Tahiti: Geochemical Evolution of a French Polynesian Volcano

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 99, NO. B12, PAGES 24,341-24,357, DECEMBER 10, 1994

Tahiti: Geochemical evolution of a French Polynesian volcano

Robert A. Duncan and Martin R. Fisk Collegeof Oceanicand AtmosphericSciences, Oregon State University, Corvallis

William M. White Departmentof GeologicalSciences, Cornell University, Ithaca, New York

Roger L. Nielsen Collegeof Oceanicand Atmospheric Sciences, Oregon State University, Corvallis

Abstract. The island of Tahiti, the largestin French Polynesia,comprises two major volcanoes alignedNW-SE, parallelwith the generaltrend of the SocietyIslands hotspot track. Rocksfrom this volcanicsystem are basaltstransitional to tholeiites,alkali basalts,basanites, picrites, and evolvedlavas. Through K-Ar radiometricdating we haveestablished the ageof volcanicactivity. The oldestlavas (--1.7 Ma) cropout in deeplyeroded valleys in the centerof the NW volcano (Tahiti Nui), while the main exposedshield phase erupted between 1.3 and 0.6 Ma, and a late- stage,valley-filling phase occurred between 0.7 and0.3 Ma. The SW volcano(Tahiti Iti) was active between0.9 and 0.3 Ma. There is a clear changein the compositionof lavas throughtime. The earliestlavas are moderatelyhigh SiO2,evolved basalts (Mg number(Mg# = Mg/Mg+Fe2+) 42-49), probablyderived from parentalliquids of compositiontransitional between those of tholeiitesand alkali basalts.The main shieldlavas are predominantlymore primitive olivine and clinopyroxene-phyricalkali basalts(Mg# 60-64), while the latervalley-filling lavas are basanitic (Mg# 64-68) andcommonly contain peridotitic xenoliths (olivine+orthopyroxene+ clinopyroxene+spinel). Isotopic compositions also change systematically with time to more depletedsignatures. Rare earthelement patterns and incompatible element ratios, however, show no systematicvariation with time.We focusedon a particularlywell exposedsequence of shield- buildinglavas in the PunaruuValley, on the westernside of Tahiti Nui. CombinedK-Ar agesand magnetostratigraphicboundaries allow high-resolutionage assignments to this-0.7-km-thick flow section.We identifiedan early periodof intensevolcanic activity, from 1.3 to 0.9 Ma, followedby a periodof moreintermittent activity, from 0.9 to 0.6 Ma. Flow accumulationrates droppedby a factorof 4 at about0.9 Ma. This changein rateof magmasupply corresponds to a shiftin activityto Tahiti Iti. We calculatedthe compositionof the parentmagma for the shield- buildingstage of volcanism,assuming that it was in equilibriumwith Fo89olivine and that the mostprimitive aphyric lavas were derived from thisparent by the crystallizationof olivine alone. The majorityof the shieldlavas represent 25 to 50% crystallizationof thisparent magma, but the mostevolved lavas represent about 70% crystallization.From over 50 analyzedflow unitswe recognizea quasi-periodicevolution of lava compositionswithin the early,robust period of volcanicactivity, which we interpretas regular recharge of themagma chamber (approximately every25 + 10 kyr). Volcanicevolution on Tahiti is similarto the classicHawaiian pattern. As the shield-buildingstage waned, the lavasbecame more silica undersaturated and isotopic ratios of the lavasbecame more MORB-like. We proposethat the Societyplume is radially zoneddue to entrainmentof a sheathof viscouslycoupled, depleted mantle surrounding a centralcore of deeper mantlematerial. All partsof the risingplume melt, but the thermaland compositional radial gradientensures that greaterproportions of meltingoccur over the plume center than its margins. The changingcomposition of Tahitianmagmas results from lithosphericmotion over this zoned plume.Magmas erupted during the main shield-buildingstage are derived mainly from thehot, incompatibleelement-enriched central zone of theplume; late-stage magmas are derived from the cooler,incompatible element-depleted, viscously coupled sheath. A correlationbetween Pb/Ce and isotoperatios suggests that the Societyplume contains deeply recycled continental material.

Introduction southcentral region known as French Polynesia.Here, several separate, subparallel island chains strike west-northwest, Linear chains of volcanic islands and seamounts are a aligned in the direction of Pacific plate motion. Young to conspicuousfeature of the Pacific Oceanfloor, especiallyits active volcanism occurs at the southeast end of each of these groups. It is now generally accepted that such volcanic Copyright 1994 by the AmericanGeophysical Union. lineaments are the consequenceof plate motion over long- Paper number 94JB00991. lived hotspots, which are proposed to be sublithospheric 0148-0227/94/94JB-00991 $05.00 thermal anomaliesmaintained by rising convectiveplumes of deeper mantle material [Morgan, 1971, 1972].

24,341 24,342 DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAHITI

Hotspot-related oceanic islands are also significant as radiometric dating has shown that island ages become geochemical records of mantle-lithosphereinteraction, mantle progressively older to the northwest, at a rate of 10 to 11 melting and magma transportprocesses, and the temporal and cm/yr [Duncan and McDougall, 1976]. The orientationof the spatial scales of mantle heterogeneity.In contrastto the first- chain and the age distribution of the volcanism are compatible order uniformity of compositionsof mid-ocean ridge basalts with Pacific plate motion over a stationary hotspot [Duncan (MORBs), ocean island basalts (OIBs) show tremendous and Clague, 1985]. variability. The compositionalrange of volcanic productsin Tahiti, the largest of these islands, lies near the southeast French Polynesia reflects melt contributions from several end of the chain. This island comprisestwo major coalesced mantle reservoirs that have remained isolated from one another volcanoes,aligned NW-SE, parallel with the general trend of for periods appropriate to mixing times for whole mantle the Society hotspot track. Volcanic activity here ceased at convection [Gurnis and Davies, 1986]. The origin of mantle about0.2 Ma [Dymond, 1975; Duncan and McDougall, 1976; heterogeneities may lie in melting and fractionation Becker et al., 1974; Gillot et al., 1993]. The present-day processes, recycling of lithosphere, metasomatism, and position of the hotspot southeastof Tahiti is indicated by the mantle-lithosphere assimilation. However, understandingthe young island of Mehetia (0.0 to 0.3 Ma [Binard et al., 1993; nature and significanceof distinct sourcesfor melt generation R. Duncan, unpublisheddata, 1993]), intense seismicactivity requires resolution of the volcanic histories of individual [Talandier and Okal, 1984], and active submarine volcanoes oceanic islands. [Chemineeet al., 1989; Devey et al., 1990]. This paper describes the volcanic history of Tahiti, the Tahiti is roughly 30 km wide and 55 km along the axis of largest island in the Society Islands chain. We begin with the two volcanoes and has a total volume of about 6 x 104 km3, radiometric dating to establish the time frame of volcanic of which about 2% lies above sea level (Figure 1). The activity. Next, we present whole rock and mineral major and volcanoesare composedpredominantly of basaltic lava flows, trace element abundancesfrom which we infer parent magma but pyroclastic rocks related to localized phreatic eruptions compositions.Isotopic data are used to evaluate mantle source occur in coastal outcrops,and subordinatedifferentiated rocks variability throughout the constructionof Tahiti. Finally, we (hawaiite-mugearite-trachyte-phonoliteseries) are common as assess fractionation, recharge, and assimilation effects in a plugs, sills, and irregular flows [DeNeufbourg, 1965]. 700-m-thick section in the Punaruu Valley. We combine Remnants of the original shield ("planeze") are separatedby compositional variations with age determinations to estimate deeply incised, radial valleys that provide spectacular recharge and fractionation rates. topographic relief (-2 km) and offer extraordinary accessto the volcanic stratigraphy. Gillot et al. [1993] reported east- west trending rift zones that cut the shield and evidence of Geological Setting and Previous Work caldera collapse at the main eruptive centers. Posterosional eruptions from these centers have partially filled old river The Society Islands chain is a -700-km lineament valleys ("remplissage des vallees"). Tahiti is remarkable in composedof islands, atolls, and seamounts,all of which are displaying the most extensive suite of plutonic rocks among predominantly volcanic. Individual volcanic centers rise Pacific islands. These are found in the eroded centers of both separatelyfrom abyssaldepths of 3.5 to 4 km below sea level volcanoes, having crystallized some 2 to 3 km below the and were erupted onto -60 Ma oceanic lithosphere.Previous original summitsof the shields[Williams, 1933].

Tahiti Nui

i Tahiti Iti

ß ..

. ß ß ß ß

10 km "'. ."

