatio measurementswere thosedescribed by Hart and Brooks but reconnaissancesampling suggestscomparatively uniform [1974] and Whitford et al. [1978]. These measurementswere tholeiitic compositions.On the basisof well-preservedflow-top done on single collector mass spectrometers.Many of the older morphologies,the most recent eruptionsprobably occurred no DTM data were reanalyzedat MP! and Cornell, and except as more than a few thousandyears ago. Mud in the bottomof saline noted below, the reanalysesagreed within the older analytical Arcturus Lake, which fills a crater in the center of Genovesa, is error, once interlaboratorybiases were taken into account. At less than 6000 years old [Colinvaux,1984]. This also suggests MPI, Sr and Nd isotope measurementswere carried out on Holocene volcanic activity on Genovesa. The most distinctive FinniganMAT261 massspectrometers. Details of the procedures featureof the northernislands is the abundanceof plagioclase-ult- givenby Whiteand Patchett[ 1984]. Pb isotoperatios were deter- raphyric basalts (abingtonites),which can contain up to 40% mined usingthe techniquedescribed by Whiteand Duprd [ 1986]. percentlarge (> 1 cm) plagioclasephenocrysts [Cullen et al., Cornell University data were acquiredusing a V.G. Sectormass 1989]. spectrometer.At both MPI and Cornell, all sampleswere leached With the notable exception of the northern islands, the dis- in hot 6 N HC1 for 30-60 min prior to dissolution. Chemical tributionof rock typesin the chainas a wholeparallels the nearly separationswere performedin the same manner as describedby east-westtrend of the CarnegieRidge. Petrographicvariations are White and Patchett [1984]. After separation,Sr was loadedon crudely symmetricalacross the long axis of the platform with singletungsten filaments with TaF5 activator.Data were acquired differentiatedrocks concentratedin a belt along part of the east- using a dynamic triple collector technique,with each analysis west axis. Plagioclaseis more commonas a phenocrystin lavas consistingof 100-120 mass scans. Nd was loaded on triple from the centerof the platform and lesscommon on the southern filaments (Ta sides and Re centers) with a small amount of islands and on to the north. On the whole, H2P20 4 or on Re single filaments with I-I2P204 and a small however, the prevalence of plagioclase and the paucity of quantityof ion exchangeresin [Walker et al., 1989], which serves clinopyroxene phenocrysts in the more primitive lavas as a reducingagent. Data were acquiredusing a dynamicquadru- distinguishesthe Galfipagosfrom mostother oceanic islands. For ple collector technique,each analysisconsisting of 120-150 mass example, in basalt from the Makaopuhi lava lake on Hawaii, scans. Pb was separatedusing the techniquedescribed by White olivine crystallizesalone down to 7% MgO whereit is joined by and Duprd [ 1986]. After separation,Pb was loadedon singleRe clinopyroxeneand then by plagioclaseat 1175øC [Wright and filaments with silica gel, and data were acquiredusing a static Fiske, 1971]. In contrast,crystallization of plagioclasealways quadruplecollector technique. It differedfrom that describedby precedesclinopyroxene in Galfipagoslavas, exceptin the most White and Duprd [ 1986] only in that 80-100 ratio measurements undersaturatedbasalts from Floreana,and it commonlyoccurs in were made rather than 60. A substantial fraction of the mea- lavaswith over 8% MgO. This inferredparagenetic sequence has surementswere replicated;values reported here are averagesof been confirmed by melting experimentsof Bow [1979], which thesereplicates. indicate that the first appearanceof plagioclaseoccurs around Rare earth elements, K, Rb, Cs, Sr, and Ba concentrationswere 1190øC. determinedby isotopedilution at DTM and at MPI. At DTM, light (Ce-Eu) and heavy (Gd-Yb) rare earth fractions were ANALYTICAL METHODS separatedon an AG 50W-X8 ion exchangecolumn and loadedon doubleRe filamentsfor isotopicanalysis following a technique Samplescollected during the initial reconnaissancestudy of the describedby Shimizu[1974]. At MPI, the techniquedescribed by 1960s[McBirney and Williams, 1969] were analyzedby classical Whiteand Patchett[ 1984] was employedfor rare earthanalysis. wet methodsby K. Aoki in his laboratoryat ScrippsInstitute of The two techniquesdiffer in severalrespects. For the rare , Oceanography.Samples collected during subsequent expeditions the MPI techniqueused a secondseparation to isolate the rare by McBirney and his studentswere analyzed by X ray fluo- earthsinto three fractionsand includedthe analysisof La and Lu, rescence(XRF), atomic absorption,and neutron activation in which were not determined at DTM. In both laboratories, a 19,538 WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THE GAL/I•PAGOS techniquederived from that of Hart and Brooks [1974] was used plotted againstMgO in Figure 2. Figure 2 illustratesthe wide for analysisof K, Rb, Cs, St, and Ba. The MPI techniquecor- rangeof compositions,much of which can be attributedto low- rected for mass fractionation for Sr, K, Cs, and Yb, whereas the pressurefractional crystallization. Given the proximity of the DTM techniqueemployed a massfractionation correction only for islandsto the GalfipagosSpreading Center and the similarityin Sr. This correctionresults in only a minor improvementin data trace element and isotope geochemistryof many of the lavas to quality. Estimatedanalytical uncertainties, based on replicationof MORB, it seems appropriate to compare the major element standards,range from 0.1% for Sm to 3.2% for Rb. chemistry to that of MORB, for which there is now a very Trace elements were analyzed at Cornell by inductively extensive data base. The Galfipagosdata overlap the MORB coupled-plasmamass spectrometry(ICP-MS) using a V.G. fields extensively,but there are a numberof subtlebut important Plasmaquad2+ instrument.One hundredmilligram samples were differences. First, thoughit is only partly apparentfrom Figure 2, digestedin HF and dried with HC104. They were thendissolved the Galfipagos data show somewhat more scatter about the in 100 mL of 1% HNO3. Ba and the rare earthswere analyzed fractionalcrystallization trends than MORB. Part of thisis due to using an external calibration technique with drift correction the Gal•ipagosanalyses being whole rock analyseswhile the [Cheathamet al., 1993]. Rb, Sr, and Y were measuredby MORB fields shownare for glassanalyses. Someof the scatter standardaddition. Estimated analytical uncertainties,based on thereforereflects accumulationof mineral phases(two obvious replicationof standards,range from 0.9% for Sr to 4.1% for Ba. examplesare the Al-rich lavas from ; their Al-rich Radiometricages were determinedon whole rock samplesby nature reflects plagioclaseaccumulation). Much of the scatter, the K-Ar method. Rocks assessed to be unaltered on the basis of however,seems to reflectgreater variation in parentalcomposition petrographic examination were crushed to 0.5-1.0 mm size andthe fractionalcrystallization process in the Galfipagos.A sec- fraction,ultrasonically washed in distilledwater, and dried. A split ond difference between the Gal,fpagos and MORB is that of this cleaned,crushed fraction was removed and powdered for K Gal,fpagosdata extendto moreMgO-rich compositions.A few of analysisby atomicabsorption spectrophotometry. Approximately the MgO-rich samplesare olivine cumulates, but others are 5 g of the crushedrock was loaded into a Mo-crucible for each aphyricor nearlyso, and the wholerock analysis approximates the argonanalysis and a 1 x 10-8 tort vacuumwas achieved by liquid composition.A third differenceis the systematicallylower overnightbaking (175øC) of the glassextraction line. Samples SiO2concentration of the Gal•ipagoslavas compared to MORB of were then fused by radio frequencyinduction heating of the similar MgO content. Figure 2 shows that at a given MgO crucible,and activegases were getteredover a hot Ti-TiO2 metal concentration,the most SiO2-richGalfipagos lavas have only the sponge.The isotopic compositionof argon was then measured averageSiO 2 concentrationof MORB, while mosthave 1-2% less massspectrometrically with an AEI MS-10S instrumentat Oregon SiO2than is typicalof MORB. Finally, the Gal•ipagoslavas seem StateUniversity, connected on-line with the extractionsystem. on averageslightly richer in FeO, CaO, Na20, and A1203 than The majority of sampleswere analyzed in the conventional MORB at comparableMgO concentrations. manet,introducing a known amount of 38Atspike during the In Figure 3, two minor oxides, K20 and TiO2, are plotted fusionto determinethe concentrationof radiogenic4øAt againstMgO; fields for MORB are shownfor comparison.TiO2 [Dalymple and Lanphere, 1969]. For very young and low-K varies by a factor of 2 for a given MgO concentrationin the samplesthe analyticalprocedures of Cassignoland Gillot [1982] Gal•ipagosdata; the variationin K20 is even greater. Lavasfrom were adopted. This procedure yields improved precision for Genovesa, the northeasternmostisland in the archipelago, samples with low concentrationsof radiogenic argon and is generallyhave the lowestK20 andTiO2 for a givenMgO concen- differentin severalways. First, 38At spike was not added because tration, while lavas from Floreana, the southeasternmostisland, of smallbut significant amounts of 4øArand 36At in thespike. have the highestK20. Genovesais actuallypoorer in K20 than Instead,sample 40At concentrationwas determinedfrom manyMORB of comparableMgO content.While thereis consid- instrumentsensitivity, which varied only 1-2% over long-term erableoverlap with the MORB field, the Gal•ipagoslavas are on measurementsof 3BAr peak heights in spikedanalyses and 4øAt averagericher in both K20 andTiO2 thanMORB. peak heightsin atmosphericargon calibrations. Second, aliquots Figure 4 is two diagramsthat have traditionallybeen used to of air argonwere routinelymeasured both before and after each discriminatealkali and tholeiitic basalts,and the Gal,fpagoslavas sample analysis in order to make the important correctionfor are againcompared to MORB. Figure4a showsthat thoughthere atmospheric40Ar contamination asprecisely as possible. This is considerablescatter, the GahSpagoslavas tend to clusteron the approachresults in lower errors in correctingfor atmospheric olivine-diopsidejoin (the alkali basalt--tholeiite divide), while argonthan spiking. Proportions of radiogenic 40At as small as a most MORB lie in the tholeiitic field in the center of the olivine- fewtenths of 1%of thetotal 4øAt signal are thus detectable and hypersthene-diopsidefield. Similarly, in the alkali-silica plot reproducible[Levi et al., 1990; Gillot and Nativel, 1989;Prestvik (Figure 4b) the Gal,fpagoslavas scatterto both sidesof the line and Duncan, 1991]. that divides alkali basalts from tholeiites in Hawaii [Macdonald and Katsura, 1964]. As may be seenin Figure4b, the Gal,fpagos RESULTS lavas are both poorerin SiO2 and richer in alkaliesthan most MORB. Major ElementChemistry It is also useful to compare the chemistry of Gal,fpagos Table 1 lists major elementanalyses and CIPW normsrep- magmaswith thoseof Kilauea [Wright and Fiske, 1971], a more resentativeof the moreprimitive and moreaphyric samples. (A familiar and better understood example of oceanic island completetable of 353 majorelement analyses used in thisstudy is volcanism. At 8% MgO, only the lowestNa20 corcentrationsin availableon microfiche. l) Five oxides and CaO/A1203 are shown GahSpagoslavas overlap those from Kilauea,and GahSpagos lavas are systematicallypoorer in SiO2 than ones. For the remaining •Tableof analysesis availablewith the entir• article on microfiche. oxides,the GahSpagosrange encompasses Kilauean values, but on Order from AmericanGeophysical Union, 2000 Florida Avenue,N.W., average,the GahSpagosare richerin Na20, CaO, and A1203and Washington,DC 20009. DocumentB93-005; $2.50. Paymentmust poorerin TiO2 and I;FeO than Kilauean lavas at the sameMgO accompanyorder. concentration (8%). The higher A1203 concentrationsin WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THEGAL/•PAGOS 19,539

TABLE 1. Representative Major Element Analyses and CIPW Norms of Galfipa[•os Lavas, G86-2 G86-3 E-41 FL-3 E-134 E-152 E-242 E-5• CA89-2 sN-1 Espanola Espanola Femandina Floreana Isabela Isabela Isabela Isabela Isabela Isabela Alcedo VolcanWolf VolcanDarwin VolcanEcuador CerroAzul SierraNegra SiO2 47.56 46.87 48.37 46.14 47.84 48.10 49.47 47.20 47.88 49.24 TiO2 1.32 1.21 2.80 1.19 3.19 2.67 3.02 3.35 2.31 2.85 A1203 15.98 14.46 15.49 14.23 13.42 16.01 12.42 15.24 16.23 13.77 Fe203 5.01 2.12 2.75 2.09 6.97 3.45 3.85 3.50 3.34 2.32 FeO 5.13 7.17 8.59 7.06 6.14 7.20 8.70 8.42 7.24 10.70 MnO 0.22 0.18 0.18 0.18 0.10 0.17 0.19 0.17 0.16 0.21 MgO 8.48 12.79 6.72 12.59 6.21 6.32 6.24 6.37 6.37 5.58 CaO 10.08 10.68 11.60 10.51 11.56 11.48 11.27 10.45 12.48 10.30 Na20 3.37 2.60 2.77 2.56 2.63 2.87 2.97 3.11 2.81 3.10 K20 0.79 1.11 0.42 1.08 0.29 0.42 0.53 0.78 0.51 0.57 P2o5 0.40 0.81 0.05 1.10 0.21 0.32 0.31 0.40 0.28 0.42 H2O+ 0.49 0.26 0.21 1.17 0.18 0.12 0.88 0.22 0.24 Total 98.83 100.00 100.00 98.94 99.73 99.19 99.09 99.87 99.83 99.30 CIPW Norms ne 2.62 1.77 5.00 0.32 0.86 or 4.70 2.71 2.48 6.52 1.76 2.52 3.18 4.70 3.01 3.41 ab 23.84 19.27 23.25 12.95 22.73 24.61 25.49 26.09 22.19 26.54 an 26.29 30.36 28.34 24.73 24.46 29.95 19.26 25.64 30.16 22.22 di 16.90 19.34 22.13 21.94 27.02 21.05 29.22 19.94 24.56 22.19 hy 6.19 10.81 5.20 10.42 12.57 ol 18.56 18.44 8.49 24.71 4.67 9.27 4.10 14.22 11.92 4.73 mt 3.02 3.05 3.19 1.33 1.84 1.51 1.79 1.70 1.65 1.88 il 2.52 3.65 5.28 2.32 6.20 5.15 5.83 6.46 4.39 5.49 ap 1.08 1.09 0.57 0.52 0.51 0.78 0.75 0.97 0.61 1.02 Si8.0 48.10 47.10 48.05 48.56 47.00 47.73 49.12 46.50 48.71 48.46 Fe8.0 9.12 9.63 9.53 9.82 10.25 8.42 10.22 9.62 10.93 10.00 Na8.0 3.53 2.92 2.44 3.07 2.43 2.71 2.80 2.93 3.07 2.85 K8.0 0.83 0.53 0.33 1.29 0.19 0.33 0.44 0.69 0.66 0.44 Mg # 61.1 64.7 54.6 71.5 47.1 52.2 47.8 49.5 52.6 43.8

E- 169 M-34 P-42 PS-16 E-165 SC-61 SC-75 E-1 G86-1 SC-196 SC-78 Genovesa Marchena Pinta Pinzon Rabida San Cristobal San Cristobal Santa Cruz Santa Cruz Santa Cruz Santa Cruz SiO2 48.65 49.00 49.71 47.85 48.87 47.72 47.42 46.14 47.14 48.56 46.44 TiO 2 1.36 2.67 2.41 1.78 0.69 1.41 1.08 2.01 1.87 2.61 1.03 A1203 16.25 14.57 15.30 15.79 17.72 16.74 16.86 16.10 15.67 14.76 15.92 Fe203 3.03 3.50 2.14 6.13 2.35 2.08 1.32 2.26 3.70 2.20 1.58 FeO 7.57 9.34 8.64 5.05 5.20 6.61 7.38 9.07 7.70 11.04 9.31 MnO 0.18 0.21 0.18 0.19 0.12 0.17 0.17 0.19 0.20 0.23 0.19 MgO 8.09 5.78 6.41 7.89 8.73 9.73 11.62 10.43 8.62 6.98 12.36 CaO 12.01 10.85 11.21 11.37 12.21 12.12 11.84 9.19 10.42 10.69 10.73 Na20 2.64 3.43 3.13 2.56 2.74 2.53 2.06 3.64 2.94 3.20 2.28 K20 0.11 0.35 0.74 0.27 0.17 0.49 0.18 0.26 0.35 0.35 0.08 P205 0.10 0.13 0.36 0.28 0.14 0.19 0.10 0.24 0.26 0.26 0.08 H2O+ 0.01 0.30 0.05 0.23 0.21 0.22 0.17 0.48 0.54 Total 100.01 100.13 100.3 99.39 99.15 100.00 100.20 100.01 99.41 100.88 100.00 CIP W No rms ne 1.54 1.58 5.05 0.39 0.57 or 0.30 2.07 4.37 1.62 1.02 2.88 1.06 1.54 2.10 2.04 0.50 ab 19.32 29.02 26.49 21.98 23.48 18.62 17.44 21.68 24.54 26.89 18.24 31.58 23.33 25.51 31.24 36.00 33.00 36.25 27.00 28.95 24.71 33.01 di 27.90 23.73 22.79 19.76 19.77 21.13 17.64 14.08 17.82 21.72 15.99 hy 2.04 3.59 6.99 0.76 2.28 1.24 ol 15.28 11.84 10.27 12.77 16.22 18.44 21.78 24.61 20.36 15.99 27.98 mt 1.44 2.01 1.70 1.55 1.07 1.24 1.24 1.62 1.62 1.88 1.56 il 2.42 5.07 4.58 3.44 1.33 2.69 2.06 3.85 3.61 4.93 1.96 ap 0.24 0.63 0.79 0.68 0.34 0.45 0.24 0.58 0.63 0.62 0.20 Si8.0 48.71 47.61 48.72 48.11 49.21 48.55 48.78 47.17 48.02 48.55 48.02 Fe8.0 10.35 10.98 9.49 10.56 8.20 9.17 9.29 11.80 11.55 12.33 11.46 Na8.0 2.66 2.96 2.79 2.57 2.82 2.80 2.45 3.95 3.10 2.98 2.72 K8.0 0.11 0.32 0.72 0.27 0.21 0.63 0.35 0.42 0.36 0.34 0.26 Mg # 59.9 45.2 52.0 57.1 68.0 67.2 70.7 62.6 58.2 48.9 67.2

G86-5 SF-13 BL-1 JH-86 SH-7 W90-4 DA-2 E87-3 E-156 SantaFe SantaFe SantiagoSantiago Santiago Wolf Island DarwinIsland CowleyIsland RocoRedonda SiO2 47.77 46.72 47.94 46.46 47.26 49.09 49.91 47.30 44.19 TiO2 2.99 1.90 1.27 1.26 2.42 1.13 1.52 2.61 2.21 Ai203 15.12 15.60 16.53 16.00 15.40 15.23 15.26 14.28 11.45 Fe203 6.77 3.41 4.17 2.95 3.17 6.09 7.45 3.24 3.21 FeO 5.03 7.69 6.15 7.37 9.38 5.25 2.74 9.19 9.69 MnO 0.19 0.18 0.17 0.18 0.21 0.18 0.19 0.17 0.19 MgO 5.59 9.69 9.34 10.79 7.58 6.50 7.30 7.83 16.98 CaO 9.12 10.23 11.76 11.97 11.26 12.19 12.40 10.18 7.47 19,540 WHITEET AL.: PETROLOGY ANDGEOCHEMISTRY OFTHE GAL,•PAGOS

TAB LE 1. (continued)

G86-5 SF-13 BL-1 JH-86 SH-7 W90-4 DA-2 E87-3 E-156 Santa Fe Santa Fe SantiagoSantiago Santiago Wolf Island Darwin Island Cowley Island Roco Redonda Na2¸ 3.51 3.03 2.64 2.28 3.24 2.95 2.90 3.07 2.32 K20 1.03 0.55 0.10 0.06 0.31 0.29 0.32 0.45 0.64 P205 0.77 0.35 0.18 0.10 0.27 0.48 0.20 0.26 0.32 H2O+ 1.40 0.10 0.09 0.27 0.23 0.22 0.04 0.67 Total 99.29 99.45 100.34 99.69 100.50 99.61 100.41 98.62 99.34 C IP W Norms ne 2.49 1.04 2.79 1.85 or 6.09 3.28 0.59 0.36 1.83 1.14 1.71 2.71 3.87 ab 29.70 21.29 22.37 17.54 22.21 21.11 24.96 26.43 16.55 an 22.46 27.60 33.00 33.52 26.50 51.50 27.46 24.27 19.24 di 14.65 17.36 19.74 20.85 22.61 14.84 24.67 20.84 13.13 hy 3.34 0.12 1.97 3.68 0.32 ol 11.85 21.92 19.90 22.56 17.07 5.84 11.42 17.98 38.50 mt 1.79 1.58 1.44 1.47 1.77 0.89 1.73 1.79 1.85 il 5.68 3.65 2.42 2.42 4.60 2.16 2.15 5.05 4.27 ap 1.68 0.85 0.43 0.24 0.65 0.55 1.05 0.63 0.78 Si8.0 48.13 47.81 48.66 47.59 47.47 48.15 49.47 47.59 47.07 FeB.0 9.75 11.51 10.59 10.73 11.95 9.71 8.97 12.77 13.38 NaB.0 3.08 3.31 2.87 2.61 3.15 2.63 2.75 3.20 3.08 KS.0 1.02 0.70 0.24 0.22 0.30 0.27 0.31 0.58 0.87 Mg# 47.3 61.6 62.7 65.7 52.5 51.9 58.0 53.6 70.6 Mg#= (Mg2+/(Mg2++Fe2+))atomicandnorms calculated assuming ferric is 10% of total iron. E-1 is fromMcBirney andWilliams [1969]; PS- 16,JH-86, and SH-7 are from Baitis [1976]; FL-3, SC-78, and SC-196 are from Bow [1979]; SC-61 and SC-75 are from Geist et al. [1985a], M-34 is fromVicenzi [1985], P-42 is from Cullen [1985]. See text for an explanation ofSi8. 0, etc. A completetable of 353 major element analyses used in thisstudy isavailable onmicrofiche. Thesedata, aswell as data inTables 3and 5are also available uponrequest asMicrosoft Excel© files on floppy disk from W. M. White.

Galfipagoslavas explain the early appearance and predominance equivalent to thoseof the differentiatedtholeiitic lavas. This of plagioclasediscussed earlier. restrictionof stronglydifferentiated magmas to thecenter of the Figures2-4 illustratethe contrasts in compositionalvariability archipelagois part of a moregeneral geographic pattern in which in Galfipagosvolcanos. Data from the large western volcanos on the minimum observedMg concentrationor Mg# decreases Isabela and Fernandina, as well as Pinta and Marchena, form towardthe centerof the archipelago. reasonablytight clusters on these diagrams. These volcanos have eruptedonly fractionated basalt of a limitedcompositional range, Geochronology withMgO generally in therange of 4-7%. Alcedois somewhatof anexception as it haserupted differentiation products which range Table 2 listsnew K-Ar agedeterminations for rocksanalyzed through . Basalts from Genovesa,though less in this study. In Figure5 mostof theseages are plotted against differentiated,also have a fairly restrictedrange (8.5-6% MgO) distancefrom Fernandina, presumed to be the leadingedge of the andform reasonably coherent trends on MgO variationdiagrams. hotspot,in the directionof motionof the Nazcaplate (100.5 ø The two northernmost islands, Darwin and Wolf, also seem [Grippand Gordon, 1990]). Ages from Wolf andDarwin Islands compositionallyuniform, once the effectsof plagioclaseac- arenot plotted, as these volcanos lie "upstream"from the hotspot cumulationare discounted,but samplingof thesevolcanos is less and cannot be related to the remainder of the islands by plate complete.In contrast,the productsof theremaining volcanos, motionalone. Also plotted are the previously measured K-Ar ages mostnotably Santa Cruz, San Cristobal, Santiago, Santa Fe, and from the islands[Cox and Dalrymple,1966; Swanson et al., 1974; Floreana,are distinctlymore heterogeneous both in MgO andin Bailey, 1976], recalculatedto currentlyaccepted decay and the valueof otheroxides at a givenMgO content.Espafiola seems abundanceconstants (see Table 2). The two setsof age de- to fall into the lattergroup, in contrastto theconclusion of Hall terminationsoverlap completely, although analytical uncertainties [1983]that Espafiola magmas are relatively homogeneous, but our for the earlier measurementsare often much larger than for the data for that island are limited. A range of parental magma newage determinations; thetwo oldest ages reported (Santa Cruz: compositionsis required on each of theseislands. 4.2 +1.8 and 3.8 + 1.8 Ma [Bailey, 1976]) are not plottedbecause Lavas more differentiated than basalt are found only in the of their imprecision.Likewise, a previouslyreported age for centralbelt extendingeast from on Isabela.The SantaFe (3.9 + 0.6 Ma, Geistet al. [1985b])had a largeanalytical rhyolitesof Alcedo,near the centerof Isabela,are the mostex- uncertaintyand was remeasured at 2.50 + 0.08Ma (sampleSF-13, treme volcanic differentiates, but Pinztn and Rfibida, a short Table2). Thetwo analysesdo not agree within experimental error; distancedownstream (east), consist predominantly of ferrobasalts, thesecond analysis ispreferred because amuch larger proportion icelandites,and daciteswith lesseramounts of siliceoustrachyte. ofradiogenic 4UAr was measured. The adjacentisland of Santiagohas a mildlyalkaline series of While,in general,the age of Galfipagosvolcanism shows little hawaiites,mugearities, and trachytes. Ferrobasalts and hawaiites relationshipto distance,the ages of theoldest flows at each of the alsooccur on SantaCruz, just to theeast. Plutonicxenoliths have islandsform a reasonableprogression from youngest in thewest to beenejected from explosive vents on all islandson thiscentral oldest in the east. The oldest flows on a volcano are likely to be axis, but the most differentiatedrocks, includingferrogabbro, buriedby subsequentactivity, so our dates certainly overrepresent ferrodiorite,and quartz syenite, are foundin a smallcinder cone theyounger flows. Given this sampling bias, the geochronological on the northern coast of Rfibida island. Their compositionsare dataare entirely consistent with the simple model of inceptionof WH•E ET AL.: PETROLOGYAND GEOCHEMISTRY OF THE GAL•PAGOS 19,541

1 9

o• ...-'.:-5:"'...... :.::...... ::::;'•. .-'....:.....'?:i•. 'ø ,,',.:,; ..-•:--•:,. ,,•.r .m • ©0

.•:i•::•; ...... -•:•:•. • •,-• I I .... • ---:.---'t', • •'.* • '-,' ,

0 5 10 15 0 5 10 15

22 4

":-::•:•:••..::.:•'•::•:•:•:•:.• :•:..• ...... , • •l 8 I0 0•::•?".?..:•:.•.o., .:.-.:::•:..-:...... •%•.:..';•i'•::'.....? •::-:.::?: • :,'...... ,•; c% •:,'• ' "ß , • • • 0 • .• • I'"'"'•"'"'""•"•':':••••"••'•q•;•:::'•::"'•::;•:-•;; •,i• :" • ;•:"*'•• :::•:- :;" • I

C 10 r .... I .... I .... I 2 :• .... I .... I .... / 0 5 10 15 0 5 10 15 15: • P o C , 75 • I, ,• ; ...... • ...... •* ß L ,,,.• ,.,•....:..:... I ... -: I 0 ....."?:' "'

0 , :::• 45 ••- x,.ß ' '-'" : '"'•,., •i,.• ;" , 0 0 5 10 15 0 5 10 15 MgO MgO ,.I., Espa•ola ID Genovesa • S.Cristobal ß Santiago ¾rabidaPinzon & A Fernandina ,a Marchena [] S.Cruz Plat.1 S. Cruz O Isabela I• Floreana ß Pinta • S. Fe • Wolf-Darwin I

Fig. 2. Variationof majoroxides as a functionof MgO in Galfipagoslavas. Light and dark shaded areas enclose 99% and 95%, respectively,of 1000analyses of MORB fromthe East Pacific Rise. Lavasfrom Baltra, Seymour, and Las Plazas Islandsare groupedin this and subsequentfigures with the SantaCruz PlatformSeries. Similarly,Roca Redonda is groupedwith lsabela,and Cowley Island and Nameless Rock are grouped with Pinz6nand Rfibida. volcanismon a plate moving over a stationarymantle plume. A Contaminationof IsotopeRatios in SubaerealOceanic Basalts secondfeature of Galfipagosvolcanism apparent from Figure5 is the extensive period over which volcanism has occurred on Table 3 reportsisotope ratios. Duringthe courseof this study, several islands. Volcanism has continued on San Cristobal for a we found in a few casesthat leachingsamples in hot 6 N HC1 periodof over 2.4 m.y., and extendedover a periodof at least2 beforedissolution lowered Sr isotoperatios significantly. Table 4 m.y. on Santa Fe. Santa Cruz, Pinta, and Floreana all have been compares87Sr/86Sr ratios of leachedand unleached analyses for active for periodsof nearly a million yearsand more. Clearly, thosecases where significantdifferences were found. In a few mostof the datedflows erupted at locationswell to the eastof the cases, such as SC-196 and E-111, the samplesshowed some leadingedge of the hotspot,indicating either an extendedzone of evidenceof weatheringand alteration. Thesesamples may have melting "downstream"from the plume or postshieldextensional been submergedduring high standsof sealevelin the past,where volcanismacross the platformregion that allowedasthenospheric theywould have been exposed to seawater.Leaching probably re- meltsto reachthe surfacethrough old shieldvolcanos. moved Sr of seawaterderivation in secondaryphases in these 19,542 WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THEGAL,•PAGOS

rainwater than halite, the principal phasein sea salt. Repeated cycles of deposition of sea salt and washing by rain over thousandsof years could leave a residue consistingof calcite, gypsum, and other relatively insoluble minerals in which the concentrationof Sr might exceed 1000 ppm. The presenceof amountsof this salt too small to noticecould still shift Sr isotope ratiossignificantly. For example,if the salt contained1000 ppm Sr with87Sr/86Sr = 0.7092, about 2% of it wouldbe required to producedthe isotopicshift observedin sampleE-4. The substan- tial chloride contentof rainwater [e.g., Brown et al., 1989] sug- gestscontamination of Sr isotoperatios by this mechanismmay be a problemeven 100 to 150 km from coasts. Another potential source of seawater Sr contamination is guano. This appears to be a problem for sample WO-1, a vesicularlava from Wolf Island with guanoin the vesicles. Its 87Sr/S6Sris significantly higher than ratios from other samples from Wolf and Darwin Islandsas well as dredgedsamples from + Espafiota [] GEnOYESa the Wolf-Darwin Lineament[Harpp et al., 1990], thoughits œNd a FErnandina ß MarS-lEna and Pb isotoperatios are similar. This suggeststhat our leaching ß SanTiago ß PinTa procedurefailed to removeall the guano. • o S. CrisTobal [] Floreana SampleE87-1 is a lava bombin palagonitetuff from Nameless 6 Wolf-Darwin ß S. Cruz Rock, a smalltuff ring remnantnow only severaltens of metersin :iil}:ß * SantaFe [] S.Cruz diameter. The samplewas collectedfrom within a few metersof •2 ß Pinzon& sealevel andshows obvious signs of alteration.Its extremelyhigh !.••;: oIsabela Plat. 87Sr/S6Sr(0.70714) suggests that significant amounts of seawater- derived Sr remained after leaching. Because we suspect -.======.:.•.======s!::::..":•:::.. contamination for both WO-1 and E87-1, the Sr isotoperatios of :::::::::::::::::::::::::::::::::::...... _.---:: ...... • • • these samples are not plottedin any of the figuresor consideredin .:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: ...... •.•,•"•' :--.•.'-:•,,** . our subsequent discussion...... ":?:':"': :';:-.-.::--.:::-:. :L•-'-":•,•.,".:.,•.: -' • o ' ...... •-:•:::•::•;•:::• .:.:ss::g.ss: O:'...:s:::.•'.. •,,. -' .,• •..• ....'•:•'•' -½:'•. ' x , j • Whatever the source of the contamination,it appearsthat leachingin hot HC1 before Sr isotoperatio analysisis advisable 0 5 10 15 even for apparentlyfresh basalts. If the contaminantSr is either MgO residualsea salt or guano,washing in distilledwater might not be sufficientto remove Sr-bearingcompounds having only limited Fig. 3. Variationof minoroxides K20 andTiO2 with MgO in Galfipagos solubility. lavas. Shadedregions are the fields for MORB as in Figure2.

IsotopeRatios

Sr and Nd isotoperatios plotted in Figure 6 reveal a range of isotopiccompositions between those typical of MORB and values cases.However, in mostof theother samples whose 87Sr/86Sr near the median of the oceanic island basalts. Most of this decreasedafter leaching, there was no visible evidenceof alter- variation occurs between volcanos. Individual volcanos or islands ation or weathering. Indeed, most appearquite youngand were are more homogeneousthan the dataset as a whole,but significant undoubtedly never directly exposed to seawater. A similar variations nevertheless occur between lavas of individual volcanos phenomenonwas observedduring reanalysisof samplesfrom on someislands, e.g., Floreana,San Cristobal,and Pinta. In other Easter Island (original data reported by White and Hofmann cases,such as Espafiolaand Femandina,the eruptiveproducts of [1982]):878r/86Sr of leached samples were systematically 0.0002 the volcanos appear virtually homogeneouswithin analytical to 0.0003 lower than from unleachedsamples., and $. R. Hart error. Thoughthis may refl•t the limited samplingin somecases, [personal communication, 1992] has observed the same in others, e.g., Marchena, an island for which we have an phenomenonwith youngbasalts from the ReykjanesPeninsula of extensivedata set, variability is still quite limited, indicatingthat Iceland. Galfipagosvolcanos can be nearly isotopicallyhomogeneous. In For the young, apparentlyunaltered basalts, the origin of the general, those volcanos whose eruptive productshave limited relativelyradiogenic Sr removedby leachingis not obvious. On rangesof major elementcomposition are alsoisotopically homo- oceanicislands composed entirely of youngbasalt, seawater is the geneous, while those with a greater range in major element only source of radiogenic Sr. We suggestthat the seawater compositionare isotopically heterogeneous. Thus most of the contaminantis delivered as an aerosolderived from breaking large western volcanos are isotopically homogeneous. Volcan waves. This seawatermist is depositedon the rocks, where the Ecuadorand Volcan Alcedo are exceptions,as their lavas exhibit water would evaporate,leaving a residue of sea salt. Seawater significantisotopic heterogeneity. The Alcedo rhyolite, E-135, containsonly 8 ppm Sr, much less than the basalts,but sea salt has isotope ratios in the middle of the range for the Alcedo contains230 ppm Sr, a concentrationcomparable to that in the basalts,consistent with its evolutionby fractional crystallization rocks. The concentrationof Sr in the residualsalt may well be en- from a basalticparent. hancedif the rocksare washedby rainwater.Sr in seasalt may be Floreana,Santa Cruz, SanCristobal, Santa Fe, andSantiago all largelyheld in calcite,which would dissolve much more slowly in exhibit a significant range in isotope ratios. The variation in WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THEGALAPAGOS 19,543

Di Nœ O

.. ß ß ß Ne • ß:• .... ßß0 ß

O[ Hy

1 44

ß Espa•ola [] Genovesa ß S. Cruz o S. Cristobal ß Pinzon& Rabida a Fernandina ß Marchena © Santiago 6 Wolf-Darwin a Floreana ß Pinta o Isabela• Santa Fe [] PlatformS.Cruz

Fig. 4. (a) Explodedview of the normativebasaltic tetrahedron projected from plagioclase[Yoder and Tilley, 1962], and (b) alkali-silicadiagram. Norms were computedassuming ferric iron is 10% of totaliron. Shadedregion are the fieldsfor MORB as in Figure 2. Line dividingHawaiian alkali basaltsfrom tholeiites[Macdonald and Katura, 1964] is shownfor comparison.