ß ,

Figure 1. Map of Tahiti showingmajor featuresmentioned in the text. Heavy shadingindicates areas of plutonic rocks. Light shadingis the area of oldest central lavas. Dotted lines are flinging reefs. Sample locations(latitude, longitude,and elevation)are listed in Table A1. Crossmarks a xenolith locationdiscussed in the text. DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAHITI 24,343

Early geologic and petrologic studies [Lacroix, 1910; (sample numbers other than T73-, T74-, and T85-) are those Marshall, 1915; Williams, 1933; DeNeufbourg, 1965; describedby Chauvin et al. [1990]. McBirney and Aoki, 1968] established that most Tahitian The thick sequenceof seawarddipping lava flows exposedin rocks are moderately evolved alkali basalts and low-pressure the Punaruu Valley (Figure 2) was erupted during the main fractional crystallization derivatives from them. The late- phase of shield construction. Rapid erosion during periodic stage, valley-filling lavas are relatively primitive basanitic storms and active quarrying have provided fresh outcrop rocks. An important aspect of these late-stage flows is that surfaces. Flow orientations and thicknesses were recorded to they commonly contain ultramafic xenoliths of dunite, construct measured sections; dips are typically 8 to 10ø, and harzburgite, wehrlite, and spinel lherzolite [Duncan, 1975; flows are 1 to 3 m thick. Individual flows can be traced for Tracy, 1980]. The plutonic rocks are chemical equivalentsof hundredsof meters along the canyon walls. Although there are alkali basalts and basanites,but most appear to be related to several discordancesin the section due to infilling of erosional the posterosionalvolcanic phase.Lava flows from the deepest surfaces,these are relatively minor, and the grossstructure is a stratigraphic level exposed on the island are hypersthene- monotonic sequence of flows (Figure 2). The 700-m-thick normative, high-silica basalts, transitional between alkali section is nearly entirely lava flows with rubbly tops and basalt and tholeiitic compositions [Duncan et al., 1987]. bottoms and blocky to columnar centers. Dikes are rare, Given the predominance of tholeiitic lavas in the shield- indicating that the eruption center was distant. Valley-filling building stage of Hawaiian Islands [Macdonald and Katsura, lava flows cover the inland extremity of this section. 1964] and the discovery of tholeiitic compositions in early Paleomagneticdirections were determinedfrom orientedcores lavas from the Marquesas[Duncan et al., 1986; Desonie et al., after the cores had been subjected to standard magnetic 1993] and Samoan Islands [Natland and Turner, 1985], it seems cleaningmethods [Chauvin et al., 1990]. possible that submarine portions of the Society islands are also tholeiitic. However, Devey et al. [1990] found no tholeiitic basalts among dredged samples from the young seamounts southeast of Tahiti, so the nature of early Radiometric Age Determinations volcanism at Society volcanoesremains enigmatic. Timing for Tahitian Volcanism Sampling and Field Studies Shield lavas, valley-filling lavas, and differentiated rocks Most of the samples used in this study (prefix T85-) were were selected for K-At age determinationsafter petrographic collected in 1985 during a joint U.S.-French program of field examination for degree of alteration and crystallinity. Rocks measurementsand sampling. Seven river canyon sectionswere were trimmed to remove any weatheredsurfaces, crushed and examined to distinguish shield and valley-filling sequences. sieved to --0.5 mm size fraction, and ultrasonically washed in We also visited the central region of the island (Figure 1) to distilledwater. A split 0-5 g) was thenpowdered for K analysis examine the plutonic complex and sample what turn out to be by atomic absorptionspectrophotometry. Because xenolithic the oldest lavas on the island. Xenoliths were collected from material contains mantle-derived Ar [Jager et al., 1985], valley-filling lavas in the Fautaua, Punaruu and Papenoo porphyritic sampleswere avoided or hand-pickedto remove valleys. Additional samples(prefixes T73- and T74-) are from olivine and pyroxene.Ar was extractedand purifiedfrom 1 to 5 Duncan's [1975] thesis collection at the Australian National g whole rock samples by radio frequency (RF) induction University and come primarily from the uppermost Punaruu heating in high-vacuum glass lines with Ti getters. The section. Paleomagneticcores from Punaruu Valley lava flows isotopiccomposition of 38At-spikedAr was measuredusing an

0.92

•30km• 0.62 POLARITY • normal reversed • transitional •0•90' I100 m 500 rn

Punaruu River

0.92 0.96 1.o4 1.oo 0.99 1.05 1.11 1.o9 1.Ol 1.o8 1.19 0.96 1.35

Figure 2. Schematiccross section of the PunaruuValley and locationsfrom which dated sampleswere collected.Average flow dips are 10ø;one dike is nearly vertical,as pictured.Although shown as a simple montonic successionof flows, the section does exhibit evidenceof minor inversionsof relief due to periods of erosionand valley-filling flows. Magnetostratigraphicdata are from Chauvinet al. [1990]; sampleages (in millions of years) are indicated. 24,344 DUNCAN ET AL.: GEOCHEMICALEVOLLrIION OF TAHITI

Table 1. K-Ar Age Determinationsfor VolcanicRocks From Tahiti, French Polynesia

%K Radiogenic40Ar (rad) Age + 1c•, Sample Location (xl0 -7 cm/•[) % Ma

Tahiti Nui T85-41 center, flow 1.57 0.9967 17.1 1.67 + 0.05 T85-43 center, dike 1.81 0.5041 9.6 0.73 + 0.03 T85-44 center,flow 1.78 0.7773 12.6 1.12 + 0.03 T85-7 Punaruu,flow 1.3 2 0.4812 13.4 0.96 + 0.04 BKV-185* Punainu,flow 0.83 0.3845 27.9 1.19 + 0.02 T85-72' Punainu,flow 0.86 0.3529 15.2 1.08 + 0.02 T85-76' Punainu,flow 1.03 0.3964 10.8 1.01 + 0.05 BKG-85* Punaruu,flow 2.07 0.8767 24.4 1.09 + 0.02 PKB-14* Punainu,flow 1.16 0.4991 18.4 1.11 + 0.03 T85-81 * Punaruu,flow 1.03 0.4092 24.4 1.05 + 0.02 T85-87 Punainu,flow 0.91 0.3526 33.1 0.99 + 0.02 T85-93' Punainu,flow 1.28 0.4872 19.8 1.00 + 0.02 T74-431' Punainu,flow 1.62 0.6396 17.0 1.04 + 0.02 T74-434' Punainu,flow 1.16 0.4233 11.8 0.96 + 0.02 T74-435' Punainu,flow 1.34 0.4701 17.8 0.92 + 0.02 T85-61 Punainu,flow 1.49 0.5307 3.8 0.92 + 0.06 T74-437' Punaruu,flow 1.66 0.4928 14.2 0.78 + 0.01 T85-67' Punaruu,flow 2.19 0.5296 6.6 0.62 + 0.02 T85-19 Papenoo,flow 0.96 0.4715 2.3 1.27 + 0.12 T85-28 Papenoo,flow 0.88 0.3710 7.9 1.09 + 0.05 T85-34 Papenoo,flow 0.84 0.2496 4.7 0.76 + 0.05 T85-101 Papenoo,flow 1.19 0.2225 6.0 0.49 + 0.02 T85-102 Papenoo,flow 3.14 0.6222 4.7 0.52 + 0.02 T85-24 Papenoo,flow 1.31 0.2630 7.4 0.52 + 0.01 T85-25 Papenoo,flow 1.28 0.2542 1.9 0.51 + 0.07 T85-12 Papenoo,flow 1.16 0.1678 2.6 0.37 + 0.07 T85-33 Papenoo,flow 1.31 0.1273 2.7 0.25 + 0.02 T85-95 Fautaua,flow 0.99 0.2991 32.4 0.78 + 0.01 T85-96 Fautaua,flow 0.88 0.2220 1.6 0.65 + 0.08 T85-47 Vaihiria, flow 1.57 0.5404 2.5 0.89 + 0.07 T85-111 Vaihiria, flow 1.08 0.3445 5.3 0.82 + 0.03 T85-103 Vaihiria, flow 1.44 0.3754 4.2 0.67 + 0.03

Tahiti Iti T85-116 Ahaavini, cobble 3.72 0.7416 4.0 0.51 + 0.03 T85-117 Vaitepiha,dike 1.24 0.4562 1.4 0.95 + 0.18

Age calculationsare basedon the followingdecay and abundance constants: )•œ= 0.581x10-1øYr-l' )•15= 4-962x10-•øYr-•;4OK/K = 1.167x10-4mol/mol. *Reportedinitially by Chauvinet al. [1990].