Santiagois largely geographic:basalts from the southeasthave above,but it remainsimperfect. For example,Floreana samples distinctlymore depletedisotopic signatures than thosefrom the havesystematically higher 87Sr/•6Sr than Pinta, even though the northwest. Within each subset,there is no systematicvariation 143Nd/144Ndratios of thetwo are comparable. Samples from the with age:the older lavashave similar isotope ratios to the historic two southern volcanos on Isabela Island, Cerro Azul and Sierra flows in that part of the island. This suggeststhat Santiagomay Negra, plot betweenPinta and Floreana. consist of two coalesced volcanos, a southeastem one and a north- Figure 7 showsPb isotoperatios. The data covera substantial western one, which have been fed by distinct mantle sources partof theoceanic basalt range, with 2ø6pb/2ø4pb varying from throughouttheir histories. This situationis reminiscentof the 18.4 to 20.0. As with Sr and Nd isotope ratios, most of this Hawaiian volcanos. The variation in Santa Cruz (including the variation is between, rather than within, volcanos. Though a adjacentsmall islandsof Baltra, Seymour,and Las Plazas)is also strongoverall correlationexists among the threePb isotoperatios, geographic,but this variationis probablyfundamentally temporal: individual islands or volcanos can deviate from this correlation. the Platform Series lavas of the northeastare about 1 m.y. old, For example,samples from Darwin, Wolf, andPinta islands have while the remainderof the island,composed of the Shield Series, somewhathigh 2ø7pb/2ø4pband 2ø8pb/2ø4pbfor a given is younger. In SanCristobal, there is no clearrelationship of iso- 2ø6pb/2ø4pbratio when compared tothe other islands. tope ratios to age or location. On Floreana,the variationalso The relationshipsbetween Pb and Sr andNd isotoperatios are appearsunrelated to age or location:ranges for the Main and shown in Figure 8. Pb isotope ratios show a strong positive Flank Seriesoverlap almost completely. correlationwith 87Sr/•6Sr and a stronginverse one with e•d- Asin The correlationbetween Sr and Nd isotoperatios is somewhat the previous figures, individual islands and volcanos tend to betterthan reported by Whiteand Hofrnann[ 1978], a resultof the exhibit better correlations than the entire data set. removal of contaminant seawater Sr by leaching as described The overlap of the Galfipagosdata with the MORB field is 19,544 WHITE ET AL.' PETROLOGYAND GEOCHEMISTRY OF THE GAL,4,PAGOS

TABLE 2. K-At Age Determinationson GalfipagosBasalts earthpatterns is apparent,ranging from fairly stronglylight rare Percent Percent Age earth element (LREE) enriched basalts from Floreana to the Sample Location K Radiogenic4øAr +la, Ma LREE-depletedbasalts of Genovesa.Only moderatelyto strongly Espanola LREE-enriched lavas have been observed on Floreana, Pinta, G86-2 GardnerBay 0.645 3.08 2.61_+0.15 Espafiola, Isabela, and Fernandina. The Marchena basalts are G86-3 Punta Suarez 0.391 24.18 2.77_+0.04 slightlyconcave downward, with a maximumat Nd or Sm; that is, Floreana FL-3 Main Series 0.894 3.66 1.52_-+0.08 the lightestrare earthsare depletedrelative to the intermediaterare E-110 Flank Series 0.365 0.12 0.08_+0.13 earths but enriched relative to the heavy REE. Only LREE- E-108 Flank Series 0.365 0.72 0.35_-+0.11 depletedlavas occur on Genovesa,while bothLREE-enriched and Genovesa LREE-depleted lavas occur on Santiago,Santa Cruz, SantaFe, G86-8 0.002 0.00 0.00 Isabela and San Cristobal. Theselatter four islandsalso show the greatest E-237 Volcan Darwin 0.582 0.00 0.00 diversityin major elementchemistry and isotoperatios. The rare E-142 Volcan Alcedo 2.440 7.88 0.15_-+0.02 earth patterns of the large western volcanos of Fernandina and 6.97 0.12_-+0.01 Isabela are particularly uniform, as are their major element DG-48 Volcan Alcedo 2.930 3.90 0.074_-+0.024 chemistryand isotoperatios. The rare earth patternsof the seven DG43 Volcan Alcedo 0.430 0.00 0.00 DG-27 Volcan Alcedo 0.450 0.52 0.15_-ff).05 basaltsanalyzed from Volcan Darwin are all essentiallyparallel, Marchena though the absoluteabundances vary by a factor of nearly three E-10 0.172 0.88 0.1•.04 between E-217 and E-224. The patternsof the two basaltsana- M-34 0.299 4.69 0.56_-+0.04 lyzed from Fernandinaare virtually indistinguishable.Thus the M-51 0.241 1.42 0.11_-+0.02 same contrastsin compositionaldiversity observedfor major M-67 0.176 0.73 0.12_-+0.04 M10 0.172 0.88 0.1•.04 elementsand isotopesalso extend to the rare earths. Pinta The REE patternsin Genovesabasalts, and somebasalts from E-8 0.627 0.28 0.036_+0.025 Santiago and Santa Cruz, are noteworthy in that they are P-21 0.357 2.81 0.89_-4-0.24 indistinguishablefrom those of typical MORB. The concentra- P-50 0.400 0.00 0.00 tions of K, Rb, Cs, and Ba in the Genovesa basalts, as well as Pinz6n E-68b 0.504 3.24 1.40-&-0.08 someof the SantaCruz shield(e.g., SC-78, SC-100) and Santiago E-70 1.20 3.04 1.04_+0.00 (e.g., SH-75a) basalts,are very low, and are also well within the San Cristobal rangeof typicalMORB. Sr concentrationsare slightlyhigher than SC-5 Southwest 0.838 1.43 0.77_+0.18 that of averagenormal MORB (N-MORB), which in somecases SC-20 Southwest 0.888 6.82 1.33_+0.04 SC-23 Southwest 0.473 33.27 2.35_+0.03 could be a result of plagioclaseaccumulation as some of these SC-28 Southwest 0.639 15.36 0.89_+0.03 lavas are plagioclase-rich. Thus in all their trace element and SC-34 Southwest 0.539 21.04 2.33_+0.039 isotopiccharacteristics, a numberof basaltsfrom the centerof the SC-61 Northeast 0.540 0.27 0.050-•.039 GalfipagosArchipelago are virtually indistinguishablefrom N- Santa Cruz MORB, a featurethat appearsto be uniqueto the Galfipagos. G86-1 Las Plazas 0.271 3.67 1.31_+0.10 SC-108 NE Platform 0.573 6.26 1.12_+0.07 As expected, there is a general correlation between trace SC-151 South Shield 0.662 0.31 0.03_+0.025 elementabundances and isotoperatios. Thus Floreana,Pinta, and 1.19 0.024_+0.011 the southern Isabela volcanos, which have the most enriched Santa Fe isotopic signatures, are all LREE-enriched, while Genovesa G86-4 0.516 16.57 2.76_-/-0.04 basalts,with a depletedisotopic signature, are LREE-depleted. In

G86-5 0.800 1.35 0.72_+0.09 detail, however, this correlationis very imperfect. For example, SF- 13 0.515 7.27 2.50-•-0.08 although basaltsof Volcan Darwin and Volcan Wolf both have Darwin Island fairly depletedisotopic signatures, they are LREE-enriched. More DA-1 0.303 0.45 0.39_+0.15 surprisingly,some of the SantaCruz basaltswith MORB-like iso- DA-2 0.266 0.48 0.41_+0.16 Wolf Island tope ratiosare LREE-enriched. W90-2 0.249 4.92 1.60-ñ-0.07 Among the LREE-enrichedbasalts there is a subtledifference W90-4 0.241 3.97 0.88_+0.13 in patterns. All LREE-enriched lavas with depleted isotopic Roco Redonda signatureshave concave-downwardLREE patterns,that is, they E-156 0.534 0.19 0.053_+0.054 flatten toward La. Lavas with less depletedisotopic signatures Ages calculated using the following decay and abundance have straightor concave-upwardlight rare earthpatterns. A good constants:)•e= 0.581x10-•ø yr-1; )• = 4.962x10-1ø yr-• 4øK/K-- 1.167 x 10 -4 mol/mol. example of this difference can be seen in Figure 9, where the concave-downwardpatterns of Volcan Wolf lavas, which have ENd~ +8, contrastwith the straightpatterns of lavasfrom Volcan particularly noteworthy. Although Zindler et al. [1979] also re- portedrelatively low 87Sr/86Srand high 143Nd/144Nd forIceland, Ecuador,which lies only 10 km to the east,and Roca Redondato the north,both of which haveENd ~ 6.5. the data reported here, particularly those for the Genovesaand someof the SantaCruz basalts,are the lowest87Sr/86Sr and There is also a systematicvariation in heavy rare earth pat- highest143Nd/144Nd yetfound on oceanic islands. Pb isotope terns. Among the LREE-enrichedbasalts, those from the southern part of the archipelago, most notably basalts of Floreana, ratiosfrom theseislands also overlap the MORB fields. Espafiola,and San Cristobal,have comparativelyflat heavy rare

Trace Elements earth (Gd-Lu) patterns, while LREE-enriched basaltsfrom the remainderof the archipelagogenerally have inclinedheavy rare Table 5 reports concentrationsof rare earthsand other trace earth patterns. This differenceis most easily seenwhen the rare elements. Chondrite-normalizedrare earth element (REE) pat- earthpatterns of Floreanaor Espafiolaare comparedwith thoseof terns are shown in Figures 9-12. A remarkablevariation in rare Rfibidaor Pinz6n in Figure 10. WHrI'E ET AL.: PETROLOGYAND GEOCHEMISTRYOF THE GAL•PAGOS 19,545

300

San CrisTobal 200 - . ++.ß ii Espanola

CoEnovEsA I I Santa Fe

SanTa Cruz Floreana Marchena Pinzon rabida Santiago Pinta

Isabela

1 2 3 4 5 Age (millionyears)

Fig. 5. Presentdistance of Galfipagoslavas from the westernmosteruptive center (Fernandina) plotted against their K-Ar ages. Shadedregion shows the positionpredicted from 37 mrn/yrmotion of the Nazca plate [Gripp and Gordon,1990] over a fixed hotspotcentered beneath Fernandina. Small dotsare datafrom the literature;larger symbols are from Table 2. Dashedline illustratesthe pre-3 Ma motionof 65 mm/yr [Minsterand Jordan, 1978].

RelationshipsAmong Major Elements,Trace Rlements, and mixed results. For SiO2, Na20, K20, La, Sm, and Yb, applying IsotopeRatios the corrections reduced the scatter of the data for individual vol- canosas well as reducingcorrelations with MgO and Mg#, sug- Becausefractional crystallizationcan obscurecompositional gesting the correction procedureat least reduced the effects of relationshipsthat might, in primitive rocks,elucidate variations in fractional crystallization on the abundancesof these elements. sourcecomposition, depth of melting, and degreeof melting, The abundancesof theseelements bear comparativelysimple rela- Kleinand Langmuir[ 1987]corrected MORB compositionsto 8% tionships to MgO concentration(Figures 2, 3, and 13). Ba, MgO usinga linearapproximation to the slopeof the liquidline of however, was not significantlycorrelated with MgO, so we did descent. Geist [1992] used a similar procedure,basing his not apply a correction; apparently, the effects of fractional correction on a theoretical evolution model for Galfipagos crystallizationon Ba are smallcompared to degreeof meltingand magmas. We have adoptedthis approachand correctedSiO2, sourceeffects. In the caseof EFeO, a significant,though reduced, FeO, Na20, and K20 abundancesusing Geist's equations for the correlationremained even after correction. This may reflect the Galfipagos(after first recalculating analyses to 100%). Sis.0, FeB.0, correlationbetween MgO andFeO in primarymelts [e.g., Hanson NAB.0,and K8.0 for representativesamples are reportedin Table 1. and Langmuir, 1978], or it could reflect the complexityof the For incompatible elements, however, the linear correction relationshipbetween FeO andMgO in the raw data(Figure 2) and functionsof Klein and Langmuir[1987] and Geist [1992] are not the inadequacyof the correctionwe applied. Geist [1992] also satisfactoryfrom boththeoretical and observationalperspectives. reportedcorrection terms for the CaO/A1203 ratio, but we found Theoretically,fractional crystallization produces an exponential that applyingthis correctionresulted in a strongercorrelation with increasein incompatibleelement abundances as a functionof the MgO than existedin the raw data. The complexrelationship be- fractionof liquid remaining.An exponentialincrease is precisely tween CaO/A1203 and MgO (Figure 2) suggestsno simple what we observewhen incompatibleelements are plottedagainst correctioncan be appliedto this parameter,so we did not consider MgO (Figure 13). Thus to facilitatecomparison of incompatible it further. By computingcorrected La, Sm, and Yb abundances, elementabundances and minimize the effectsof fractionalcrys- we were also able to calculate corrected La/Sm and Sm/Yb ratios. tallization, we also applied corrections to La, Sm, and Yb The correctedLa/Sm ratioscorrelated slightly better with isotope abundancesusing equations of theform Ms.0 = M x ea(Mgo-8.0), ratiosthan did the uncorrectedratios. This suggestsour correction where M is the concentration of the element of interest and the was essentiallysound, since La/Sm dependsstrongly on source coefficient a was derived empirically from the relationship composition.In the discussionthat follows we usethe notationof betweenthese elements and MgO for the entire Galfipagosdata Klein and Langmuir[ 1987] to denotecorrected variables. set. To examinethe relationshipbetween the variouschemical and Obviously,caution must be usedin applyingsuch corrections isotopic parameters, we calculated average values for each so as not to imposeartificial relationshipsbetween variables, and volcano (Table 6), and then computed the correlation matrix not to obscurereal ones. Our attemptat correctingthe data had (Table 7) for these averages. La/Sm ratios and Ba and La 19,546 WH1TE ET AL.: PETROLOGY AND GEOCHEMISTRY OFTHE GAL•PAGOS

TABLE 3. IsotopeRatios in GalapagosLavas Sample Location 87Sr/86Sr 143Nd/144Nd 206pb/204Pb207pb/204Pb 208pb/204Pb Espafiola E-102 Summit 0.70294 0.513024 G86-2' GardnerBay 0.70298 0.513027 18.725 15.520 38.289 G86-3' Punta Suarez 0.70292 0.513044 18.762 15.535 38.380 H334 Punta Suarez 0.70293 0.513021 18.742 15.522 38.309 Fernandina E-42 0.70312 0.51294 19.064 15.553 38.696 F436t 1961 Flow 0.70319 F3N3t 0.70323 E-87 P. Espinoza 0.70320 0.512934 E-41 So. slope 0.70319 0.512951 19.114 15.567 38.778 Floreana E-98 wehrilite xenolith 0.70352 0.512937 19.970 15.664 39.721 FL-3 Main Series 0.70366 0.512905 20.002 15.657 39.739 FL-19* Main Series 0.70340 0.512928 19.812 15.642 39.469 FL-25* Flank Series 0.70339 0.512968 19.606 15.622 39.254 FL-26 Flank Series 0.70343 0.512929 FL-27 Flank Series 0.70338 0.512971 FL-29* Main Series 0.70353 0.512936 20.004 15.666 39.746 E-110 Flank Series 0.70360 0.512975 19.785 15.636 39.516 E-108 Flank Series 0.70341 0.512954 19.605 15.605 39.283 Genovesa E-169 0.70272 0.513127 18.387 15.511 37.941 E-171 0.70266 0.513119 18.433 15.517 37.976 E-172 0.70259 0.513119 18.443 15.524 38.005 G303 0.70274 0.513142 G86-8' 0.70275 0.513116 18.439 15.511 37.967 G304 P.rincePhilip's Steps 0.70270 0.513107 18.427 15.509 37.952 Isabela E-132' Volcan Alcedo 0.70331 0.512939 19.337 15.576 38.955 E-134 Volcan Alcedo 0.70327 0.512996 E-135' Volcan Alcedo 0.70316 0.512988 19.119 15.559 38.673 E-137 Volcan Alcedo 0.70309 0.512978 19.191 15.58 38.806 B056t Volcan Ecuador 0.70297 B155t Volcan Ecuador 0.70297 0.513006 E-54 Volcan Ecuador 0.70328 0.512940 19.232 15.597 38.951 E-56 Volcan Ecuador 0.70322 0.512932 19.274 15.593 38.990 ZOB6• Cerro Azul 0.70332 0.512931 CA90-1 * Cerro Azul 0.70336 0.512933 19.317 15.568 38.955 CA90-6 * Cerro Azul 0.70334 0.512954 19.322 15.573 38.947 NOC5? SierraNegra 0.70338 0.512922 SN-1 SierraNegra 0.70330 0.512915 19.372 15.602 39.058 SN-3 SierraNegra 0.70332 0.512931 19.375 15.604 39.073 E-152 Volcan Wolf 0.70275 0.513047 18.971 15.588 38.474 E-153 Volcan Wolf 0.70269 0.513077 18.822 15.554 38.426 E-155 Volcan Wolf 0.70273 0.513038 18.908 15.542 38.38

E-151* Volcan Darwin 0.70285 0.513005 18.788 15.525 38.277 E-216' Volcan Darwin 0.70289 0.512982 18.813 15.542 38.364 E-217' Volcan Darwin 0.70282 0.513022 18.79 15.545 38.331 E-224' Volcan Darwin 0.70286 0.513005 18.801 15.545 38.43 E-237' Volcan Darwin 0.70276 0.513053 18.686 15.519 38.167 E-242' Volcan Darwin 0.70289 0.512994 18.765 15.514 38.231 E-243' Volcan Darwin 0.70285 0.513005 18.794 15.544 38.341 E-63 Volcan Darwin (TagusC) 0.70278 0.513016 18.787 15.533 38.318 Marchena E-15 0.70276 0.513048 M-28 0.70280 0.513035 M-34 0.70280 0.513047 M-39 0.70287 0.513046 M-51 0.70281 0.513027 M-63 0.70277 0.513036 18.923 15.557 38.521 M-67 0.70278 0.513039 M071t 0.70281 0.513043 M10 0.70287 0.513021 18.91 15.558 38.522 M12 0.70291 0.513036 M31 groundmass 0.70276 0.513037 M31 p.lagioclase 0.70277 0.513033 M069• Caldera Wall 0.70285 0.70329 0.512902 E-8 0.70331 0.512887 19.346 15.631 39.302 P-24 plagioclase 0.70314 0.512922 P-24 groundmass 0.70313 0.512943 P-42 0.70338 0.512892 19.228 15.63 39.208 WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OF THE GAL•PAGOS 19,547

TABLE 3. (continued) Sample Location 87Sr/86Sr 143Nd/144Nd 206pb/204pb 207pb/204pb 208pb/204pb P-50 0.70319 0.512923 19.139 15.604 39.024 San Cristobal E-103 Southwest 0.70281 0.513042 18.838 15.536 38.386 SC-9 * Southwest 0.70287 0.513074 18.888 15.543 38.491 SC-20 * Southwest 0.70315 0.513023 18.974 15.548 38.558 SC-28 * Southwest 0.70284 0.513074 18.800 15.516 38.334 SC-34 * Southwest 0.70312 0.513032 18.794 15.550 38.433 SC-123 Northeast 0.70327 0.513047 18.776 15.558 38.430 SC-59 Northeast 0.70310 0.513030 18.729 15.549 38.369 SC-61 Northeast 0.70293 0.513006 18.824 15.562 38.484 SC-75 Northeast 0.70293 0.513052 18.702 15.552 38.378 C-332 Northeast 0.70312 0.513063 Pinzon E-68b 0.70298 0.512988 E-70 0.70311 0.513018 19.005 15.557 38.569 PS-16* 0.70291 0.512997 19.031 15.568 38.610 Rabida Island E-51* 0.70300 0.512980 19.028 15.553 38.611 E-174 0.70297 0.513013 G86-10' 0.70300 0.513013 19.142 15.568 38.746 Santa Cruz E-1 Academy Bay (Shield) 0.70263 0.513077 18.514 15.52 38.048 SC-46 Academy Bay (Shield) 0.70262 0.513088 SC-64 NW (Shield) 0.70274 0.513059 SC-78 N. Coast (Shield) 0.70262 0.513101 SC-100 NE Shield 0.70266 0.513074 18.656 15.521 38.147 SC-151 South Shield 0.70268 0.513094 SC-161 Summit (Shield) 0.70263 0.513085 18.511 15.517 38.032 SC-163 SE Shield 0.70267 0.513063 18.555 15.508 38.016 SC-200 Summit (Shield) 0.70261 0.513086 18.513 15.515 38.019 SC-108 NE Platform 0.70289 0.513020 18.774 15.564 38.45 SC-174 North Central, Platform 0.70286 0.513043 18.714 15.543 38.288 SC-196 NE Platform 0.70283 0.513014 SC-83 NE Platform 0.70281 0.513040 18.774 15.564 38.45 G86-1' South Las Plazas (Platform) 0.70296 0.513022 18.686 15.531 38.266 E-167 SeymourIsland (Platform) 0.70312 0.512997 18.946 15.566 38.553 E-31 (Platform) 0.70306 0.513034 18.719 15.556 38.337 E-111 0.70300 0.513033 G86-4' 0.70318 0.512980 18.814 15.548 38.504 G86-5' 0.70330 0.512934 18.965 15.584 38.807 SF-13 0.70293 0.513005 18.722 15.569 38.475 Santiago Island SH-7 * NW Highlands 0.70281 0.513053 18.843 15.559 38.446 SH-75a 1897 Flow (NW) 0.70275 0.513045 18.599 15.534 38.197 E-16 James Bay (NW) 0.70279 0.513027 18.856 15.549 38.404 E-20 Marmalade Pot Flow(NW) 0.70294 0.512980 19.022 15.582 38.726 E-76 Buccaneer Cove (NW) 0.70286 0.513002 19.05 15.58 38.656 E-24 Sullivan Bay (SE) 0.70272 0.513029 JL-94* SE flank 0.70277 0.513045 18.701 15.543 38.285 JH-86* EasternHighlands 0.70280 0.513057 18.554 15.550 38.164 BL-I* Bartolom6, S.E. 0.70279 0.513062 18.628 15.539 38.180 Wolf & Darwin Islands DA-1 Darwin 0.703008 0.512949 19.01 15.607 39.032 DA-2 Darwin 0.703032 0.512954 19.044 15.596 38.983 E-35 Darwin 0.70302 0.512984 E-39 Wolf 0.70300 0.513002 19.085 15.609 38.911 WO-1 Wolf 0.70337 0.512950 19.074 15.597 38.874 Smaller Rocks and Cones E87-2' Cowley Island 0.70306 0.512981 19.184 15.583 38.803 E87-1' Nameless Rock 0.70714 0.513051 18.716 15.545 38.281 E-156 Roca Redonda 0.70311 0.512964 19.334 15.599 39.048 Exceptwhere noted, values were determinedat the Max-Plank-Institutin Mainz (MPI). 87Sr/86Sris correctedfor mass fractionationbynormalizing to 86Sr/88Sr= 0.11940. Values are reported relative to 87Sr/86Sr = 0.70800 for theEimer and Amend (E&A) Sr isotope standard. Actual measuredvalues were 0.70792 at DTM, 0.70800 at MPI, and 0.70802 at Cornell. Estimated analyticaluncertainties are +0.00006 for theDTM dataand +0.000035 for theMPI andCornell data (2 (•). 143Nd/144Ndis corrected formass fractionation by normalizingto 146Nd/144Nd= 0.72190. Values are reported relative to 143Nd/144Nd= 0.51264 for U.S. GeologicalSurvey standardBCR-1. Actual measuredvalues for BCR-1 were 0.512695 at DTM, 0.512647 at MPI, and 0.512643 at Cornell. Measuredvalues for the La Jolla Nd isotopic standardwere 0.511847 at MPI and 0.511840 at Cornell. Estimated analytical uncertaintiesare +0.000040 for the DTM data, +0.000020 for the MPI, and +0.000016 for the Cornell data (2(•). Pb isotoperatios are reported relative to valuesof 2ø6pb/2ø4pb= 16.937, 2ø7pb/2ø4pb = 15.493, and ø8pb/2ø4pb = 36.705 for theNBS- 981 standard.A fractionationcorrection of 1.40%oper amu and 1.10%oper amu has been appliedto the MPI and Cornell data, respectively. Estimatedanalytical uncertaintiesare 206 Pb/204 Pb +0.010, 207 Pb/204 Pb: _-_4-0.016,and 208 Pb/204 Pb: _-_-•0.035(2 (•) * Determinedat the Departmentof TerrestrialMagnetism (DTM). * Determinedat CornellUniversity (Cornell). 19,548 WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OFTHE GAL•PAGOS

TABLE4. Comparisonof 87 SffaSSr in Leachedand Unleached nearlyparallel to the lithosphericfault systemthat M. A. Feighner GalapagosBasalts and M. A. Richards(submitted manuscript, 1993) infer divides Unleached Leached stronglithosphere to the southand westfrom weak lithosphereto E-4 0.70387 0.70329 the noah and east,with the more depletedcompositions occurring SC-59 0.70329 0.70310 in the region of weak lithosphere. There are important SC-196 0.70399 0.70293 differences, however. Wolf and Darwin volcanos have fairly E- 111 0.70341 0.70300 depletedisotopic signatures yet lie uponstrong lithosphere while WO-1 0.70345 0.70337 FL-3 0.70395 0.70366 Pinta, which hasa stronglyenriched isotopic signature, apparently lies on weak lithosphere. concentrationsstrongly correlate with isotope ratios. With the Figure 15 illustratesanother intriguing feature of Galfipagos exceptionof K8.o, none of the major element abundancesare geochemistry:the mostenriched isotopic signatures on the GSC do not occurat its closestapproach to the Galfipagos,rather they correlatedwith isotoperatios or with LaJSm. Na8.0 correlateswith K8.0 andweakly with Sm8.0 andYb8.0. Si8.0shows a weakinverse lie noahwestof the archipelago,at the intersectionof the Wolf- Darwin lineament with the GSC. If there is flow of plume correlationwith Srn/Yb8.o. Fe correlateswith no otherparameter. Within individual volcanic suitesNa8. 0 does seem to correlate :i:!11:i?i...... '-.-.-.-'•::•-::•-,•;•r [] SantaCruz* Pinta [] SantaCruz stronglywith LaJSm8.0 in somecases (Figure 14). The SantaCruz ::11!I::::!!I...... ' '-•'":'•::•••'•••:::i:•x•-GSC * Santa Fe [] Genovesa Platform shieldseries, San Cristobal,Cerro Azul, Santiago,and SantaFe all :•..".."s..'".-'"½i:ii•ii'"-'"'•"•':':'••'"•:•"•••••il• ß Marchenao I•bela + espanola ...... •- "'"'•"• nss'•:•:••••••:i•:-•© Santiago • Floreanaa Fernandina appear to exhibit such a correlation. This correlationprobably ...... •••••••••••:•'•'.•.:• • o San ß Pinzon, results from the melting process, which affects both Na20 :::::::::::::::::::::::::::::::::::::::::::::::::::::'%"•i?:•i•i•::•::?:?:?:•'•::•:: o • Darwin, concentrations and La/Sm ratios. In most other cases, there are too few data to establish whether a correlation exists. No within- i ...... :...... ::::::::::::::::::::::::::::• o suitecorrelations were observedbetween major elementsand iso- œNd7 • -ø•:• o o • •o • ffi tope ratios. o • • • O ffi

GeographicPatterns

Perhapsthe most remarkablefeature of the Galfipagosis the systematicvariation in isotoperatios, in which the mostdepleted 4 I • , • I I isotopicsignatures occur in the centerof the archipelagoand the O.7025 O.7030 O.7035 O.7040 most enriched signatures occur on the southern, western, and 87Sr/86Sr eastern margins. This pattern was first noted by White and Fig. 6. œNdand 87Sr/86Sr ratios of Galfipagoslavas. Lightshaded region Hofmann [1978] for Sr isotoperatios. The presentresults entirely showsthe field for East Pacific Rise MORB; dark shadedregion is the confirm the earlier findings. Further, this zonation can be field for GSC MORB. observedin Pb and Nd isotope ratios and incompatibleelement ratiosas well. The patternsfor 87Sr/86Sr,end, La/SmN, and 40 Sm/YbN are shownas computer-generatedcontour plots in Figure

15. Contoursare basedon averagesfor individual volcanos,with 39.5 locationsbacktracked to the point wherethe lava was erupted(i.e., correctedfor plate motion). Separateaverages were made for 39 %* ocø islands showing a range of ages or a geographiccompositional 8 O t• I [] S.Cruz Platform II '•' I[]s.cruz Os. cristO•a, I variations (Floreana, San Cristobal, Santiago,Santa Cruz). Data _.• • o I• Santa Fe [] Floreana I for basalts dredged from the Galfipagos Spreading Center •38.5 GSC •:'"'•. '••/":.. [* Pinta ß Espanola I /'""'"'"•'cY'•••-••i•s:• I ß Marchenaa FernandinaI [Schillinget al., 1982; Vermaand Schilling, 1982; Vermaet al., Pinzon, • GEnovEsa • 1983] were also usedin the contouring. Contourswere closedon 38 '•'•:•••••::"'":•EPR I RabidaI /••"•:/ I' Santiago Darwin, I the assumptionthat normal (i.e, average MORB) surroundsthe GalfipagosPlatform. 37.5 The geographicpatterns of 87Sr/86Sr,end, and La/SmN are 18.5 19 19.5 20 20.5 15.7 similar: they are horseshoe-shapedwith the most depleted sig- naturesin the northeasternand central partsof the archipelago.

Theydiffer in detail:maximum 87Sr/86Sr and La/SmN occur on 15.65 Floreanain the south,while two minima for ENdoccur; one on Pinta in the noah, and the other on Floreanain the south,though a secondarymaximum for 87Sr/86Sroccurs on Pinta. Thepattern •15.6 for Pb isotoperatios (not shown)is essentiallysimilar to that of 87Sr/86Sr,with the maximumon Floreana. Contoursof other incompatible element ratios (e.g., Rb/Sr) and incompatible 15.55 elementconcentrations generally show the samehorseshoe shape. The pattern for Sm/YbNs.0differs somewhatfrom thoseof the 15.5 otherratios shown; rather than a horseshoe,Sm/Yb•8.0 is highest 18 18.5 19 19.5 20 20.5 in the westernpart of the archipelago,with maximaat CerroAzul 206pb/204pb and Volcan Ecuador, the noahern and southernmost volcanos on Fig. 7. Pb isotoperatios of Galfipagoslavas. Light shadedregion shows Isabela. the field for East Pacific Rise MORB' dark shadedregion is the field for The contoursfor isotopeand incompatibleelement ratios are GSC MORB. WHITE ET AL.: PETROLOGYAND GEOCHEMISTRY OF THEGAL•,PAGOS 19,549

incompatible element characteristics indistinguishable from

SC MORB; the geographic zonation of isotope ratios and incompatible element concentrationsand ratios, with the most depletedsignatures occurring in the central and northeastpart of the archipelago;and the north-southdifferences in geochemistry:

7 e.g., lavas from the southernislands, particularly Floreana, have higherLa/Sm N, 2ø6pb/2ø4pband 87Sr/86Sr at the sameENd comparedwith northernislands.

• ß OCiC•)0 ffi I• o GeographicPatterns of IsotopeRatios and Plume-Asthenosphere Dynamics 4 . , . . I i , , , I , , , , I , , , , i , , i i 18 18.5 19 19.5 20 20 5 2o6Pb/204Pb The geographic elemental and isotopic variations raise the most fundamental questions, and we discuss them first. We 0.704 o Isabelaß SanTa Cruzo San consideredfour types of modelsto explain the observedchemical FI• s*,T, FE ø SantigaO •C.ristobal variations. The first model is that of a plume heterogeneouson a "I v,.•...... :. ß Pinzon, etc I'1' p' scale of 10-100 m, with incompatibleelement-enriched veins or marchena+ _Espa onl..a ' •Wo•lf lensesthat melt at lower temperaturethan the more depletedma- 0 7035 I_LE•_enovesaa Fernandi_na_a S.CruzI?T. trix in which they are embedded,i.e., the veined,or plum-pudding mantle consideredby Hanson [1977], Morris and Hart [1983], Batiza and Vanko[ 1984], and Zindler et al. [ 1984]. The plumeis