AEI MS-10S mass spectrometerwith digital data acquisition. gave ages of 1.67 + 0.05 and 1.12 + 0.03 Ma. Both samples Precise ages (typically 1-5% analytical uncertainty) were produced low total oxides (-96 wt. %), indicating some obtained for rocks as young as 0.3 Ma by virtue of relatively replacementof igneousphases with hydrousalteration phases. high sample K contents and daily monitoring of the mass This is a concern because of possible radiogenic 40Ar loss fractionationof our instrument[e.g., Levi et al., 1988]. and/or K mobilization, both of which would disrupt the K-Ar New age determinations, together with some previously system. The K contents of these samples, however, are reporteddata [Chauvinet al., 1990] are presentedin Table 1. In concordant with overall variation of K20 with MgO within general, they are compatiblewith earlier reportedradiometric shield lavas, and we conclude that K loss (or addition) has not agesfrom Tahiti [Beckeret al., 1974; Dymond, 1975; Duncan been significant. Radiogenic 4OArloss is probablebut harder and McDougall, 1976; Leotot et al., 1990]. A few, much older to evaluate without additional age determinations from the ages[Krurnrnenacher and Noetzlin, 1966;Dyrnond, 1975] from central section of lavas. Hence we view the 1.67 Ma age as plutonic rocks are certainly not reliable crystallizationages approximate, but it is probably a minimum age given the due to excess4OAr trapped during cooling.Because they are tied likelihood of some Ar loss. This section is cut by a dike of to the volcanic stratigraphy through outcrop sampling, the basanitic composition,dated at 0.73 + 0.04 Ma, which we new ages place a definitive time frame on the volcanic surmise fed valley-filling flows. The Punaruu, Papenoo,and evolution of the island. Vaihiria canyon sequencesall producedages in the 1.3 to 0.6 The oldestexposed rocks are from the erodedcenter of the Ma range,which we ascribeto the main shield-buildingphase. island. Few of these were fresh enough for dating; zeolites, The older central lavas may sensiblybe included in this phase clays and carbonatehave replacedmuch of the groundmassand on structuralgrounds but have been separatedhere on the basis olivine phenocrysts.Two of the freshest samples,however, of compositions[Duncan et al., 1987; Cheng et al., 1993] DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAHITI 24,345

(also see below). Valley-filling flows began about 0.7 Ma and table includes rocks from a number of locations from both continued to at least 0.25 Ma. Hence shield and valley-filling volcanoes of Tahiti. The second half of the table includes phasesoverlapped for a short time, at 0.6 to 0.7 Ma. Volcanic extrusive rocks of the shield-buildingphase of volcanism from activity at Tahiti Iti was contemporaneouswith that at Tahiti the PunaruuValley, arrangedin stratigraphicorder. Nui but occurred over a shorter period, from 1.0 to 0.3 Ma Several laboratories contributed XRF analyses. Samples [Duncan, 1975; Leotot et al., 1990; this study]. analyzed by more than one laboratory show good agreement except that a correction was needed for an interlaboratorybias The Punaruu Valley Section of SiO2 (+2.0 wt. %) between the Oregon State University analyses and those of other laboratories. Comparisons Age determinationson 16 samplesfrom the Punaruuvalley between other analytical techniques are equally favorable, section appear in Table 1, and their locations are shown in except that microprobe analysesof K20 in fused glasseswith Figure 2. The age succession is consistent with the high K20 (> 4 wt. %) are consistently lower than the XRF stratigraphicorder of the lava flows, with the exception of analyses of the same samples. This difference, due to the sample T85-7, which probably comes from a flow filling an volatilization of K20 during melting of the samples, has no old erosional depression (P.-Y. Gillot, personal impact on the conclusionsof this paper. communication, 1994). Additional age control is provided by Trace element concentrations were determined by XRF, flow-by-flow magnetostratigraphy [Chauvin et al., 1990]. DCP, and inductively coupled plama mass spectrometry Three magneticpolarity reversalswere observedin the section (ICPMS) at Cornell University and are presentedin Table A3. and correlate with the Brunhes-Matuyama, Matuyama- For ICPMS analyses, approximately 200 mg of sample were Jaramillo, and Jaramillo-Matuyama transitions. An apparent digestedwith a 4:1 HNO3-HF mixture in Savilex capsulesfor reversed-transitional-reversed(R-T-R) excursion was identified 48 hours. After evaporating to dryness, they were redissolved lower in the sequence,and K-Ar age determinationsaround 1.1 in 8 mL HNO 3 and diluted to 200 mL with distilledwater. The Ma suggest that it corresponds to the Cobb Mountain solutionswere then analyzed using a FISONS Plasmaquad2+ subchron. The Brunhes-Matuyama and the lower Jaramillo mass spectrometerequipped with a high performanceinterface. transitions are defined by only a few intermediate directions The analysis employed the drift-corrected external calibration while many intermediatedirections were observedfor the upper Jaramillo transition and the Cobb Mountain excursion. These technique of Cheatham et al. [1993]. External calibration standards were matrix matched (i.e., solutions of differences were attributed to variations in the rate of eruption interlaboratory or in-house basalt standardswere employed). during construction of the shield sequence. To maximize precision and minimize residual drift errors, each The 700-m-thick Punaruuvalley sectionformed between 1.3 sample was analyzed between 3 and 5 times. A blank and a and 0.6 Ma, yielding an average accumulation rate of 1.0 quality control standard were analyzed with each batch of km/m.y., or approximately one flow every 2000 years. samples. Analytical uncertainty, as judged from repeated Evidence from the radiometric dating and the paleomagnetic analysesof U.S. Geological Survey standardBHVO-1, ranges studies, however, is that flows were erupted in pulses with from 1.1% for Yb and Lu to 5.4% for Th. Accuracy has been significanttime gaps rather than at regular intervals.This may assessedfrom the mean of the analyses of BHVO-1. These reflect morphologic evolution of the volcano and preferential values are in excellent agreement with those determined at channeling of flows or may reflect changes in location of Cornell by the more precise methods of thermal ionization eruptive centers. The majority (80%) of the section formed mass spectrometry isotope dilution and ICPMS standard between 1.3 and 0.9 Ma, indicating an early period of intense addition [Cheatham et al., 1993; White et al., 1993]. activity, followed by a period of intermittent activity between Sr, Nd, and Pb isotopic compositionswere determined for a 0.9 and 0.6 Ma. Flow accumulationrates droppedby a factor of number of samples (Table A4). Values for interlaboratory 4 between these two intervals. The timing of change in magma standards and estimates of analytical uncertainty based on supply correspondsapproximately with a shift in shield- replication of those standardsare given. In addition to samples building 'activity to Tahiti Iti and with the beginning of collected in 1985, we include in this table several samples valley-filling volcanism at Tahiti Nui. obtained from the Smithsonian Institution. Most samples were analyzed on a VG Sector mass spectrometerat Cornell Geochemistry' Methods and Results using methods described by White et al. [1990]. A few were run on a VG Sector 54 mass spectrometer at Cornell using Geochemical data in supplementaryTables A1-A4 (major essentiallyidentical methods. We analyzed two samples(T85- and trace element concentrations and isotopic compositions) 41 and T85-19) also analyzed by Cheng et al. [1993]. Our Nd for samplesanalyzed here are availableelectronically 1 or as isotopes ratios agree well with those they reported, and our hard copies from the authors. The major element 87Sr/86Srratio for T85-41 (after leaching) agrees exactly with concentrationswere determined by X ray fluorescence(XRF), the ratio they report. Our 87Sr/86Srratio for T85-19 is about by microprobe of fused glasses (MFG) at Oregon State 0.00004 less than the value they report. The difference may University, and by direct currentplasma spectroscopy (DCP) at reflect slight contamination; we leached samples before the Lamont-Doherty Earth Observatory of Columbia analysis, but Cheng et al. did not. The Pb isotope ratios we University. The major element analyses are presented determined for T85-41 are significantly lower than the values electronically, divided into two parts. The first half of the reported by Cheng et al. [1993]. The difference may also reflect slight contaminationthat was removedby leachingin IAn electronicsupplement of thismaterial may be obtainedon a our procedure.Alternatively, it may reflect a differencein the disketteor AnonymousFTP from KOSMOS.AGU.ORG. (LOGIN to way the fractionation correction was performed or an AGU's FTP account using ANONYMOUS as the usernameand interlaboratory bias. GUEST as the password.Go to the right directoryby typing CD A diagram of total alkalis and silica (Figure 3) showsthat APEND. Type LS to see what files are available.Type GET and the the oldest rocks (central) fall within a restrictedrange of 4.6 to name of the file to get it. Finally, type EXIT to leave the system.) (Paper 94JB00991, Tahiti: Geochemicalevolution of a French 5.8 wt. % alkalis and 48 to 49.5 wt. % SiO2. Most rocks from Polynesianvolcano, by Robert A. Duncan et al.) Diskette may be the shield-building stage of volcanism have lower silica and orderedfrom American GeophysicalUnion, 2000 Florida Avenue, total alkalis than do the central rocks. The composition of N.W., Washington, DC 20009; $15.00. Payment must accompany rocks from the last stage of volcanism (valley-filling) order. overlapswith that of the shield-buildingstage but is at the low 24,346 DUNCAN ET AL.: GEOCHF_3/IICAL EVOLUTION OF TAHITI