0.703 hottestin its center,so melting there would proceedto a higherde- gree, and melts producedwould reflect a higher contributionfrom '! '..:..'iil the depleted, higher solidus matrix. In the second model, a chemically homogenous plume impinges on the overlying 0.7025 I lithosphereand heatsand melts it. The westernvolcanos would 1• 185 19 19.5 20 20.5 reflect melting of the plume, while thoseto the east and over the central axis of the hotspot track would reflect an increased Fig. 8. ENdand 87Sr/86Sr for Galfipagoslavas plotted against 206pb/204pb. contribution from remelting of the depleted lithosphere(M. A. Light shadedregion showsthe field for East Pacific Rise MORB; dark Feighnerand M. A. Richards,submitted manuscript, 1993). In the shadedregion is the field for GSC MORB. third model, isotopicand incompatibleelement variations reflect material from the Galfipagos to the GSC, as Schilling and entrainment of asthenosphereinto the center the plume. In the coworkers have suggested[Schilling et al., 1982; Verma and fourth model, relict asthenosphericupwelling, reflecting the Schilling, 1982; Verma et al., 1983], then flow is "upstream"in former position of the GSC, producesextensional tectonism and persistentvolcanism of more MORB-like compositionalong the the sensethat regional plate motion, and presumablyuppermost centralaxis and northernmargin of the platform. asthenospheremotion, is to the east-southeast(100.5ø). The first of these models predicts a relationship between Geographic patterns for the major elements are not as well degreeof melting and isotoperatios, while the secondpredicts a developedas thoseof isotopeand incompatibleelement ratios, but relationshipbetween magma chemistry, particularly isotope ratios, somegeneral features are apparent:Na8. 0 is lowest in the west centralvolcanos but alsodecreases toward the GSC; Si8.0 appears and depthof melting. Degree of melting hasa stronginfluence on to show a broad, weak minimum over the west centralpart of the magma chemistry,independent of sourcecomposition, which af- archipelagoand may also increase toward the GSC; FeB. 0 is lowest fords an opportunity to test these models. These influencesof depth and degree of melting on magma chemistry have been in the northwesternvolcanos and highestin the volcanosof the central part of the archipelago,with a maximum on Santiago. reviewedby Klein and Langmuir[1987, 1989], and Geist [1992] hasdiscussed them in the contextof Galfipagosgeochemistry. As Thesepatterns are similar to thosefound by Geist [ 1992] usinga Klein and Langmuir [ 1987, 1989] havepointed out, Na is a moder- different, but overlapping,data set. Geist [1992] concludedthat ately incompatible element, and its concentration should be magmas from the center of the archipelagoare derived by a inverselyrelated to degreeof melting. Experimentalobservations relatively low average extent of melting and deeper average show that the Na20 concentrationof partial melts is essentially extractiondepths. independentof pressureand that SiO2 and FeO are sensitiveto DISCUSSION both degreeand depthof melting [e.g., Jaquesand Green, 1980]. SiO2 increaseswith increasingdegree of melting and decreases The distributionand compositionof volcanic activity in the with increasingdepth, while FeO increaseswith depth and de- Galfipagosregion present a number of puzzling features that creasesas degreeof melting increasesin an upwellingcolumn. requireexplanation. These includethe following: the numberof If the isotopic variations in the Galfipagosreflect varying simultaneouslyactive volcanosin the Galfipagosarchipelago; the degreesof meltingof a plum-puddingmantle, 87Sr/86Sr, for east-west differences in size and structure of volcanos; the example, shouldshow a positivecorrelation with Na20, and an extended period of volcanism on a number of islands, most inverseone with SiO2 if depthof meltingwere constant.Table 7 notably San Cristobal, SantaCruz, and SantaFe; the restrictionof showsclearly that this is not the case. Indeed, averageNa8.0 is differentiatedrocks to the center of the archipelago;the lower highest on Santa Cruz, where isotoperatios approachthose of SiO2 contentof Galfipagoslavas relative to MORB; the great MORB, and lowest on the central and southwestern volcanos and variety of isotopeand trace elementcompositions displayed by Fernandinain particular. If the Na8.0 value can be taken as an Galfipagoslavas; the presenceof lavas with isotoperatio and indicatorof relative degreeof melting as suggestedby Klein and 19,550 WHITE ET AL.' PETROLOGYAND GEOCHEMISTRY OF THE GAL,•PAGOS WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THEGALJ•PAGOS 19,551

• .

--00660606666 660 6o 006 0060 60060606 o 6 19,552 WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OFTHE GAL,&PAGOS

0 0 •:• 0 0 O•

o o o c5 c5 o

o • T-- c5 c5 c5 o c5 c5 o o o c5 o c5 o o o c5

• o

• 0 WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THE GAL/•PAGOS 19,553

300 sphereis one possiblemechanism for this dramaticdifference in •••.•1Fernandina A!cedo•_ SierraNegra lithosphericstrength. We believe, however,that it is unlikely that _• E-42 O E-132 _' SN-1 E-õ7_• E-134-" SN-3 remelting of the lithosphere can account for the observed lOO Cerro Azul geochemicalpatterns. Emermanand Turcotte[1983] concluded 35n• NOC5ZOB6 that conduction alone is unlikely to supplysufficient heat to thin the lithosphereabove a plume. The only alternativemechanism to conductionis heatingby plume-derivedmelts passing through the lithosphere. It is unlikely that thesemelts are superheated;thus the energy for heatingand melting the lithospherecan be derived

lO only from their crystallization. Since energy for melting is most SouThern Isabela & Fernandina abundantin direct proximity to conduitsand diapirsof thesemelts,

• I I I I I I I I I I I I I I it seemslikely that meltsproduced in the lithospherewill mix with ta Ce Pr Nd PM SM EU Cod Tb Dy Ho Er TM Yb Lu the plume-derivedones.

100 To evaluatethis model, we needto estimatethe compositionof NorThern Isabela magmasthat could be producedby remeltingthe lithosphere.As noted,the lithospherewould have gonethrough extensive melting at the GalfipagosSpreading Center. If the melt producedby this processwere extractedwith 100% efficiency, the residualmantle would be highly depleted. If some melt remained trapped and crystallized in place, the depletion would be less extreme. We R. Redonda consideredthe case where 1% melt remainstrapped in the litho- • E-156 sphereafter generationand extraction of a reasonablyprimitive V. Darwin _• E-224 V. Wolf 10 O E-63 • E-237 • E-152V. Ecuador LREE-depleted MORB (e.g., 5 ppm Nd) by 12% melting. .• E-216 O E-242• E-153 ß B155 •- E-217 O E-243 --. E-155 a E-56 200

I I I I I I I I I I I I I I I Floreana 5 100 La Ce Pr Nd PM SM Eu Gd Tb Dy Ho Er TM Yb Lu BasalTs • FL-27 -' FL-3 -'- FL-29 Santiago I_- e-20 • e-?6-o--sH-? II •- '- FL-19 --6- E-110 •' L a E-24+ -* JL-94JH-õ6• SH-75aBL-1 I •• '- FL-25 xenoliTh ß FL-26 .-.o- E-98

10

' _ 5 LA Ce PR Nd PM SM EU Gd Tb Dy Ho ER Tb Yb Lu 5 I I I I I I I I I I I I I I I 70 La Ce Pr Nd PM SM EU Gd Tb Dy Ho Er Tb Yb Lu Espa•ola SanCrisTobal '- El02 ß C332 -o-- SC-75 ? G86-3* E-103-• SC-9 Ill••'••'"-'""-"'"'•:--'•'•';•:•:.••.••:::...::.• • H 334 --• SC - 59 -• SC - 123 Fig.Fernandina9. Chondrite-normalizedand Santiagolavas. abundancesNote the parallelismofrare earths of patternsinIsabela, from 'i _ Isabelaand Fernandina anddiversity ofpatterns from Santiago. Shaded • areashows therange ofrare earth patterns fromCerro Azul (D. J. Geist, • unpublisheddata,1992). Chondritic normalizing factorsbased on the • averageof20 ordinary chondrites ofNakamura [1974]. • 10 ß Langmuir [ 1987], then theserelationships indicate that isotopera- b SanCristobal &Espa•ola tios, which presumablyreflect source compositions, are unrelated 5 I I I I I I I I I I I I I I I La Ce Pr Nd PM SM EU Gd Tb Dy Ho Er TM Yb Lu todegree of melting. 300 Finally, both rare earth abundancesand Na8.0 suggestthat magmasof the Santa Cruz shield serieshave been generatedby widely varying extentsof melting (Figure 14 and subsequent .• rAbida&Pinzon c discussion).Yetthe isotopic compositions ofthese magmas show '• only limitedvariations. Indeed, there are no correlations• whatsoeverbetween isotope ratios and NaB.o, SiOs.o, or FeB.0 • withinindividual volcanic suites. Theseobservations are in- • r/,bida...... •::•:;:::::::•i::i•::i::::??:i?:i•ii::iiii!i•!!!!!•!i•iiiiii•i•!!?•?•ii•??:•::•::•::•::•*:•.•.•.!i! ...... ::::::::!::iiiiiiii!ii!•:•iii::i::::::::•::¾.....•...... iil.'"-iii:!ii!i!i!!!!? consistent with the veined or plum-pudding plume model. 10 Although small-scale hetergeneitiesprobably do exist in the mantle beneath the Galfipagosand contribute to local magma -o--- E87-1 • PS16 e E87-2 I ...... variability,we concludethat variableextent of meltingof themis 3 I I I I I I I I I I I I I I I LA Ce PR Nd PM SM EU Gd Tb Dy Ho ER Tb Yb Lu not responsiblefor the large scalepatterns observed in Figure 15. The secondmodel, remeltingof the lithosphere,receives some Fig. 10. Chondrite-normalizedabundances of rare earthsin lavasfrom the support from the analysis of geophysicalobservations in the southernislands of Floreana,San Cristobal, and Espafiola and the central islandsof Rfibidaand Pinz6n. Shadedarea in Figure 10b is the field of Galfipagosby M. A. Feighner and M. A. Richards (submitted San Cristobal rare earth patternsof Geist et al. [1985a]. Shadedarea in manuscript,1993). The lithosphereappears to be particularlythin Figure 10c showsthe rangeof rare earthpatterns reported by Shimizuet underthe centralpart of the archipelagoand remeltingof the litho- al. [1981] from Rfibida. 19,554 WH1TE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THE GAL•PAGOS

lOO Experimentalstudies, reviewed by Klein and Langmuir [1987], . PinTa& Marchena Pinta marchenaj showthat melting in this pressureregime shouldproduce magmas ]"::?'....":. .'.....'•.'•..•..• .... --P-,o--E,OI dramatically poorer in FeO than those producedin the pressure i"":':'""":'':'"•'.-••.I•.:.•...'-•-..••';•:'"'":'":.:•''"'•';•':'...-":::'"' '-'-... ' ß o E4 -• E-15I regime of 0.5-1.5 GPa, which is more typical of MORB "•...... + E8 [] M071I production[Klein and Langmuir,1987]. For instance,a 3% melt at 0.5 GPa shouldhave an SiO2 concertrationof about49% and an FeO concentrationof about6%. This contrastswith meanSis.0 and FeB.0 of 47.6 and 11.4, respectively,for the SantaCruz shield • -"__•_ --.--•--.--.•series. The low SiO2 and high FeO concentrationsof SantaCruz I a I I I I I I I I I I I I I I magmasare very difficult to reconcilewith suchshallow melting. Thus consideration of the likely melting mechanism, the La Ce Pr Nd Pm sM Eu Cad Tb D¾ Ho Er Tb Yb Lu 40 consequentmixing relationship,and compositionalrelationships lead us to reject remelting of the lithosphereas a causeof the GENOVESa compositionaland isotopiczonations observed. We concludethat the geographicpattern of isotope and in- compatible-elementratios in the Galfipagosmust reflect large- scale chemical variations in the mantle. White and Hofinann : E-171 -' Ca303 [ 1978] pointedout that the geographicvariation in isotoperatios is b •- E-172: Ca304 opposite that predicted by the model of a rising mantle plume I I I I I I I I I I I i I I I mixing radially outwardwith surroundingdepleted asthenosphere, La Ce Pr Nd PM SM Eu Cad Tb D¾ Ho Er Tb Yb Lu 40 which Schilling [1973] proposedto explain geographicvariation Darwin& Wolf m.•in wou •1 in rare earth abundancesin Iceland and the ReykjanesRidge. - o e-:l Geist et al. [1988] proposedthat the geographicpattern results -a_. DA-1 -• WO-11 from the Galfipagos mantle plume being torus-shaped. This

_• A A possibilityemerges from fluid-dynamicexperiments of Griffiths [1986] and occurswhen the medium surroundinga rising ther- mally buoyantdiapir is conductivelyheated and entrainedby the diapir. The resultis a segregatedtorus of originaldiapir material surroundinga center of entrainedmaterial. Geist et al. [ 1988] ar- I I I I I I I I I i I I I I I gued that in the caseof an incompatibleelement-enriched mantle La Ce Pr Nd PM sM Eu Cad Tb D¾ Ho Er Tb Yb Lu plume rising through and entrainingdepleted asthenosphere, the Fig. 11. Chondrite-normalizedrare earthpatterns of the northernislands mostdepleted magmas should erupt above the centerof the plume, of Pinta,Marchena, Genovesa, Darwin, andWolf. Shadedarea in Figure 100 1lc showsthe rangeof rareearth patterns from Marchena determined by Vicenziet aL [1990]; cross-hatchedarea is the rangeof rare earthsfrom Santa cruz Pintareported by Cullenand McBirney[1987]. Rareearth patterns from individualvolcanos are parallel.

Assuming5% clinopyroxenein this residue,a 1% "remelt"would have a La/SmN ratio of 0.5 and a Nd concentrationof 3.7 ppm (higherpercent melts would be moredepleted). Assuming a .7• primitiveplume-derived magma with a composition similar to that of FL-3(ENd = 5.5)and that the lithosphere hasENd of 10,a 1'1 mixture of remelted lithosphereand plume-derivedmagma (a morerealistic estimate would be 1' 10 or 1' 100) would have ENd of PlaTform Series Shield Series 6.6 and La/SmN of 2.6. Assuming0.1% remeltingof the litho- _' SC-108 _- SC-174 [] E-1 .• SC-151 A SC-83 .-' G86-1 .-. SC-46 .•- SC-161 sphereor a somewhatmore enriched composition (e.g., ENd= 8) o SC-196 Seymour Is. -•--SC-78 •_ SC-163 does not substantiallychange these results. Thus becausethe _' SC-64 • E-167 = SC-100 O SC-200 lithosphereshould be highly depleted, and because any melts pro- 2 ! I • I • • I I I I I I I I I duced within it will mix with more enriched plume melts, La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tb Yb Lu remeltingof the lithospherecannot produce the magmacom- •o0 positionsobserved in the centralGalfipagos. These arguments do ••-• SantaFe : sF-I•I notpreclude the possibility that heat lost from ascending magma is ._• -.- I important in thinning of the lithosphere beneath the central Galfipagos'weargue only that this process will have an in- eruptedsignificantmagmas. effecton isotope andincompatible elementratios or the < lo The lithosphere-remeltinghypothesis also predictsa specific geographicpattern for thoseelements sensitive to depthof La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tb Yb Lu melting:mean depthof equilibrationshould be shallowestin the Fig. 12. Chondrite-normalizedrare earthpatterns of the centralislands of centraland northernparts of the plateauwhere the lithospherehas Santa Cruz (including Seymour, Baltra, and Las Plazas) and Santa Fe beenthinned. The "missing"lithosphere in the depthregion of 6- The diversityof thesepatterns contrasts with thoseof the large western 12 km corresponds to pressures of less than 0.5 GPa. volcanos. WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OF THE GAL•PAGOS 19,555

5O relative to the Galfipagosplume complicatethe recent volcanic historyof the area,so that the plumeflux is difficult to estimate. Thougha recentincrease in plumeflux cannotbe ruledout, there is little compellingevidence for it. Anotherdifficulty with thismodel arises if theplume is subject to shearfrom horizontalmotion of asthenosphere.Geist et al. [1985a] suggestedthat the Galfipagosplume may be shearedby asthenosphericflow, and this may account for the unusual geochemicalpatterns in the Galfipagos.Experiments by Richards •o NmoS.ß mm and Griffiths[ 1989]with toroidaldiapirs subject to velocityshear show an asymmetricdistribution of original diapir materialand entrainedasthenosphere with the highestconcentrations of diapir 1 material"downstream" of thehotspot center. The Nazcaplate has 20 movedeast-southeast at 37 mm/yrover the last 3 m.y. [Gripp and Gordon,1990], suggestingthere should be a significanteastward asthenosphericmotion. Shearof the Galfipagosmantle plume re- lO sultingfrom this motionwould not producethe patternobserved. Thus the toms plume model proposedby Geist et al. [1988] encounterssubstantial, and possiblyinsurmountable, difficulties. The sameseries of fluid dynamicalexperiments by Richards and Griffiths[ 1989] suggestsan alternative,though fundamentally similar, mechanismfor entrainmentof asthenosphereby the plume. They found that steadystate plumes subject to velocity shear will entrain surroundingmaterial in a manner somewhat analogousto that found in the diapir experimentsof Griffiths , , , I , , , I , , , I , , m I m m m I , , m I m m m [ 1986]. As the surroundingasthenosphere is conductively warmed by a plume, it becomesthermally buoyant. As the plumeis bent by velocity shear,this buoyantasthenosphere becomes entrained in the center,and the original plume materialconcentrates in the edgesand a thin veneer acrossthe top. The result shouldbe a hotspottrace with stripesof "enriched"compositions bounding a centralstripe of more depletedmaterial in the center. The idea is illustratedin Figure 16, which showsnorth-south and east-west cross sections of the bent-plume hypothesis applied to the Galfipagos. The first part of the plumeto reachthe depthwhere melting beginsis the upper veneerof original plume and entrainedas- thenospherebeneath it. This melts to produce basalt with