18 increasingcompatibility. All groupshave similar abundances of these elements, although there are several subtle but significant differences. First, the central lavas show a 15 negative Sr anomaly. We suspect this may reflect fractionation of plagioclase, which is an occasional o • phenocrystin the central lavas but not in other lavas. Second, • 12 the central lavas have a slight positivePb anomaly. Though not large, we believe this feature is real and associatedwith the • 9 Alkaline a e 6

14 ß o Subalkaline ¸ o oß ß •e•> o oo 0 ß ' ' ' ' ' J ! I I I I I g 40 45 50 55 60 • 10 SiO2, wt. % • 8 Figure 3. Silica-alkali diagram for volcanic rocks from Tahiti. Patternedsquares are from the centralarea of Tahiti Nui (oldest lavas), open circles are from the shield-buildingphase ,g 6 of volcanism (intermediate age), and the open diamondsare from the valley-filling series (youngestlavas). Solid circles 4 are from the PunaruuValley (0.6 to 1.2 Ma). Solid diamonds are from Tahiti Iti. Dashed line separates alkaline from subalkalinemagmas. Heavy solid line indicatesthe 2 fractionation trend from a parental shield magma (Table 2) 0 5 10 15 20 under relatively oxidizing conditions(two log units above the quartz-fayalite-magnetite,QFM, buffer). The light solid line MgO, wt. % showsthe liquid line of descentof a magma with lower silica andunder more reducing conditions (onelog unit above QFM) b than the shield magma.

end of the SiO2 range of these lavas (Figure 3). These three groupsof lavas describea trend of decreasingsilica with time, althoughthe PunaruuValley sequencealone coversmuch of the compositionalrange of the central, shield, and valley-filling 4 lavas. These relationshipsalso exist in MgO variation diagrams (Figure 4), which show that the central lavas are more evolved than most of the shield and valley-filling lavas. The shield lavas extend to more evolved compositionsthan do many of the valley-filling lavas. (Here degreeof evolutionis indicated, to first order, by MgO abundance.) Phenocrystsin theserocks ooo are primarily olivine and pyroxene; considerableaccumulation of these phases is evident in some rocks (picrites and ß o ankaramites). Rare earth patterns are shown in Figure 5. With the exception of T85-102, the most rare earth rich sample, the 0 5 10 15 20 patterns are remarkably similar despite the wide variation in major element and isotopic (see below) compositionof these MgO, wt. % samples.All samplesare strongly light rare earth enrichedand have steepmiddle and heavy rare earth patterns,as reportedfor Figure 4. MgO variation diagramsfor Tahiti volcanic rocks. all volcanoesin the Society hotspot lineament [Dostal et al., (a) MgO-FeO. The fractionationlines indicatethe changein 1982]. Assumingthe heavy rare earth patternof the sourceis magmacomposition with about80% crystallizationof a parent flat or nearly so, the steepheavy rare earth patternand the low magma at two different oxygen fugacitiesusing the model of concentrationsof the heavy rare earth elements suggestthat Nielsen [1990]. The model predicts that olivine crystallizes these magmaswere generatedin the garnet stability field. T85- first, followedby pyroxene,followed by magnetite,a sequence 102 is a mugearitewhose high overall abundanceof rare earths in agreementwith the bulk rock chemistryand the petrography and light rare earth enrichment is undoubtedlya result of of the samples. Higher oxygen fugacities result in earlier extensive fractional crystallization. crystallizationof magnetite(heavy line). (b) MgO-K20. Liquid Figure 6 displays average incompatible element lines of descentfor two starting compositionswith different concentrationsfor central, shield, late-stage, and Tahiti Iti K20 abundance(Table 2) are shown. Symbolsare the sameas groupsin a "spiderdiagram," in which elementsare orderedby in Figure 3. DUNCAN ET AL.' GEOCHEMICAL EVOLUTION OF TAHITI 24,347

500

100

5 I I I I I I I I I I I I I I I La Ce Pr Nd Pm $m Eu Gd Tb Dy Ho Er Tm Yb Lu

Figure 5. Chondrite-normalizedrare earthelement patterns of Tahitianvolcanics. Normalization values are slightlymodified from the average of 20 ordinarychondrites reported by Nakamura[ 1974]. mantle source, as Pb/Ce ratios in Society Islands lavas have a negativeK anomaly. The centrallavas at first appearto correlate with isotope ratios (W. M. White, manuscript in be an exception; however, we believe these lavas also are K preparation, 1994). Third, the central lavas have somewhat depleted,but this is disguisedby low abundancesof Ta and Nb. lower abundances of Ta and Nb. This feature is also associated In addition,only the central lavas are depletedin Rb relative to with the source, as Nb/U ratios in Society Islands lavas Ba. correlate with isotope ratios (W. M. White, manuscript in Three observations suggest that the K depletion is not preparation, 1994). Fourth, Zr and Hf of all groups are related to weathering. First, the oldest (central) least depleted overabundantcompared to rare earthsof similar compatibility lavas (Figure 6) shouldhave the greatestK depletionbased on (e.g., Sm). Since Hf/Sm ratios in Society Islands lavas also their degreeof weathering. Second,weathering should produce correlatewith isotoperatios, we believe this featureis real and greater depletionsin Rb than K, but there are actually slight associatedwith the Societiesmantle plume. Finally, all lavas enrichments of Rb in the shield, late-stage, and Tahiti Iti

l ooo

RbBaTh U NbTa K LaCePbPrSrNdPmSmZrHf EuGdTbDy Y HoErTmYb Lu

Figure 6. "Spider" diagram showing average abundancesof incompatible elements in central, shield, late- stage, and Tahiti Iti lavas, normalized to a primitive mantle composition. Normalization values are from Sun and McDonough [1989]. 24,348 DUNCAN ET AL.: GEOCHEMIC• EVOLUTION OF TAHITI lavas. Third, depletionof K by 30% (suggestedby comparing with that of Cheng et al. [1993], are shownin Figure 10. The K abundancewith Ta and La abundance,Figure 6) would result combineddata clearly show a general decreasein 878r/86Srwith in the calculatedages of the PunaruuValley basaltsbeing 30% time but do not support the idea that volcanic stages too old and inconsistent with the well-dated paleomagnetic beginningat -1.7 Ma, 1.25 Ma, and 0.8 Ma had initially high boundaries (Figure 2). The close agreementof the ICPMS 87Sr/86Srthat subsequentlydecreased. The temporalevolution potassium analyses with the atomic absorption of magmageochemistry in Tahiti appearsto be broadlysimilar spectrophotometricanalyses (Table A1) shows no analytical to that observed in Hawaii [e.g., Chen and Frey, 1985; bias. Hence we conclude that the K depletion of the Tahiti Lanphere and Frey, 1987; Hofmann et al., 1987]. In both lavas must be a primary feature of these rocks. The only Hawaii and Tahiti, radiogenicisotope ratios shift toward more observedeffect of weatheringis the slight depletionof Rb in depleted signatures with time, and the magmas become the central lavas. increasingly silica undersaturated. In Figure 7, œNdis plottedagainst 87Sr/86Sr. The Tahiti data fall entirely within the field defined by previouslypublished Society Islands and Seamountsdata [White and Hofrnann, Discussion 1982; Devey et al., 1990]. Figure 8 displays Pb isotope ratiosplotted in the usual manner. Again, our data fall with the Punaruu Sequence field previously defined for the Societies. In contrastto the range of Sr and Nd isotope ratios, the range in Pb isotope This sequenceis our best sampled eruptive series on the ratios is notably restricted, a feature characteristic of the island and providesan opportunityto model the evolutionof Societies in general [White, 1985; Devey et al., 1990]. The the magmatic systemthat producedthese lavas. Much of the 207Pb/204pband 208pb/204pbisotope ratios reportedby Cheng et sequencewas erupted over a period of 300,000 years. The al. [1993] are generally higher than our values, as discussed chemicalstratigraphy is shownin Figure 11; K20 is usedfor earlier. Sr and Nd isotope ratios are strongly correlated, a this time seriesfor a numberof reasons. (1) It is a major feature suggestiveof two-componentmixing. However, the element (> 1 wt. % in most samples),so it has been analyzed correlationsbetween Pb and Sr isotoperatios (Figure 9) and Pb in all samples(only a subsetof the rocks was analyzed for and Nd isotope ratios (not shown) are very much poorer and trace elements), and there does not appear to be any would preclude the possibility that the source of Tahitian interlaboratory bias or significant differences between the magmas,and Society magmasgenerally, is a mixture of two various analytical methods. (2) K20 is an incompatible homogeneouscomponents. elementin the primary fractionatingminerals, olivine, augite, Cheng et al. [1993] found that Sr and Nd isotoperatios in and magnetite, so its abundancein the host magmaticliquid Tahitian magmas decreasedand increased,respectively, with can be estimatedfrom the bulk rock K:O and the phenocryst time. Cheng et al. [1993] also argued that the three volcanic abundances.(3) Pressuresin crustal magma chambershave stagesidentified by Duncan et al. [1987] began with high Sr little effect on the partitioncoefficient of K:O in the primary and low Nd isotope ratios that respectively decreasedand phenocrysts,so pressurehas little effect on the K:O in the increasedwith time. Our Nd isotope and age data, combined magmas. This is important because even though our

I .....,.....• ....• , 6 gNd 4

-2 o.•o2• o.•o• o.•o• o.•m o.•o4• o.•o• o.•o• o.•o, o.•o,• o.•o• 8]SR/g6SR

Figure 7. Sr and Nd isotopic ratios of volcanic rocks from Tahiti. The group "Tahiti other" includes the Smithsoniansamples (Table A4), whose ages and stratigraphicpositions are u•nown, as well as data from Chenget al. [1993]. Fields for literaturedata for mid-oceanridge basalts(MORB), Galapagos,Kerguelen, Saint Helena, Tubuai and Mangala, Raratonga,San Felix, and San Ambrosio, and other Society Islands and seamountsare shownfor comparison. Note the large rangeof the Tahitian data md the generallinearity of the Societies a•ay. DUNCAN ET AL.: GEOCHF_3/IICAL EVOLUTION OF TAHITI 24,349

39.3

39.2 39.1.:,:,,,.,.,• 38.9 38.8 .....

38.6Ga•pa• ' &ster 38.5 38.4 38.3

38.238.1 Pac• MORB 38 18.5 18.6 18.7 18.8 18.9 19 19.1 19.2 19.• 19.4 19.5 206p b/204p b 15.7

......

18.5 18.6 18.7 18.8 18.9 19 19.1 19.2 19.3 19.4 19.5

206p b/204p b

Figure 8. Pb isotoperatios in Tahitianlavas. Fieldsfor literaturedata for mid-oceanridge basalts (MORB), Galapagos,Easter, Tubuai and Mangaia, Raratonga, San Felix, andSan Ambrosio, and other Society Islands and seamountsare shownfor comparison.Note that the rangeof theseplots is ratherlimited, reflecting the limitedrange in Pb isotoperatios in Societieslavas. For example,the field for Pb isotoperatios in G•lapagos lavasextends off both endsof the plot. 2 tT error barsare shownfor reference.

fractionation model was calibrated with data from 1-atm and eruption. The modified model relies on experimental experiments,the models will be applicable to fractionationin phase equilibrium and phase compositionsof alkali basalts a crustalmagma chamber. and on the observed phase equilibria of the Tahiti samples. Our assumptionthat K:O has behavedas an incompatible Sample outputs of this calculation (for simple fractional element (and appropriate to monitor magma evolution) is crystallization)are shownin Figures 3 and 4. justified in a plot of P2Os/K:O versus K:O (Figure 12). The model requires that a number of initial conditionsbe P•Os/K:O is uniform (at -0.35) over a wide rangeof K:O; hence specified:the compositionof the parent magma,the oxygen P:Os and K•O are equally incompatiblewithin the range of fugacity within the magma chamber, and the processes fractionation,especially within the Punaruusection. We note, operatingwithin the magmachamber. The processesinclude: again, that there is no evidence for K loss from weathering, (1) homogeneousfractional crystallization with no rechargeor even in the central flows. A few late-stagelavas have higher eruption, which is depictedin Figures 3 and 4; (2) magma P:O:/K:O, due to higher P:Os. K:O is also highly correlated chamberrecharge; and (3) eruptionfrom the magmachamber. with other incompatible elements (e.g., Zr) for the Punaruu If processes(2) and (3) are permitted,then assumptionsabout flows, reiteratingthe consistentlyincompatible behavior of K their timing and the relative proportionsof crystallizationand during fractional crystallization. eruption relative to recharge are required. In the models The computermodel used to evaluate the evolution of the discussedhere, rechargeis assumedto occurafter an amountof Tahiti magmatic system is a modified version of the crystallization (for example, 10, 20, or 30%) within the differentiation model of Nielsen [1990]. This model calculates magma chamber. However, the amount of crystallization the composition of magmatic liquids resulting from low- between recharge events is allowed to vary in a Gaussian pressurefractional crystallization, magma chamber recharge, distribution about these specified values. The total of 24,350 -DUNCAN ET AL.: GE(•HEMICAL EVOLU'I•ON OF TAHITI

0,707

0.706

0.705

RAI"eOTONqA Felbc & S. A•bnoslo

0.704,

0.703 18.5 18.6 18.7 18.8 18.9 19 19.1 19.2 19.3 19.4. 19.5 206p b/204p b

Figure 9. Relationshipbetween Sr and Pb isotoperatios in Tahitian lavas. Fields for literature data for Galapagos,Easter, Tubuai and Mangaia, Raratonga, San Felix and SanAmbrosio, and other Society Islands and seamountsare shown for comparison.The correlationbetween 87Sr/86Sr and 206pb/204pb is poor, particularly in comparisonto thatbetween 87Sr/86Sr and end. crystallizationplus eruption relative to rechargedetermines range of FeO (total) abundancein rocks with high MgO is whether the magma chamberis growing, dying, or at steady relativelyrestricted, especially for the shieldlavas (Figure 4a), state. We have examineddying magmachambers (no recharge, and this trend is due to the accumulationor crystallizationof Figures3 and 4) and steadystate magma chambers (recharge olivine. The most primitive olivines found in this suite of equalscrystallization plus eruption,Figure 11). lavas are Fo89.This constraint,along with the limited rangeof The parentalmagma was estimatedwith a methodpreviously FeO (11 to 12 wt. %) and the establishediron-magnesium K D used for a suite of ocean island basalts[Fisk et al., 1988]. The for olivine-basalt equilibrium [Roeder and Emslie, 1970],

•;Nd 3

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 AgE, Ma Figure10. end plottedagainst age for Tahitianlavas. "Other" includes data from Cheng et al. [1993].There is a systematicchange to more depleted,or moreMORB-like, isotopiccompositions With time, althoughthe two lavaswith the leastradiogenic compositions erupted between 1.2 and 1.0 Ma. DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAHI• 24,351

3.5/A DATA •J. MODELS 3.0 Rh 4 2.5

1.5

1.0

0.5 - a • , s • s s s 1..Ma.... 5 10 15 20 25 30 35 40 45 50 • 6 4 R 0 Sequence number Volume crystallized

3.5 3.0 B DATA MODELS 2.5

1.5

1.0

0.5 5 10 15 20 25 30 35 40 45 50 2.4 1.8 1.2 0.6 0 Sequence number Volume crystallized Figure 11. K20 variationwith positionin the PunaruuValley sectionand computer-generatedmagma compositions.The sequence numbers correspond to those in electronicsupplement Table A1. In theleft hand panelsthe light line is thewhole rock abundance, and the heavy line withsolid circles has been corrected for phenocrystcontent. The line at thebottom shows the 0.3 m.y.interval over which most lavas were erupted. Lavas3 to 23 wereerupted in a periodof about50,000 years. (a) The rightpanel shows the effect of changing theratio of crystallizationto eruption(C/E) in a steadystate magma chamber. As C/E increases,the steady stateK20 abundanceincreases. In all modelsthe amount of crystallizationbetween recharge events averages 20%. Volumecrystallized indicates the amountof crystallizationrelative to the originalmagma chamber volume. (b) The right panelshows the effectof changingthe averageamount of crystallizationbetween rechargeevents in the casewhere C/E = 2.

¸ Punaruu Valley 0.8 ß other dated samples

0.6

¸ ¸o ¸ 0.,1 o0o%0 o 0.2 o oC%o% o

i i i i I i ß ß ß ß ß ß ß ß i ß ß ß ß I .... i . . .