, , , I , , , I , , , I , , m I , m m I m m m I m m m moderately enriched trace element and isotopic characteristics, 0 2 4 6 8 10 12 14 suchas that eruptingon Femandina.Farther to the east,melting MgO. •/t o• along the central axis taps entrainedasthenosphere with only a small admixtureof plume material. Basaltproduced in this area Fig. 13. Abundancesof La, Sm, and Yb as a functionof MgO in will have more depletedsignatures, similar to thoseerupting on Galfipagoslavas. Linesdrawn through the datashow the slopeof the functionsused to correctthe abundancesof the elementsto an MgO Volcan Darwin. Fartheralong the axis of the plume to the east, contentof 8%. Theseequations are LaB. 0 = Lax e0-1827(MgO-8-0),Srng.0 = melts are derivedentirely from this entrainedasthenosphere as it Smx e0-1443(MgO-8-0),Ybs.0 = Yb x e0.07719(MgO-8-0)- Rareearth data from crossesthe minimumdepth for melting,while the originalplume Cullenand McBirney[1987], Vicenziet al. [1990],Geist et al. [1985a], material above it becomesexhausted of its basalticcomponent. and unpublisheddata from Cerro Azul of D. J. Geist are also includedon thisand subsequent figures. Thus along the axis of the plume, the isotopicand traceelement signatureof thebasalts becomes increasingly depleted to theeast. A north-southcross section at the longitudeof Isabelashows with moreenriched ones on the periphery.In a generalway, this that volcanosalong the axis of the plume, suchas Volcan Wolf matchesthe patternobserved in the Galfipagos.Duncan et al. and Volcan Darwin, tap meltsproduced primarily by meltingof [1986] invokeda similartoms-plume model to explaintemporal entrainedasthenosphere. Alcedo, which sits over the boundary variations in the isotopic compositionsof basalts from the betweenplume and entrainedasthenosphere, has erupteda series MarquesasIslands. of magmasof intermediateisotopic composition that are notably Griffith's [1986] experimentalresults apply to the transient more heterogeneous than those of its northern and southern caseof a thermalplume head rather than a continuousplume. The neighbors. To the north and south, volcanos such as Roca Cocosand Carnegie ridges, however, preserve a recordof essen- Redondaand Sierra Negra are tappingmelts of predominantly tially continuousactivity of the Galfipagosmantle plume for originalplume material. Thoughthe cartoonin Figure 16 greatly perhapsthe past25 m.y. Experimentsby Richardsand Griffiths simplifiesthesituation, weconclude that the 87Sr/86Sr, end, and [1989] showthat similar toroidialdiapirs can be producedby La/SmN patternsshown in Figure 15 are generallyconsistent with thermalconduit waves resulting from suddenincreases in plume thatexpected for meltingof a plumebent by velocityshear. flux. Changesin the locationof the GalfipagosSpreading Center Figure 15 indicatesthat the plumehas a substantialinfluence 19,556 WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OFTHE GAL/•PAGOS

TABLE 6. VolcanoAverages Volcano Si8.0 Fe8.0Na8.0 K8.0 87Sr/86Sr œNd 206pb/204pb Ba La8.0 Sm8.0 Yb8.0 LaJSmN8Sm/YbN8 Santa Cruz Platform 47.32 10.42 2.87 0.38 0.70294 7.54 18.769 80.91 10.20 4.99 3.31 1.26 1.64 Santa Cruz Shield 47.65 11.49 3.4 0.38 0.70265 8.64 18.550 43.40 9.66 4.83 2.84 1.24 1.84 Santa Fe 47.66 10.35 3.18 0.65 0.70310 6.83 18.834 195.04 14.45 5.78 3.14 1.54 2.00 Marchena 47.87 10.44 2.96 0.31 0.70281 7.79 18.917 43.28 5.58 3.90 2.94 0.88 1.43 Pinta 47.98 9.74 2.79 0.69 0.70324 5.30 19.238 170.71 15.88 5.49 2.80 1.79 2.12 Genovesa 48.14 10.14 2.75 0.11 0.70269 9.41 18.435 8.91 2.12 2.87 2.90 0.46 1.07 Volcan Ecuador 47.07 9.09 3.06 0.58 0.70311 6.27 19.253 164.00 15.62 5.52 2.44 1.75 2.45 Volcan Wolf 47.36 8.37 2.68 0.37 0.70272 8.11 18.900 60.50 10.43 4.92 2.33 1.31 2.29 VolcanDarwin 47.53 9.39 2.80 0.40 0.70284 7.35 18.778 69.82 11.60 5.30 2.80 1.35 2.05 Volcan Alcedo 47.95 10.28 2.62 0.43 0.70321 6.57 19.216 151.14 9.07 3.63 2.16 1.54 1.83 SierraNegra 47.04 10.95 2.51 0.58 0.70333 5.55 19.374 128.93 13.53 5.29 2.80 1.58 2.05 Cerro Azul 47.93 9.81 2.90 0.53 0.70334 5.88 19.319 136.50 15.99 5.29 2.33 1.87 2.47 Fernandina 47.43 9.18 2.43 0.29 0.70319 5.92 19.089 82.33 10.30 4.30 2.17 1.48 2.15 San Cristobal 48.45 10.44 2.97 0.61 0.70305 7.78 18.805 69.83 11.45 4.31 3.03 1.64 1.54 Floreana 48.20 10.47 3.34 0.99 0.70348 5.98 19.826 268.48 25.15 4.21 2.39 3.69 1.91 Espafiola 47.40 10.45 2.97 0.59 0.70294 7.63 18.743 87.47 20.08 5.62 3.30 2.20 1.85 SantiagoNW 47.01 11.57 2.98 0.45 0.70285 7.36 18.943 47.16 13.53 6.03 3.01 1.39 2.17 SantiagoSE 47.39 10.98 2.86 0.34 0.70276 7.99 18.621 18.74 5.45 3.42 2.49 0.98 1.49 Pinzon 47.26 9.97 2.77 0.38 0.70300 7.08 19.018 77.73 11.24 4.90 2.57 1.42 2.07 Rabida 48.50 8.18 3.47 0.78 0.70299 7.05 19.085 163.35 16.94 7.08 3.46 1.48 2.22 Darwin-Wolf 48.96 9.36 2.56 0.29 0.70309 6.43 19.053 65.66 8.08 3.75 3.00 1.33 1.35 Na8.o, Si8.o, K8. o, and Fe8. o arethe concentrations of Na20, SiO 2, K20, and ZFeO corrected for fractional crystallization tothe equivalent of 8% MgO [Kleinand Langmuir,1987]. Lavasthat were obviously cumulates, those with <5% MgO, andthose with >13% MgO werenot usedin computingNa8. o, Si8.o, andFe8.o. Corrections were made using the equations of Geist[1992], which are for MGO>9%(olivine only saturated); Y8.o= Y +a(MgO-9)+b; andfor MGO<9% (multiply saturated); Y8.o = Y +b(MgO-8).For the volcanoes of Isabelaand Fernandina as well as Pintaand Rfibida, bNa - 0.114,bsi = 0.471,bFe - 1.21,b K = 0.571,and a wasnot applicable. For the remaining volcanoes, bNa = 0.211,bsi - 0.625,bFe = 0.678,b K = 0.139;aNa- 0.0687,asi--0.282, aFe = 0.0153,and a K = 0.0115. on the geochemistryof the mantlebeneath the GSC, asSchilling et solutionto this dilemmais flow of plumematerial to the ridgeat a al. [1976] previouslyobserved. Vermaet al. [1982] andMorgan much deeper level, perhaps related to deep asthenospheric [1978] suggestedthat flow of the plume to the ridge was counterflowtoward the ridge. channeled at shallow levels beneath the Wolf-Darwin Lineament. Indeed, Morgan [1978] suggestedthe Wolf-Darwin Lineament Degreeand Depth of Melting was the manifestationof this shallow flow. However, Figure 15 suggeststhe flow occursover a muchbroader area. Furthermore, Degree and depth of melting can be assessedfrom a con- Harpp et al. [ 1990] found that isotoperatios in basaltsdredged siderationof rare earth and Nd isotopesystematics. Figure 17a from the Wolf-Darwin Lineament become more plume-like showsENd plotted against [La/Sm]N8.0. The two are correlated,as toward the GSC. This observationis not obviouslypredicted by would be expectedfrom plume-asthenospheremixing. There is the channelizedshallow flow hypothesis. nonethelessconsiderable scatter, which can be modeled as variable The most plume-likeisotope ratios on the GSC occurnear it's extents of partial melting superimposedon this mixing. The intersection with the Wolf-Darwin Lineament [Verma and simplecalculations shown in Figure 17 illustratethis. Schilling,1982; Vermaet al., 1983] northwestof the archipelago. A partial melt will have higher La/Sm than its source,and This requires plume flow in a direction nearly oppositeto the henceany œNd-La/Smcurve for mixing betweensources should absolutemotion of the Nazca plate, and presumablythe shallow define the minimum La/Sm ratio observedin basaltswith a given asthenosphere,since the latter'smotion is undoubtedlyviscously end. We constructeda binary mixing model based on this coupled to the overlying plate. We suggestthat a possible observationand on the maximum and minimum observedend.

TABLE 7. CorrelationMatrix for VolcanoAverages 87Sr/86Sr ENd 2ø6pb/2ø6pbLa/SmN8.0 Sm/YbN8.0 Ba La8.0 Sm8.0 Yb8.0 Si8.0 Fe8.0 NaB.0 Ks.0 87Sr/86Sr 1.00 end -0.90* 1.00 2ø6pb/2ø6pb 0.88* -0.85* 1.00 La/Smss.0 0.71* -0.56* 0.75* 1.00 Sm/Yb$8.0 0.38t -0.55* 0.50t 0.38t 1.00 Ba 0.80* -0.71' 0.80* 0.78* 0.49t 1.00 LaB.0 0.62* -0.56* 0.66* 0.90* 0.60* 0.79* 1.00 Sms.0 0.14 -0.31 0.19 0.24 0.73* 0.39t 0.62* 1.00 Ybs.0 -0.31 0.31 -0.41t -0.19 -0.31 -0.14 0.05 0.41? 1.00 Sis.0 0.17 -0.01 0.10 0.10 -0.43? 0.15 -0.03 -0.26 0.20 1.00 FeB.0 -0.10 0.20 -0.19 0.01 -0.36 -0.20 -0.10 -0.23 0.13 -0.27 1.00 NaB.0 -0.11 0.22 0.00 0.33 0.13 0.35 0.45t 0.43t 0.40? 0.17 0.12 1.00 Ks.0 0.63* -0.53? 0.66* 0.83* 0.45? 0.86* 0.90* 0.55* 0.12 0.15 -0.06 0.57* 1.00 Valuesare basedon 124 analysesaveraged among 21 volcaniccenters. Darwin andWolf Islandswere combinedin a singleaverage. Lavas from Santiagowere divided between northwest and southeast, those from SantaCruz weredivided between Platform and Shield series. *Correlationssignificant at the 0.5% level. ?Correlations significant atthe 5% level. WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THEGAL,•PAGOS i 9,557

spinel peridotitemelting). As may be seen,melting of garnet peridotiteproduces strong enrichment of Sm over Yb in the melt, ß Santiago ß PInta ß Plnzon-Rablda [] Genovesa whereasmelting of spinel peridotitedoes not. Most Galfipagos • Wolf-DarwIn ß Marchena basalts,particularly those from Marchena,Santa Cruz, Santiago, o S Crlstobai • Santa Fe ß S Cruz Platform [] Fioreana Fernandina,and most of the Isabela volcanos,have high Sm/Yb ß SantaCruz .!. Espa•la ratios,implying that they were generatedprimarily in the garnet 0 Cerro Azui ß V Wolf stability field. Basalts from the southernislands, particularly [] SIerra Negra ß V DarwIn ,x Fernandlna ß V Aicedo those from Floreana and San Cristobal, have lower Sm/Yb ratios, [] [] [] ß .!. o o indicating they were producedprimarily in the spinel stability o 0 _Ooo field. Bow [1979] and Bow and Geist [1992] also concludedthat Floreanamagmas were producedwithin the spinelstability field, whereas Geist et al. [1985a] concludedSan Cristobal magmas were producedwithin the garnet stability field (however,D. J.

[][] Geist (personal communication, 1992) now agrees that San Cristobal magmas were derived primarily within the spinel sta- 0 I I I I I I I I I I I I I m I m bility field). Darwin and Wolf islandsalso appearto be products 2 3 4 5 of relatively shallowmelting. NA8.0 When end,La/Srn, and Sm/Yb are consideredsimultaneously, Fig. 14. Variationof La/SmN8.0with Na8.0 in Galfipagoslavas. Although mostGalfipagos basalts cannot be modeledas meltsof eitherjust there is no overall correlation,there are apparentwithin-suite correlations garnetperidotite or just spinelperidotite. They are morelikely the for SantaCruz shield, Santiago,San Cristobal,Cerro Azul, and perhaps SantaFe. These correlationsare consistentwith the predictedeffects of productsof polybaricmelting that beginsin the stabilityfield of variableextent of partial melting. garnetand endsin that of spinel,which is expectedfor a rising mantle diapir. Salters and Hart [1989] found suchpolybaric The compositionsof the depletedasthenospheric and enriched melting necessaryto explain Lu-Hf and Sm-Nd systematicsin plumecomponents, which are basedon thesecriteria, are givenin MORB. Figure 17 also showscalculated melting pathsfor three the caption to Figure 17. We then calculatedthe effects of examplesof "progressivemelting". In eachcase, 1% batchmelts equilibriumpartial melting of this mixture. While the absolute were produced,extracted, and pooled;the curvesshow the com- degrees of melting are arbitrary and depend on the choice of positionof the pooledmelts. In curveA, the first 6% melting(i.e., mantlecomponents, Figure 17a neverthelessallows comparison of first six increments)occurs in the garnet stability field, with the relative degreesof melting. In general,the northern(Marchena, remainderin the spinelstability field. In curvesB andC, the first Genovesa, Pinta, Wolf, and Darwin) and western volcanos, 2% and 1% melt increments, respectively,occur in the garnet particularlyFemandina and Sierra Negra, are the productsof the stabilityfield, and all subsequentmelts form in the spinelstability largestdegrees of melting. The least melting occursunder Santa field. Theseexamples of progressivemelting show that rare earth Cruz and the southernislands of Floreana, Espafiola, and San patternsretain the signature of garnet even if only a small Cristobal. Figure 17 also indicatesthat degreeof melting under percentage(10-20%) of the total melting occursin the garnet theseislands, particularly Santa Cruz, Floreana,and San Cristobal, stabilityfield. This is becausea relativelylarge fractionof the in- has beenquite variable. compatible elements are extracted in the earliest melting These conclusions,however, are based on our assumptionof increments. These early melting incrementstherefore dominate simplebinary mixing. As we discussbelow, thereis evidencethat the incompatible element budget of the pooled melts. These the plume itself is laterally heterogeneous.Our mixing model is calculations thus reinforce the conclusion that some of the basedon compositionsappropriate for the northernislands, which Floreana,Espafiola, and San Cristobalmagmas must have been may not be appropriatefor the southernislands, particularly San producedat shallow depth as they retain no gamet signature Cristobal and Floreana. It is possible,therefore, that lavas from whatsoever. these islands are generatedby larger degrees of melting than The inferenceson degree and depth of melting derived from suggestedby Figure 17. We have also assumedconstant dis- rare earth and Nd isotoperelationships in Figure 17 are generally tributioncoefficients in our modelling. However, Gallahanand consistentwith variationsin Si8.0,Na8.0, and Fe8.0,though there Nielsen [1992] have demonstrateda systematicdependence of are somedifferences. For example,the high Na8.0values for lavas clinopy.roxenerare earthpartition coefficients on temperatureand from Floreana,Santa Cruz, Espafiola,and San Cristobalare con- melt composition.The resultof thesedependencies is that parti- sistentwith low degreesof melting inferred from œNd-La/SmN tion coefficientsdecrease with increasingextent of melting. The systematics. Lowest Na8.0 values occur in in the westernand use of variable partition coefficientswould allow productionof northernpart of the archipelago,implying the greatestdegrees of high La/Sm meltsby slightlygreater degrees of meltingthan those melting, and highestmantle temperatures,in thoseareas. Again, implied by Figure 17. this is consistentwith the inference we made above from œNd- Figure 17b showsSm/YbN8.0 plotted against La/SmN8.0. Also La/SmN systematics. Low Fe8.0 in the southernisland lavas shownis the samemixing curve as in Figure 17a. Becausegarnet impliesrather shallow melting, consistent with the low Sm/YbN strongly fractionates Sm and Yb during melting whereas ratios. The highFeO8.0 ratios for SantaCruz andSantiago suggest clinopyroxene,the only othermantle phase accepting significant deeper melting beneath the center of the archipelago, as do amounts of the rare earths, does not, the Sm/Yb ratio should be Sm/Yb-La/Sm systematics.However, low Fe8.0 for lavas from particularly sensitive to the presenceor absenceof garnet and Fernandina and the Wolf and Ecuador volcanos of northern thereforeto depth of melting. Thus we have calculatedseparate Isabela suggestsshallow melting, yet theselavas generallyhave melting curves for garnet and spinel peridotite (we have not high Sm/Yb, indicative of melting in the garnet stability field. specificallyconsidered melting of plagioclaseperidotite, which The discrepancymay reflect the observationwe made above, will producefractionations similar to, but lesspronounced than, namely, that incompatible element abundances in pooled 19,558 WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OF THE GALfi, PAGOS

I I I I 93 ø W 92 ø W 91 ø W 90 ø W 89 ø W 88 ø W 93 ø W 92 ø W 91 ø W 90 ø W 89 ø W 88 ø W k ' ' I •e' e•' / •.•'0•/•' (SM/Yb)N8' 0 ß ':• {ua/Sm)n80 s '•• •x. '