0 0.5 1 1.5 2 2.5 3

K20, wt. % Figure 12. For the majorityof the Punaruubasalts and otherdated samples, P:Os/K20 is uniform(at --0.35), indicatingthat K:O and P:O5 are equallyincompatible and that K loss duringweathering has not been a significantconcern. A few samples,all from late-stagelavas, have high P:Os/K20 due to higherP20 5 contents. 24,352 DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAH1TI requiresan MgO contentof the parent magmaof 13 to 14.2 wt. The presence of 23 lavas within the Jaramillo reversed %. Other oxide abundanceswithin this parent magma can be period (about 50,000 years) suggests that these eruptions estimated (Table 2) by examining MgO-oxide variation occurred, on average, every 2000 years. Within this 50-kyr diagrams(such as Figure 4b). period there are three seriesof four or more lavas, characterized The chemical differences between the main shield-building by a progressiveincrease in K20 followed by a suddendecrease phase of volcanism and the late-stage volcanism suggestthat in K20. These series may represent episodes of magma there was a change in compositionof the parent magmas for chamber shrinkage followed by recharge, with mean duration the two series.These differencesare obviousin Figures 3 and 4 of 15,000 to 20,000 years. Leotot et al. [1990] documenteda with the late-stagebasanites having lower SiO2 and higherFeO similarly detailed history of eruptive cycles in the and K20 than do the shield-building alkali basalts. The constructionof the westernportion of Tahiti Iti, each lasting estimatedparent magmasfor the shield and late-stagebasalts approximately 13,000 years. are given in Table 2. We have modeled steady state magma chamber processes The oxygen fugacity of the magma chamber is also (Figure 11) and found that the ratio of crystallization to important for these models becauseit controls the appearance eruption (C/E) determinesthe average level of differentiation of magnetite, which when fractionated strongly affects the of the steadystate lavas. This point is illustrated(Figure 1l a, SiOn, TiO:, and FeO abundanceof the lavas (Figures 3 and 4). right panel) with the continuouscomposition of the magma In Figure 3 the liquid line of descent is shown for magmas chamber for three C/E values. The lowest curve (labeled 1) is buffered at two oxygen activities. In the more oxidized for equal amountsof crystallizationand eruption(C/E = 1), and magma, magnetite starts to crystallize shortly after pyroxene, the middle and upper curves are for C/E values of 2 and 4, resulting in a nearly continuous increase in silica with respectively. The curves all start at the K20 content of the increasing differentiation. In the less oxidized magma, parent magma and approachsteady state K20 abundancesof magnetite formation is delayed so the crystallizationsequence about 1.3 wt. %, 2.2 wt. %, and 3.2 wt. % with increasingC/E of olivine, olivine plus pyroxene, and olivine plus pyroxene value. plus magnetite results in a zigzag path of silica abundance Another feature of the steadystate modelsis that high C/E (Figure 3). These models indicate that the oxygen fugacity correspondsto a slow approachto the steady state abundance decreased as the shield-building volcanism gave way to late- of K20. The full scale in the right panel of Figure 11a stage volcanism. This supports our hypothesis (developed represents800% crystallizationof the magma chamber(eight below) that the mantle sourceof the Tahitian magmasevolved magma chambervolumes), and K20 abundancereaches a steady from OIB or plume (more oxidized [Ballhaus, 1993]) to a state value of about 3.2 wt. % when C/E = 4 only after about MORB or upper mantle source(less oxidized). 800% crystallization. For a C/E = 1 the steadystate value is The computer model calculates the composition of the approached after about 200% crystallization. Since residual liquid for every 0.2% crystallization.These calculated crystallization plus eruption equals recharge (C + E = compositionsshould be comparedwith the liquids eruptedfrom recharge), the models 1, 2, and 4 (Figure 1la) reach a steady the magma chamber. The compositions of the lavas that we state after recharge volumes that are 4, 6, and 10 times the report, however, include phenocrysts, so we cannot make a volume of the magma chamber. The best fit average K20 direct comparison of the calculated liquids from the models abundance in the Punaruu section (after correction for with these bulk rock compositions. We have therefore phenocrystcontent) appearsto be for models where C/E =1 to determined the modal abundance of phenocrysts(by point 2 (Figure 1la), in which 50 to 67% of the rechargedmagma counting) for most of the Punaruu Valley basaltsand corrected crystallizes in the magma chamber and the remainder erupts. the bulk rock chemistry accordingly. The effect of this These models match the average K20 abundance,but the correctionon K:O can be seenin Figure 11. The most evolved amplitude of the K20 variations in the steady state magma magmas (with the highest K:O) also tend to have the highest chamber (Figure 1l a, right panel) is much smaller than the phenocryst abundance so that the variation in matrix K20 observedvariations of K20 (Figure1 l a, left panel). (heavy line, Figure 11) has the same pattern as that for the In the three models of Figure 1 l a, the average amount of bulk rock K:O (light line, Figure 11) but a larger amplitude. crystallization between recharge events was 20%, which The majority of the Punaruu Valley lava flows that we results in an average amplitude of variation of K20 between examined (46 samplesin about 700 m of section)erupted over recharge events of 0.2 to 0.4 wt. %. The amount of 300,000 years. They may be considereda randomsample of crystallizationbetween rechargeevents varied randomlyabout the magmatic system at approximately 6500-year intervals. this 20% value, simulatingan unsteadyrecharge of the magma Our models have assumeda single magma chambererupting chamber and irregular variation in magma chamberchemistry. and recharging over this period, but we cannot dismiss the Figure llb shows the results of the crystallization model in possibility of several independent chambers operating which the C/E = 2 and the average amount of crystallization contemporaneouslyor in sequence.However, we see no long- between recharge events was either 10% or 30%. The term trends in composition of the majority of Punaruu lavas, amplitude of the variations of the magma chamberchemistry althoughthere is a significantincrease in P:O5 in the youngest for the 30% model is similar to that for the variations observed lavas, consistent with the shift to more alkalic lavas at about in PunaruuValley flows but the amplitudeof the K20 variations 0.9 Ma. for the 10% model is much smaller than observed.

Table 2. Starting Compositionsfor Models

Suite SiO2 A1203 FeO Fe203 MgO CaO Na20 K20 TiO2 P205 Mn

Shield 46.4 10.5 10.0 2.0 13.8 10.9 2.0 0.8 2.8 0.40 0.20 Late stage 44.0 10.8 10.7 1.9 13.8 11.2 2.5 1.0 3.3 0.55 0.20

Values are given in weight percent. DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAHITI 24,353

For the 10% model to apply to the Punaruu section, simple two-component mixture of depleted upper mantle (i.e., successive magmas must have sampled the chamber at MORB source) and the Society mantle plume. (We noted intervals separatedby large amountsof crystallization. The earlier that the Pb isotope data preclude such a simple two- 23 lavas within the Jaramillo reversed period suggest that component model of the source. We nevertheless feel that these magmaserupted in rapid sequenceand thereforedo not comparison of the data with such a model is useful, and we represent eruptions separated in time by large amounts of return to the implicationsof the Pb isotopedata at the end of crystallization and magma chamber recharge. They are this section.) For the depleted upper mantle, we assumedan therefore probably derived from a magma chamber that expresseda wide rangeof crystallizationover a shortperiod of end of 9.8 and a [La/Sm]N of 0.4, which would give rise to a time, as is illustrated by the model with 30% crystallization typical light rare earth depletedMORB upon 15% melting. For between recharge events. the plume (the enrichedend-member), we assumedan œNdof -1, which is equal to the lowest end observedin the Society The Temporal Evolution of Tahitian Volcanism Islands (sample 73-185 from Tahaa [White and Hofmann, 1982]). We then calculated the rare earth abundances for the An important feature of Tahitian rock compositionsis the sourceof 73-185 using simple batch melting. (More complex systematicvariation in time that is apparentin isotope ratios melting models do not give substantiallydifferent results.) In and major element chemistry. As lava chemistry became doing so, we assumed15% melting of garnet peridotite, based increasinglysilica undersaturated,the Sr and Nd isotoperatios on the reasoning that the plume should be at least as hot as shifted toward more depleted,more MORB-like, signatures.In surroundingupper mantle and therefore that degree of melting light of this, the lack of systematicvariation in incompatible should be at least as great as is typical of MORB, and we element abundancesor ratios is remarkable. The changes in somewhatarbitrarily assumed33% olivine fractionation. The 87Sr/86Srand œNd imply that the mantle sourcebecame more [La/Sm]N calculatedfor the plume from these assumptionswas depletedin incompatibleelements with time, which shouldbe 3.08. From the end and rare earth abundancesof the two reflected in the incompatibleelement abundancesof the lavas model end-members, we calculated the mixing line shown by if all other factors remain constant. The only reasonable the heavy line in Figure 13. Dashed lines in Figure 13 show inference that can be drawn from the constancy of the calculated composition of 10%, 5%, 2%, and 1% partial incompatibleelement ratios suchas La/Sm is that the degreeof melts of this mixture. Implicit in this model is the assumption melting must have decreasedthrough time. that the mineralogy of the plume does not differ substantially To illustrate this point, we have superimposedsome simple from that of the asthenosphere. model calculationson a plot of end versus[La/Sm]N in Figure The mixing line shows that La/Sm and end in the source 13. We assumed that the source of Tahitian magmas is a should be inversely related. No such inverse relationship is