2.O

-- 0 . •:::s. 0.• o o=', _

Fig. 15. Computergenerated contours of (a) 87Sr/86Sr,(b) ENd,(C) LaJSmN8.0,and (d) Sm/YbN8.0in the Galfipagos Archipelagoand adjacent GSC, thelatter with thedata of Schillinget al. [1982], Vermaand Schilling [1982], and Verma et al. [ 1983]. Contouringwas basedon averagesfor individualvolcanos for the islandsand individualsamples for the GSC. Locationswere correctedfor plate motionto the positionoccupied at 'the averageeruption time. Two averageswere computedfor Santiagobecause of systematicdifferences in compositionbetween the northwestand southeast. Two averages,with lavasgrouped by age,were alsocomputed for Floreana,San Cristobal, and SantaCruz. progressiveor "sequential"melts are controlledby earliermelting concentration(Figure 2). This suggeststhat the degreeof melting increments,but this is not the casefor compatibleelements such beneaththe Galfipagosis similar to that beneathmid-ocean ridges as iron. On the otherhand, our datafor thesevolcanos are sparse, but that the depth of melting is somewhatgreater (assuming the andhence we havelittle confidencein the low averageFEB.0. plume and asthenospherehave similar major elementcomposi- It is interestingto compareour inferenceson depthand degree tions). The similarityin degreeof meltingis somewhatsurprising, of meltingwith the structureinferred by M. A. Feighnerand M. A. given the high inferred temperaturesbeneath the Galfipagos Richards(submitted manuscript, 1993). We concludemelting is [Schilling, 1991]. We note, however, that both Fernandinaand shallow and of relatively small degree beneath Floreana and SierraNegra seemto have lessNa20 than averageMORB. This Espafiola,which seemsconsistent with the thicker lithosphere suggeststhat the plume component is indeed hotter than the (implyingcooler temperatures) and thinner crust inferred by M. A. asthenospherebeneath ridges. The lower temperatureselsewhere Feighnerand M. A. Richards(submitted manuscript, 1993). They in the Galfipagosmay reflect conductiveheat loss from the plume find the crustto be quitethick beneath the centralarchipelago, par- to the surroundingasthenosphere. ticularlyeastern Isabela, where it appearsto reacha total thickness We pointedout earlier that Galfipagoslavas have more Na20 of 18 km. From our inferences,we would expectmost crustal and lessSiO2 andZFeO than Kilaueanlavas, which implieslower thickeningto occur in the west, in the area of Fernandinaand degreesof melting, and perhapsgreater depths of meltingbeneath Isabela,where degreeof melting is largest. The thick crustM. A. the Galfipagos(assuming the Hawaiian and Galfipagosplumes Feighnerand M. A. Richards(submitted manuscript, 1993) find in have similar major elementcompositions). The lower degreeof the central platform was apparently produced when that melting is consistentwith the apparentlower excesstemperature lithospherewas farther to the west. Relatively smaller degree estimatedfor the Galfipagos(215øC; [Schilling,1991]) than for melts eruptedon the central platform (i.e., on SantaCruz) would Hawaii (300øC; [Wyllie, 1988]), but these lower temperatures resultin comparativelyminor additionalcrustal thickening. shouldresult in shallower melting than in Hawaii, which is not On the whole,Galfipagos lavas have similar Na20, lowerSiO2, consistentwith the lower SiO2 andZFeO of the Galfipagos. and slightly higher FeO than do MORB at similar MgO The inferred variationsin depth and degreeof melting are WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THEGALfi, PAGOS 19,559 broadlywithhighest consistentmantle withtemperatures thesheared above plume the modelascending outlined limb above,of the I FERNANdiNA SANTiAgO plumeinthe west, and lower temperatures tothe south, east, and Iw V.DARWIN SANTACRUZ north.Superimposed onthis pattern appears tobe a generaltrend i •..-:• •::..:..--:-••.... •.,..•.. •..•.•...... of decreasingdegree and depth of melting southward. This stronglysuggests atendency toward cooler mantle temperatures to /LI/MU51JMtI•t i.' / / / thesouth, away from the spreading center. That mantle tempera- • • '• '• '•'• '• '• •' • • •': turesshould decrease away from the spreadingcenter is, of course, quitereasonable. We werenevertheless surprised that this cooling is observableeven in the presenceof the Galfipagosplume, which

Thisimplies that the asthenosphere has a strongthermal influence onhasthean plume,excess temperatureapparently drawingof215øC heat accordingfrom it. toThe Schillingstrong thermal[1991]. I a interactionbetween plume and asthenosphereis consistentwith the entrainmentof thermally buoyantasthenosphere by sheared plumesas proposed by Richardsand Griffiths [ 1989]. HeterogeneityWithinthePlume In the discussion above, we assumed that the mantle source of Galfipagosmagmas is a mixtureof two homogenouscomponents: the plumeand depletedasthenosphere. The isotopedata, however, Roca v. Wolf Alcedo suggesta subtlecompositional difference between the north and N Redonda v.Darwin south limbs of the plume. For example, althoughFloreana and Pintahave similar •l'qd values, Floreana has higher 87Sr/86Sr ....:..e• .•• ...•...... :• ß" ." ." ß ." ." ." ." ." T." ." ."'•1 •'." ." (Figures6 and 14). Lavasfrom SanCristobal in the southdefine % % % % % % % % % •% % % %1% % % % % % % % % % %]% % % %1% % the high878r/86Sr side of the 878r/86Sr-•NdGalfipagos array, ':--:-:-::::'!-'-.--:: .... -••---".•':•'•' ._..• % % while the northern volcanos, such as Marchena, Wolf Island, ::::::::::'::':.--i-:i.::..•-::::::-:::::i:iiiiiii;.:'i!i:::-iii....'"'::"-'-'--'-:'•:-•:i:::i.'•..--..:•...... •l Darwin Island, Volcan Wolf, and Volcan Darwin, define the low -:iii:-'- .... ::i1' iiii:-:.i::-' -:i::.::-i'.-. .... i!::i:;:i::.] "":•"-•'•:•'•-:•!:i' I ::ii:i 87Sr/86Srside of the array. Thesenorth-south differences are ...... :.... illustrated in Figure 18. The southernIsabela volcanos,Cerro Azul and SierraNegra, plot betweenthe Floreanaand Pinta fields. A somewhatsimilar north-southvariation is apparentin the •Nd- 206pbf204pbarray (Figure 8). In Figure7, thenorthern islands of Pinta,Wolf, and Darwin have distinctly higher 2ø7pb/2ø4pb and 2øspb/2ø4pbfora given2ø6pb/2ø4pb thanthe remaining islands. Thesenorth-south differences are mostapparent in lavas with the mostenriched signatures and are orthogonalto the main arrays in isotoperatio diagrams(Figure 18). They are, therefore,not readily explainedby plume-asthenospheremixing. We conclude ...... i i i i i' ii i ....:" i .....:...... i ii • that thesedifferences reflect lateral heterogeneityin the plume itself, perhapsintrinsic to the deepmantle source of the plume. In Fig. 16. Cartoonillustrating the shearedplume model. Stippledpattern this respect,the Galfipagosseem volcanos some what analogousto representslithosphere, cross-hatched pattern is originalplume material, grayed patternedis asthenosphere,darker gray is thermallybuoyant Hawaiian ones. The emergentvolcanos of the Hawaiian chain asthenosphere.(a) East-westcross section beneath the centerof the appearto be arrangedin two chains,"Loa" and "Kea". While ar- archipelago;(b) north-southcross section at thelongitude of Isabela. rangement of volcanos in these chains appears to reflect lithospheric stresses[Moore, 1987], systematic differences in isotopic composition [e.g., Stille et al., 1986] suggestlateral A relatedfeature is a decreaseof minimumMg# for a volcano heterogeneityin the Hawaiian plume. towardthe centerof the archipelagoand the restrictionof the most differentiatedlavas to the west central volcanos. Since Mg# is effectivelyinsensitive to degreeof melting,this distributionmust Major ElementPatterns and CrustalProcesses reflect shallow level fractional crystallization. The most dif- Figure 17 shows that some volcanos (Marchena and ferentiatedlavas occur near the centerof the plume and overlap Fernandina,for example)appear to eruptproducts of a relatively the area where Na8.0 indicatesthat the degree of melting is uniformdegree of melting,while others(Santa Cruz and Floreana, highest. for example)erupt basaltsproduced by a wide rangeof melting. We believe this distribution reflects a combination of several The apparentlyuniform extentsof melting in the former group, factors: magma supply rates (which in turn determinewhether which also includes Pinta, Genovesa,and the large volcanosof permanentmagma chambers can be maintained),a tendencyfor Isabela,could reflect eitheruniform melting or homogenizationof large central volcanos to evolve toward production of diverse melting products in central magma chambers. The differentiated lavas as they age and magma supply wanes, and volcanosof this groupalso have, for the mostpart, relatively uni- relatively old and cool lithosphere. We argue that eruption of form major element and isotopiccompositions and Mg#. This differentiatedlavas can occuronly where magma'supplyrates are suggeststo us that magma chambersbeneath these islands are high enoughto sustaina permanentcrustal magma chamber. The servingto homogenizemelts passing through them. observationof Sinton and Detrick [1992] that magmasfrom the 19,560 WH1TEET AL.: PETROLOGY ANDGEOCHEMISTRY OFTHE GAL•tPAGOS

10

9 - ...... x

8

œNd 7 '::•::••i•i•::•i•!•!•isi•::•i!!•!•i• )/.•,,d:i::•i::::!i!::ii:: :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :::::::::::::::::::::::::::::::::::::::::::::::.-;."j•::.';..'i::! ....•::•:•;•j/..'/" '::•::.•:.-. ::•:••:,•:•::.'• •:::::::: i••,•::•:::: :::::::::::::::::::::::: i::•::::••::•'•'•'"'"i•i:•i•'•'•'-': :;..,-':/"•4!I:• '•:!'•:•'•!!!J!!:;•:;ii!:i•i'•i:i•:!•,i•! !:i!•,!•,'•'•('•ii'•ii'•i! i:•:•::::•.1%

6 ...... •...... ::::::::::::::::::::::::: .,.•.:•:•,•:••!•!•::?:ii?:?:!i?:?:?:i:i::?:i•:?:i?•i::!::!•i::i::!i::•i::•i:•ii•i•?:i•ii!ii•::iiii::ii•::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

5

.... I , , , , I , , , • I , , , , I .... I ..... , , , I .... 0 0.5 1 1.5 2 2.5 3 3.5 4 [La/Sm]nao

• • ...... !i:•:::•:;•':.•-..•i•:::::::.:...... ::::• ...... :::•.•:•:/.•i•::-::;'•::::::i•.-i!•:!i• ...... i!i•:i:i:!:::•!:!::::::"' '...... '-::i:'-....:-.•'...... t" .. J2% ' 0 ' E ME•T NC..• • • :::::::::::::::::::::::::::::::::::::======;=':'?•:"-'"•::::•.•:..-'" /- Spl• ß • IRid•::•:::::•:::•::•::•::•%•:.•::•::•;•.:•::: ...... •:•:•:•::•:• :::::::::::::::::::::::::::::::::::

• : .0:• ¾:•::::':':::-'.'•::•::::•:: x ':.'-" ...... •::::•::";:•:••:•::•'• ...... • "::::::•?:•?::?.•;•:?:•. ..::...... :.. . :- ..•::•:,:•.•:.::•5•:•:•...... 1.5 • :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::-...'"--•...... ::•::::-..: ...... :•--m.---:-::•..::...... •...... ::•.:-....•;:•...... :•;• ...... •• ......

0.5I I I b, 0 0.5 1 1.5 2 2.5 3 3.5 4 lLa/Sm]nao

Fig. 17.(a) ENdand (b) [SndYb]N8.0plotted against [La/Sm]N8.0. Heavy solid line is a hypotheticalplume-asthenosphere mixingcurve. The "depletedasthenosphere" end-member has ENd = 9.8, La/SmN = 0.35, SndYbN=1; the "enriched plume"end-member has ENd = 4.85,LaJSm N = 1.7,SrrdYb• = 2.2,Nd concentration3.45 times higher than in thedepleted source. Finer linesillustrate equilibrium partial melting of this mixture. Fine dashedlines represent constant melt fractions(1%, 2.5%, 5%, and 10%). In Figure17b the effects of meltingin thegarnet and spinel peridotite stability fields arecompared. Garnet strongly fractionates Sm fromYb andmelting of garnetperidotite consequently produces a much higherSm/Yb ratio in themelt. Alsoshown are the calculated paths for threeexamples of polybaricprogressive melting (dashedcurves labeled A, B, and C). In eachcase, 1% meltingincrements are produced,extracted, and pooled. The first 6%, 2%, and 1% of meltingoccur in thegarnet stability field for curvesA, B, andC, respectively,with subsequentmelting occurringin the spinelstability field. The curvesshow how the composition of thepooled melt changes with increasing total degreeof melting. fast spreadingEast PacificRise are, on average,more differen- the largewestern volcanos may eventuallyreach this stage,but at tiated than those of the slow spreadingMid-Atlantic Ridge present,only Alcedo,which has erupted a bimodalrhyolite-basalt supportsthis hypothesis.If magmasupply rate is low, magma suite over the past 100 kyr or so, is currentlyin it. Santiago remainingin the crust after an eruptiveepisode crystallizes in apparentlyalso passed through such a phasebut now appears to be placebefore the new magmais injected. At higherrates of pro- experiencinga rejuvenation.The uniformcompositions of Pinta, duction,large poolsof magmaremain in the crust,cooling and Marchena,and Genovesasuggest that thesevolcanos are also un- crystallizing,and mix with the nextbatch of magmaarriving from derlain by permanentmagma chambers,though they have not the mantle. When magmasupply eventually wanes, this mixing eruptedhighly differentiatedlavas. Either they have not yet ratioincreasingly favors the remainingdifferentiated melt. Rfibida reachedthis phase of theirevolution, or the young,warm nature of and Pinz6n seemto havepassed through such a stage. Severalof the lithospherein this area limits cooling and fractionalcrys- WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OF THEGAL/[PAGOS 19,561

of fractional crystallization, reflect variations in both source compositionand degreeand depthof melting. Both major and trace element abundancessuggest that the greatest degree of melting occurs beneath the central western volcanos. Degree of melting decreaseseastward and to the north and south. This pattern suggeststhat interaction between the plume and the surroundingasthenosphere results in significant œNd7 cooling of the plume. Superimposedon this thermal pattern producedby plume-asthenosphereinteraction is a tendencyfor meltingto be lessextensive and to occurat shallowerdepths to the south,presumably reflecting a decreasein ambientasthenospheric temperature away from the GSC. Magmas erupted on the southernislands are producedprimarily by melting in the spinel

4 stability field, while those erupted on the central and northern 0.7025 0.704 islandsare producedprimarily in the garnet stability field. The greatestdepth of melting may occurbeneath the centralislands of SantaCruz and Santiago. Fig. 18. ENdplotted against 87Sr/86Sr to illustratehow bothplume-as- Overall, our resultssuggest that the Galfipagosmantle plume thenospheremixing and north-southheterogeneity within the plume appearto be requiredto explainthe total isotopicvariation in Galfipagos interacts strongly with the asthenosphere. This extensive lavas. Data from Genovesa,Marchena, Pinta, Wolf Island, Darwin Island, interactionappears to be uniqueamong oceanic island groups. We Santiago,Santa Cruz, Santa Fe, Volcan Ecuador,Volcan Wolf, Volcan suspectthat this reflects the special tectonic setting of the Darwin, Fernandina,Pinzon, and Rfibida (the "Northern Volcanos") are Galfipagos. The plume emerges beneath young lithosphere plottedas opencircles; data from the southernvolcanos (Volcan Alcedo, Cerro Azul, Sierra Negra, Floreana, Espafiola,and San Cristobal) are adjacentto a moderatelyfast spreadingridge, but absoluteplate, plottedas opentriangles. The southernvolcanos generally have higher and presumablyshallow asthenospheric, motion is almostparallel 87Sr/86Srfor a givenENd. to the ridge. More work on the dynamicand thermalinteraction of plumes with the asthenosphereis needed before a full understandingof the Galfipagoscan be achieved. tallization. M. A. Feighner and M. A. Richards (submitted Acknowledgments.We thankMike Cheathamand ChristineMcBirney manuscript, 1993) noted that the occurrence of highly for performingmany of the majorand trace element analyses reported here differentiatedlavas seemsrestricted to the regionof thickestcrust. and Mike Cheathamand Karen Harpp for proofreadingthe manuscript. Since thick crust requires high magma supply rates, this The work benefitedimmensely from discussionswith our colleaguesand observationis consistentwith the explanationwe proposehere. coconspiratorsDave Christie, Mark Richards, Dennis Geist, Mark Feighner,Ross Griffiths, and Karen Harpp. We also thank Denny Geist Alternatively, thick crust may result in magma ponding at and Tom Simkin for providingadditional samples and for accessto their relatively great depth (due to lack of density contrast),where it unpublisheddata and the Darwin ResearchStation for assistancewith field can cool and crystallize slowly until its density decreasessuffi- work. Thanks as well to Stan Hart, Denny Geist, and Karen Harpp for ciently that it can rise to the surface. their helpful commentson an early versionof this manuscript. Formal reviewsby JamieAllen, StanHart, Emily Klein, and Tom Simkinresulted in substantialimprovements to the manuscript.This work was supported CONCLUSIONS by NSF grants INT-8315432, EAR85-07973, and OCE89-15402 to A.R.M.; EAR85-20704 to W.M.W. and RAD; OCE89-17031 to W.M.W.; The mostimportant observation made here is that isotopeand and OCE89-11286 to R.A.D. incompatibleelement ratios define a horseshoepattern with the most depleted signatures in the center of the Galfipagos REFERENCES Archipelago and the more enrichedsignatures on the eastern, Bailey, K., Potassium-argonages from the GalfipagosIslands, Science, northern, and southernperiphery. We suggestthat this reflects 192, 465-467, 1976. thermalentrainment of asthenosphereby a mantleplume undergo- Baitis, H., , petrography,and petrologyof Pinz6n and Santiago ing velocity shear. Alternativehypotheses, such as progressive islands, GalfipagosArchipelago, Ph.D. dissertation,Univ. of Oreg., meltingof a veinedmantle or remeltingof lithospherebeneath the Eugene,1976. Baitis, H. W., and M. Lindstrom,Geology, petrography, and petrologyof centerof the archipelago,predict relationships between major and Pinz6n Island, GalfipagosArchipelago, Contrib. Mineral. Petrol., 72, trace elementsand isotoperatios that are inconsistentwith those 367-386, 1980. actually observed. Baitis, H. W., and F. J. Swanson, Ocean rise-like basalts within the Superimposedon the source variations producedby en- GalfipagosArchipelago, Nature, 259, 195-197, 1976. Batiza, R., and D. Vanko, Petrology of young Pacific seamounts,J. trainment of asthenosphereare both heterogeneitywithin the Geophys.Res., 89, 11,235-11,260,1984. plume itself and regional variations in degree and depth of Bow, C. S., Geology and Petrogenesisof lavas from Floreanaand Santa melting. The plume appearsto be laterally heterogeneousin a Cruz islands,Galfipagos Archipelago, Ph.D. dissertation,308 pp., Univ. north-south direction, with the southern limb having higher of Oreg., Eugene,1979. 2ø6pb/2ø4pband87Sr/86Sr and possibly La/Sm for a givenœNd' Bow, C. S., and D. J. Geist, Geology and petrologyof FloreanaIsland, GalfipagosArchipelago, Ecuador, J. Volcanol. Geotherm.Res., 52, 83- The lack of correlationbetween isotope ratios and Na8. o, Si8.o, and 105, 1992. Fe8.0 indicatesthat neitherdegree nor depthof meltingaffect the Brown, J., A. Colling, D. Park, J. Phillips, D. Rothery,and J. Wright, isotope ratios of Galfipagos magmas. It also indicates that Seawater: Its Composition, Properties and Behavior, 165 pp., whatever differencesin Na20, SiO2, and FeO there might be Pergamon,New York, 1989. Cassignol,C., and P.-Y. Gillot, Range and effectivenessof unspiked betweenplume and asthenosphere,they are very muchsecondary potassium-argondating: Experimental groundwork and applications,in to the effects of degree and depth of melting on the latter. NumericalDating and Stratigraphy,edited by G. S. Odin, pp. 159-179, Incompatible trace element ratios, after correction for the effects JohnWiley and Sons,New York, 1982. 19,562 WHITEET AL.: PETROLOGYAND GEOCHEMISTRY OFTHE GAL•PAGOS