Figure 13. end plottedversus chondrite-normalized La/Sm ratio for Tahitianbasalts (data from differentiated rocksare not plotted). Superimposedon the data is a mixing melting model. The heavy line representsmixing betweenthe assumedcomposition of the Societyplume (œNd= -1, [La/Sm]N = 3.08, La = 0.2 ppm, Nd = 0.84 ppm, and Sm = 0.32 ppm) and depletedmantle (œNd = +9.8, [La/Sm]N = 0.4, La = 8.1 ppm, Nd = 9.3 ppm, and Sm = 1.6 ppm). Labeledticks showthe percentageof depletedmantle in the mixture. Dashedlines labeledin italics show the compositionof 1, 2, 3, 5, and 10% melts of this mixture. These calculationsassume melting occurs in the garnet stability field, but the results would not be greatly different for melting of spinel peridotite.Comparison of the actualdata with this modelshows that the percentageof meltingmust decline as the fractionof depletedmantle in the sourceincreases. 24,354 DUNCAN ET AL.: GEOCHEMICALEVOLUTION OF TAHITI

apparentin the actual data. Indeed, the most light rare earth enriched melts are associated with the highest end. MAIN ShIeld Phase Comparisonof the data with our calculatedmixing-melting model clearly indicatesthat the magmaswith highestend must have been producedby the smallestdegree of melting. Since end increaseswith time in Tahitianmagmas, this suggeststhat the degreeof meltingdecreased with time. Assumingthat our estimatesabout the degreeof melting for MORB and 73-185 are correct (i.e., both are 15% melts), then the earliest central magmaswere producedby 5-15% meltingwhile the late shield and late-stagemagmas were producedby as little as 1-2% melting. Even if our estimateof the specificamount of source meltingis incorrect,the degreeof meltingapparently declined by an order of magnitude over the course of Tahitian volcanism. Our conclusions in this respect are entirely consistentwith those of Cheng et al. [1993]. If the most radiogenicTahaa sample(73-185) representsthe isotopic compositionof the Societies mantle plume, then even the most enriched Tahitian samplesare derived from a mixed sourcethat containsnearly 40% depletedmantle. In the LaTe Phase simpletwo-component model illustrated in Figure 13, the late- stage sampleswith highest end are derived from a source consistingpredominantly of depletedmantle with less than 10% admixture of the Societies plume. The progressionto a more depletedsource and to small degreesof melting in Tahiti is similar to that observedin Hawaii. Chen and Frey [1985] proposed that the compositional evolution of Hawaiian volcanoes reflects passageof the volcanoover the plume. We adopta similar modelto explainour observationson Tahiti (Figure14). Our model differs from theirs in that we argue that the depleted signatureof the late-stagemagmas primarily reflects melting of depletedmantle entrained by the plumerather than melting of the lithosphere. In our model, volcanic evolution at Tahiti reflects the rheologicaland thermalinteraction between the plumeand the surroundingmantle. Mantle viscositiesare in the range of Figure 14. Cartoon illustratinghow volcanic evolutionof 10•8 to 1022p. Plumesappear to be typically 200øC or more Tahiti is controlledby passageover a chemicallyand thermally hotter than normal upper mantle [e.g., Klein and Langrnuir, zoned plume. Shading is used to suggest variations in 1987; Sleep,1990]. The assumedtemperature dependence of temperatureand consequentlyextent of melting. The central viscosity of core of the plume, consistingof deep mantle incompatible element enriched and radiogenic material, is hottest. Temperaturesin the surroundingincompatible element depleted tl = t!0 exp (-[•AT/T') and less radiogenic material, entrained through viscous coupling and heating by the rising plume, decreaseradially with 15= 35, T'= 2300øKand AT = 200øK[Stacey and Loper, away from the centralcore. In the main shieldphase, magmas 1983; Richards et al., 1988] implies a plume viscosityonly 1 containa relatively large fractionof high-degreemelts derived to 2 ordersof magnitudelower than surroundingmantle. These from the plumecore. In the late stage,the contributionfrom viscosities mean that a vertical, axisymmetric plume cannot the core is small, and the bulk of the magmas is derived by rise freely through the mantle; instead, surroundingmantle smaller extentsof melting of the cooler margins. will be viscouslycoupled to the plume and rise with it as a sheath-likeboundary layer. The thicknessof this entrained layer couldreach hundreds of kilometers[Hauri et al., 1993]. rising structurewill pass through the P,T conditionsfor Conductive heat loss from the plume to the boundary layer melting in the upper mantle; however, temperaturesin the causesfurther reductionin viscositycontrast and increasesthe peripheralmaterial will remainsubstantially below that of the buoyancyof the boundarylayer material [Griffiths, 1986]. plume core, so that the entrainedmantle will melt to a lesser Thus the plume will continuouslyentrain surroundingmantle degreethan the initial plume mantle. This melting range is as it rises, so that material closest to the plume core will have reinforcedby the fact that uppermantle, being more depleted been entrainedfrom the deepestlevels, while the outer part is in incompatibleelements, will melt lessat a giventemperature entrained at relatively shallow levels (Figure 14). This thanwill plumemantle. We assumethat the volcanois located produces,in plan view, a concentricallyzoned hotspot,with directlyover the plumecore during its mostvigorous period of the most plume-like material in the center,followed by lower growthin the early shieldstage (central lavas). As the volcano and upper mantle material with increasingradius. The entire is transportedaway from the plume by plate motion, small- DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAHITI 24,355