Chadwick,W. W., and K. A. Howard, The patternof circumferentialand Soc. Aust. Abtr., 27, 44, 1990. radial eruptive fissures on the volcanos of Fernandinaand Isabela Hart, S. R., and C. Brooks, Clinopyroxene-matrixpartitioning of K, Rb, islands,Galfipagos, Bull. Volcanol.,53, 259-275, 1991. Cs, Sr, and Ba, Geochim. Cosmochim.Acta, 38, 1799-1806, 1974. Chadwick,W. W. J., T. De Roy, andA. Carrasco,The September1988 in- Hey, R., Tectonic evolution of the Cocos-Nazcaspreading center, Geol. tracalderaavalanche and eruption at Fernandinavolcano, Galfipagos Is., Soc. Am. Bull., 88, 1404-1420, 1977. Bull. Volcanol., 53, 276-286, 1991. Jaques,A. L., and D. H. Green, Anhydrousmelting of peridotiteat 0-15 Cheatham,M. M., W.F. Sangrey,and W. M. White, Sourcesof error in Kb pressureand the genesis of tholeiitic basalts,Contrib. Mineral. external calibration ICP-MS analysisof geologicalsamples and an Petrol., 73, 287-310, 1980. improvednon-linear drift correction,Spectrochim. Acta Electron.,48B, Klein, E. M., and C. H. Langmuir, Ocean ridge basalt chemistry,axial E487-E506, 1993. depth, crustal thicknessand temperaturevariations in the mantle, J. Christie, D. M., R. A. Duncan, A. R. McBirney, M. A. Richards,W. M. Geophys.Res., 92, 8089-8115, 1987. White, K. S. Harpp, and C. G. Fox, Drownedislands downstream from Klein, E. M., and C. H. Langmuir,Local versusglobal variations in ocean the Galfipagoshotspot imply extendedspeciation times, Nature, 355, ridge basalt composition:A reply, J. Geophys.Res., 94, 4241-4252, 246-248, 1992. 1989. Colinvaux, P. A., The Galapagosclimate: presentand past, in Key Levi, S., H. Audunsson,R.A. Duncan, L. Kristjansson,P.-Y. Gillot, and Environments:Galapagos, edited by C. Perry,p. 15-41,Pergamon, New S.K. Jakobsson,Late Pleistocene geomagneticexcursion in Iceland York, 1984. lavas:Confirmation of the Laschampexcursion, Earth Planet.Sci. Lett., Cox, A. and G.B. Dalrymple, Paleomagnetismand potassium-argonages 96, 443-457, 1990. of somevolcanic rocks from the GalfipagosIslands, Nature, 209, 776- Lindstrom, M. M., Geochemical studies of volcanic rocks from Pinz6n 777, 1966. and Santiagoislands, Galfipagos Archipelago, Ph.D. disseration,174 pp., Cullen, A. B., Geology and Petrologyof Isla Pinta, Galfipagos,M.S. Univ. of Oreg,Eugene, 1976. thesis,77 pp., Univ. of Oreg.,Eugene, 1985. Macdonald,G. A., and T. Katsura,Chemical composition of Hawaiian Cullen, A., and A. R. McBirney, The volcanicgeology and petrologyof lavas, J. Petrol., 5, 82-133, 1964. Isla Pinta, GalfipagosArchipelago, Geol. Soc.Am. Bull., 98, 294-301, Malfait, B. T., and M. G. Dinkerman,Circum-Pacific tectonic and igneous 1987. activity and the evolutionof the Caribbeanplate, Geol. Soc.Am. Bull., Cullen, A. B., A. R. McBirney, and R. D. Rogers,Structural controls on 83, 251-272, 1972. the morphologyof Galfipagosshields, J. Volcanol.Geotherm. Res., 34, McBirney, A. R., Differentiatedrocks of the Galfipagoshotspot, J. Geol. 143-151, 1987. Soc.London, in press,1993. Cullen, A., E. Vicenzi, and A. R. McBirney, Plagioclase-ultraphyric McBirney, A. R., A. B. Cullen, D. Geist, E. P. Vicenzi, R. A. Duncan,M. basaltsof the GalfipagosArchipelago, J. Volcanol.Geotherm. Res., 37, L. Hall, and M. Estrella, The GalfipagosVolcano Alcedo: A unique 325-337, 1989. ocean caldera, J. Volcanol. Geotherm. Res., 26, 173-177, 1985. Dalymple, G. B., and M. A. Lanphere,Potassium-Argon Dating, W. H. McBirney, A. R., and H. Williams, Geology and Petrology of the Freeman, New York, 1969. GaldpagosIslands, Geol. Soc.Am. Mere. 118, pp. 197, Geol. Soc.Am., Darwin, C., The Voyageof the Beagle, 524 pp., Doubleday,New York, Boulder, 1969. 1860. Minster, J. B., and T. H. Jordan,Present-day plate motions,J. Geophys. Delaney,J. R., W. E. Colony,T. M. Gerlach,and B. E. Nordlie,Geology Res., 83, 5331-5354, 1978. of the Volcan Chico areaon SierraNegra Volcano, Galfipagos Islands, Moore, J. G., Subsidenceof the HawaiianRidge, U.S. Geol. Surv.Prof. Geol. Soc. Am. Bull., 84, 2455-2470, 1973. Pap., 1350, 85-100, 1987. Duncan, R. A., M. T. McCulloch, H. G. Barczus, and D. R. Nelson, Morgan, W. J., Convectionplumes in the lower mantle.,Nature, 230, 42- Plumeversus lithopheric sources for the melts at Ua Pou, Marquesas 43, 1971. Island, Nature, 303, 142-146, 1986. Morgan, W. J., Rodriguez, Darwin, Amsterdam.... A secondtype of Emerman,S. H. and D. L. Turcotte,Stagnation flow with a temperature- hotspotisland, J. Geophys.Res., 83, 5355-5360, 1978. dependentviscosity, J. Fluid Mech., 127, 507-517, 1983. Morris, J. D., and S. R. Hart, Isotopic and incompatibleelement con- Gallahan,W. E., and R. L. Nielsen,The partitioningof Sc, Y andthe rare straintson the genesisof arc volcanicsfrom Cold Bay andAmak Island, earth elements between high-Ca pyroxene and natural mafic to Aleutians,and implicationsfor mantlestructure, Geochim. Cosmochim. intermediatelavas at 1 atmosphere,Geochim. Cosmochim.Acta, 56, Acta, 47, 2015-2030, 1983. 2387-2404, 1992. Nakamura, N., Determination of REE, Ba, Fe, Mg, Na and K in car- Geist,D. J., An appraisalof meltingprocesses and the Galfipagoshotspot: bonaceousand ordinary chondrites,Geochim. Cosmochim.Acta, 38, Major and trace elementevidence, J. Volcanol. Geotherm.Res., 52, 65- 757-775, 1974. 82, 1992. Prestvik, T., and R.A. Duncan, The geology and age of Peter I Oy, Geist,D. J., A. R. McBirney,and R. A. Duncan,Geology and petrogenesis Antarctica, Polar Res., 9, 89-98, 1991. of lavas from San Cristobal,Galfipagos Archipelago, Geol. Soc. Am. Richards,M. A., and R. W. Griffiths, Thermal entrainmentby deflected Bull., 97, 555-566, 1985a. mantleplumes, Nature, 342,900-902, 1989. Geist, D. J., A. R. McBirney, and R. A. Duncan,Geology of SanteFe Rowland, S. K., and D.C. Munro, The caldera of Volcan Fernandina: A Island:the oldestGalfipagos volcano, J. Volcanol.Geotherm. Res., 26, remote sensingstudy of its structureand recentactivity, Bull. Volcanol., 203-212, 1985b. 55, 97-109, 1992. Geist, D. J., W. M. White, and A. R. McBirney, Plume-asthenosphereSalters, V. J. M., and S. R. Hart, The Hf-paradox,and the role of garnetin mixing beneaththe GalfipagosArchipelago., Nature, 333, 657-660, the MORB source,Nature, 342, 420-422, 1989. 1988. Schilling,J.-G., Icelandicmantle plume: Geochemical evidence along the Gillot, P.-Y. and P. Nativel, Eruptive history of the Piton de la Fournaise ReykjanesRidge., Nature, 242, 565-571, 1973. volcano, Reunion Island, Indian Ocean, J. Volcanol. Geotherm. Res., 36, Schilling,J.-G., Fluxes and exces• temperatures of mantle plumes inferred 53-66, 1989. from their interactionwith migrating mid-oceanridges, Nature, 352, Griffiths, R. W., The differing effects of compositionaland thermal 397-403, 1991. buoyancieson the evolution of mantle diapirs, Phys. Earth Planet. Schilling, J.-G., R. N. Anderson,and P. Vogt, Rare earth, Fe and Ti Inter., 43, 261-273, 1986. variationsalong the Galfipagosspreading centre, and their relationship to Gripp, A. E., and R. G. Gordon, Current plate velocities relative to the Galfipagosmantle plume, Nature, 261,108-113, 1976. hotspotsincorporating the NUVEL-1 globalplate model, Geophys. Res. Schilling,J.-G., R. H. Kingsley,and J. D. Devine, Galfipagoshot spot- Lett., 17, 1109-1112, 1990. spreading center system, 1, Spatial petrological and geochemical Hall, M. L., Origin of EspafiolaIsland and the age of terrestriallife on the variations(83øW-101øW), J. Geophys.Res., 87, 5593-5610, 1982. GalfipagosIslands, Science, 221,545-547, 1983. Shimizu,H., A. Masuda,and N. Masui, Rare-earthelement geochemistry Hanson, G. N., Geochemical evolution of the suboceanicmantle, J. Geol. of volcanicrocks from the GalfipagosIslands, Geochemical J., 15, 81- Soc. London, 134, 1-19, 1977. 93, 1981. Hanson,G. N., and C. H. Langmuir, Modelling of major elementsin Shimizu, N., An isotopedilution techniquefor analysisof the rare earth mantle-melt systems using trace element approaches., Geochim. elements,Year Book CarnegieInst. Washington,73, 941-943, 1974. Cosmochim. Acta, 42, 725-741, 1978. Simkin, T., Geology of the GalapagosIslands, in Key Environments: Harpp, K. S., W. M. White, and R. A. Duncan,Geochemistry of the Wolf- Galapagos,edited by R. Perry,pp. 15-41,Pergamon, New York, 1984. Darwin Lineamentand ridge-plumeinteraction in the Galfipagos,Geol. Simkin,T., and K. A. Howard,Caldera collapse in the GalfipagosIslands, WHITE ET AL.: PETROLOGYAND GEOCHEMISTRYOF THE GAL/•PAGOS 19,563

1968, Science, 169, 429-437, 1970. Res., 91, 5927-5941, 1986. Simkin,T., L. Siebert,L. McClelland,D. Bridge,C. Newhall,and J. H. White, W. M., and A. W. Hofmann, Geochemistryof the Galfipagos Latter, Volcanos of the World, 232 pp., Smithsonian Institution, Islands: Implicationsfor mantle dynamicsand evolution., Year Book Washington,D.C., 1981. CarnegieInst. Washington,77, 596-606, 1978. Sinton, J. M., and R. Detrick, Mid-oceanridge magmachambers, J. White, W. M., and A. W. Hofmann, Sr and Nd isotopegeochemistry of Geophys.Res., 97, 197-216, 1992. oceanicbasalts and mantlegeochemistry, Nature, 296, 821-825, 1982. Stille, P., D. M. Unruh, and M. Tatsumoto,Pb, Sr, Nd, and Hf isotopic White, W. M., and P. J. Patchett, Hf-Nd-Sr and incompatible-element constraintson the originof Hawaiianbasalts and evidence for a unique abundancesin island arcs: Implicationsfor magma origins and crust- mantle source,Geochim. Cosmochim. Acta, 50, 2303-2320, 1986. mantle evolution, Earth Planet. Sci. Lett., 67, 167-185, 1984. Swanson,F. J., H. W. Baitis, J. Lexa, and J. Dymond,Geology of Whitford, D. J., W. M. White, A. W. Hofmann, and D. E. James, Santiago,Rfibida, and Pinz6nIslands, Galfipagos, Geol. Soc. Am. Bull., Separationand isotopicanalysis of neodymium,Year Book Carnegie 85, 1803-1810, 1974. Inst. Washington,77, 620-623, 1978. van Andel, T. H., G. R. Heath, B. T. Malfait, D. F. Heinrichs,and J. I. Wright, T. L., and R. S. Fiske, Origin of the differentiatedand hybrid Ewing, Tectonicsof the PanamaBasin, eastern equatorial Pacific, Geol. lavas of Kilauea Volcano, Hawaii, J. Petrol., 12, 1-65, 1971. Soc. Am. Bull., 82, 1482-1504, 1971. Wyllie, P. J., Solidus curves, mantle plumes, and magma generation vanAndel, T., et al., Initial Reportsof theDeep Sea Drilling Project,949 beneathHawaii, J. Geophys.Res., 93, 4171-4181, 1988. pp., U.S. GovermentPrinting Office, Washington, D.C., 1973. Yoder, H. S., and C. E. Tilley, Origin of basaltmagmas: an experimental Verma, S. P., and J.-G. Schilling,Galfipagos hot spot-spreadingcenter study of natural and syntheticrock systems.,J. Petrol., 3, 342-532, system,2, 87Sr/86Srand large ion lithophileelement variations (85øW - 1962. 101øW),J. Geophys.Res., 87, 10838-10856, 1982. Zindler,A., S. R. Hart, F. A. Frey, and S. P. Jakobsson,Nd and Sm isotope Verma,S. P., J.-G. Schilling,and D. G. Waggoner,Neodymium isotopic ratiosand rare earthelement abundances in ReykjanesPeninsula basalts; evidencefor Galfipagoshotspot-spreading centre systemevolution, Evidence for mantle heterogenietybeneath Iceland, Earth Planet. Sci. Nature, 306, 654-657, 1983. Lett., 45, 249-262, 1979. Vicenzi, E., The geology,petrology, and petrogenesisof Isla Marchena, Zindler, A., H. Staudigel, and R. Batiza, Isotope and trace element GalfipagosArchipelago, M.S. thesis,96 pp., Univ. of Oreg, Eugene, geochemistryof young Pacific seamounts:Implications for the scaleof 1985. uppermantle heterogeneity, Earth Planet. Sci. Lett., 70, 175-195, 1984. Vicenzi, E. P., A. R. McBirney, W. M. White, and M. Hamilton, The geologyand geochemistry of IslaMarchena, Galfipagos Archipelago: An oceanisland adjacent to a mid-oceanridge, J. Volcanol.Geotherm. Res., R. A. Duncan, College of Oceanography,Oregon State University, 40, 291-315, 1990. Corvallis, OR 97331. Vogt, P. R., Volcano heightand plate thickness,Earth Planet. Sci. Lett., A. R. McBirney, Center for Volcanology, University of Oregon, 23, 337-348, 1974. Eugene, OR 97403. Walker, R. J., R. W. Carlson,S. B. Shirey,and F. R. Boyd, Os, Sr, Nd, W. M. White, Dept. of Geological Sciences, Cornell University, and Pb isotopesystematics of southernAfrican peridotitcxenoliths: Ithaca, NY 14853. Implications for the chemical evolution of the subcontinentalmantle, Geochim. Cosmochim.Acta, 53, 1583-1595, 1989. (ReceivedAugust 3, 1992; White,W. M., andB. Dupr6,Sediment subduction and magma genesis in revisedJuly 15, 1993; the LesserAntilles: Isotopic and trace element constraints, J. Geophys. acceptedJuly 21, 1993.)