degree melts of the cooler sheath increasingly dominate the oceanic islands such as the Galapagos. Heterogeneityin Pb waning volcanic activity. isotoperatios in either end-member,despite uniformity in Sr Cheng et al. [1993] suggestedthat the depletedcomponent and Nd isotope ratios, may explain the lack of correlation in in Tahitian lavas was derived from metasomatizedlithosphere. Figure 8. Alternatively, the Pb isotoperatios may indicate the While melting of the lithospheremay contribute in a minor presenceof a third component. One possible componentis way to Tahitian volcanism, we believe, primarily on the basis FOZO ("focal zone"), the compositionthat Hart et al. [1992] of thermal considerations,that it is a less satisfactory source hypothesize to characterize the lower mantle. The Sr and Nd for the depleted componentthan is entrained depleted mantle. isotope compositionsof FOZO lie close to the "enriched"end The 60 Ma lithosphere, particularly in its mid and upper of the MORB array. Because of this, a three-component regions, would have cooled well below its melting mixture of depletedmantle, FOZO, and the Societyplume may temperature. Chenget al. [1993] proposedconductive heating still produce a relatively linear array on a plot of Nd versusSr by the plume. However, conduction is a particularly slow isotoperatios. The Pb isotopecomposition of FOZO is more means of transmitting heat in the mantle. Emerman and radiogenic than that of MORB, so a plot of Sr versus Pb Turcotte [1983] concluded that conductive heating of the isotope ratios would show the weak correlation characteristic lithosphereby a plume is not likely to be significant. of three-componentmixtures. In the context of the model we In addition, the mid and upper portion of the lithosphere propose, the FOZO material would have been entrained from would have been severely depleted in its basaltic component the lower mantle, while depleted mantle would have been and incompatible elements by melting at a spreading center entrained at shallower levels. With the existing data, we and hence would be a poor sourceof basalt. Partly on the basis cannot distinguish between these two possibilities or verify of their analysis of Tahitian xenoliths, Cheng et al. [1993] the presenceof FOZO in the Tahitian source. Additional data, dismissed this problem by suggesting that the lithosphere particularly He isotope data, might ultimately resolve this beneathTahiti had been "metasomatized."Cheng et al. [1993, question. p. 177] suggested no specific mechanism for this metasomatismbut stated only that it "may be related to the Nature of the Society Mantle Plume 'superswell'proposed by McNutt and Fischer [1987]." On the basis of the similarity between the isotopic compositions of White and Hofmann [1982] suggestedthat the distinctive Tahitian xenoliths analyzed by Cheng et al. (eNd -+4 and isotopic compositionsof the Society mantle plume reflect the 878r/86Sr-• 0.7042) and Tahitianmagmas, we suggestthat the presenceof subductedoceanic crust and sedimentin the plume. "residual" xenoliths acquired their isotopic compositions by High Pb/Ce ratios in Society seamount lavas reported by reaction with Tahitian plume magmas that had earlier passed Devey et al. [1990] provided some support for this through the lithosphere, i.e., the agent of metasomatismwas hypothesis. Newsom et al. [1986] showedthat Pb/Ce ratios in simply earlier Tahitian magmas. We see no compelling MORB and many oceanicisland basaltshave uniform values of evidence in our data for the existence of "metasomatized" about 0.036. This value is substantiallybelow the bulk earth lithosphereprior to Tahitian magmatism. value of about 0.1 and lower than Pb/Ce ratios in the One important contrast with Hawaiian volcanism is that continental crust, which are generally in excess of 0.1. Ben while tholeiites predominate in Hawaii, alkali basalts Othman et al. [1989] found that modern marine sediments also predominatein the Societies. The data we report here, as well have values in excess of 0.1 and noted that because of the as those of Cheng et al. [1993] and LeRoy et al. [1993], relatively high concentrationsof Pb in sediment and the low demonstratethat hypersthene-normativelavas do exist in the concentration of Pb in the mantle the Pb/Ce ratio should be a earliest exposed sequences. However, even if tholeiites particularly sensitive indicator of the presence of recycled predominate at some deeper, unexposedlevel, the tholeiite- sediment in mantle plumes. White et al. [1993] have found alkali basalt transitionwould appearto occur at a much earlier that Pb/Ce ratios in Tahaa basaltsrange up to 0.064, which is point in the evolution of Tahiti than is typical of Hawaiian well above the range found in other oceanic basalts by volcanos. Newsom et al. [1986]. In Tahiti, most basalts have Pb/Ce There are two factorsthat may explainthis predominanceof ratios in the range of 0.033 to 0.042, typical of oceanic alkali basalts. First, because they reflect higher time- basalts generally. However, several of the central lavas have integratedRb/Sr and Nd/Sm, the isotopic data require that the higher Pb/Ce ratios, ranging up to 0.050. Furthermore,Pb/Ce Societyplume sourcehas been more enrichedin incompatible ratios in Societies basalts, including the Tahitian samples elements,including K and Na, than the Hawaiian plume source. reportedhere, correlatestrongly with Sr and Nd isotoperatios Thuseven at similarconcentrations of SiO 2, Tahitianmagma [White et al., 1993]. This correlation demonstrates that the compositions will be more alkalic than Hawaiian magmas. elevatedPb/Ce ratiosare a featureof the source,and it strongly Second,the Society plume has a significantlylower buoyancy suggests the incompatible element enrichment thatß flux than does the Hawaiian plume [Sleep, 1990]. This may characterizes the plume is due, at least in part, to deep indicate that the Society plume is cooler than the Hawaiian recycling of crustal material. one. If so, the degree of melting will generally be lower. Magmas producedby lower degreesof melting are typically Conclusions poorerin SiO2 and richerin alkalisand hencewill be more alkalic than higher-degreemelts. The ages of the subaeriallavas at the island of Tahiti, which We now return to the lack of correlationbetween Pb isotope makeup about2% of the volumeof the volcanicsystem, range ratios and Sr and Nd isotope ratios. This observationwould from 1.7 to 0.2 Ma. The majority of these lavas appear to seem to rule out the simple two-component mixing model have erupted between 1.3 and 0.9 Ma. This episode was proposedhere. However, we emphasizethat the total range in followed by a period of less voluminous posterosional Pb isotope ratios in Tahiti is small compared to that of other eruptions. Through the construction of the volcano the lava 24,356 DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAH1TI compositions changed from tholeiitic and alkalic basalts to (West Pacific): A geochemical study, J. Geophys.Res., 95, 5049- basanites, while corresponding Sr and Nd isotopic ratios 5067, 1990. changed to more depleted values. We interpret these trends to Dostal, J., C. Dupuy, and J. M. Liotard, Geochemistryand origin of indicate that the mantle source for melting changed from a basaltic lavas from Society Islands, French Polynesia(south central predominantly plume composition to depleted upper mantle Pacific Ocean), Bull. Volcanol., 45, 51-62, 1982. Duncan, R. A., Linear volcanismin French Polynesia,Ph.D. thesis, 151 with time. We propose that a compositional zonation occurs pp., Aust. Natl. Univ., Canberra,1975. across the Societies hotspot, as a result of entrainment of Duncan,R. A. and D. A. Clague,Pacific plate motionrecorded by linear depleted mantle around the rising plume. During the vigorous volcanic chains,in The Ocean Basinsand Margins, vol. 7A, edited by shield-building phase of volcanism the magma chamber was A. E. M. Nairn, F. G. Stehli, and S. Uyeda, pp. 89-121, Plenum,New recharged at intervals separatedby about 30% crystallization York, 1985. of the magma chamber. More than three rechargeevents are Duncan,R. A., and I. McDougall, Linear volcanismin FrenchPolynesia, observed during the 50,000 year Jaramillo (magnetic normal) J. Volcanol. Geotherrn. Res., 1, 197-227, 1976. period, suggestingthat these events occurred roughly 20,000 Duncan, R. A., M. T. McCulloch, H. G. Barsczus, and D. R. Nelson, years apart. Plume versus lithosphericsources for melts at Ua Pou, Marquesas Islands, Nature, 322, 534-538, 1986. Duncan, R. A., M. R. Fisk, and J. Natland, The development of Acknowledgments. We thank JacquesTalandier (Atomic Energy volcanismat Tahiti, FrenchPolynesia, (abstract) Eos Trans.A GU, 68, Commission, Papeete) and ORSTOM for hospitality and logistical 1521, 1987. support in Tahiti, Hans Barsczus and Jim Natland for assistancein Dymond, J., K-Ar ages of Tahiti and Moorea, Society Islands, and sampling and discussions,Pierrick Roperch and Annick Chauvin for implicationsfor the hot-spotmodel, Geology,3, 236-240, 1975. paleomagnetic core samples from the Punaruu Valley, and Mike Cheatham and Karen Harpp for assistancewith the isotopic analyses. Emerman, S. H., and D. L. Turcotte, Stagnation flow with a The paper has been improved following reviews by Fred Frey, Eric temperature-dependentviscosity, J. Fluid Mech., 127, 507-517, 1983. Hauri, Pierre-Yves Gillot, and Don Swanson. R.A.D. and M.R.F. were Fisk, M. R., B. G. J. Upton, C. E. Ford, and W. M. White, Geochemical supportedby NSF grant EAR85-09012, and W.M.W. by NSF grant and experimentalstudy of the genesisof magmasof ReunionIsland, EAR87-07484. Indian Ocean, J. Geophys.Res., 93, 4933-4950, 1988. Gillot, P.-Y., J. Talandier,H. Guillou, and I. LeRoy, Temporalevolution References of the volcanismof Tahiti Island: Geologyand structuralevolution, paper presentedat InternationalWorkshop on IntraplateVolcanism: Ballhaus, C., Redox states of lithospheric and asthenosphericupper The PolynesianPlume Province,August 2-7, Centre National de la mantle, Contrib. Mineral. Petrol., 114, 331-338, 1993. RechercheScientifique, Papeete, Tahiti, 1993. Becker, M., R. Brousse, G. Guille, and H. Bellon, Phases d'erosion - Griffiths, R. W., The differing effects of compositionaland thermal comblement de la vallee de la Papenoo et volcanismesubrecent a buoyancieson the evolutionof mantle diapirs,Phys. Earth Planet. Tahiti, en relation avec l'evolution des iles della Societe (Pacifique Inter., 43, 261-273, 1986. Sud), Mar. Geol., 16, 71-77, 1974. Gumis, M., and G. F. Davies, Mixing in numericalmodels of mantle Ben Othman, D., W. M. White, and J. Patchett,The geochemistryof convection incorporating plate kinematics, J. Geophys.Res., 91, marine sediments, island arc magma genesis, and crust-mantle 6375-6395, 1986. recycling,Earth Planet Sci.Lett., 94, 1-21, 1989. Hart, S. R., E. H. Hauri, L. A. Oschmann, and J. A. Whitehead, Mantle Binard, N., R. C. Maury, G. Guille, J. Talandier, P.Y. Gillot, and J. plumes and entrainment:Isotopic evidence, Science,256, 517-520, Cotten, Mehetia Island, South Pacific: Geology and petrologyof the 1992. emergedpart of the Societyhot spot,J. Volcanol.Geotherrn. Res., 55, Hauri, E. M., J. A. Whitehead,and S. R. Hart, Fluid dynamicconstraints 239-260, 1993. on mixing in mantle plumes, paper presented at International Chauvin, A., P. Roperch, and R. A. Duncan, Recordsof geomagnetic Workshopon IntraplateVolcanism: The PolynesianPlume Province, field reversals from volcanic islands of French Polynesia, 2, August 2-7, Centre National de la RechercheScientifique, Papeete, Paleomagneticstudy of a flow sequence(1.2-0.6 Ma) from the island Tahiti, 1993. of Tahiti and discussionof reversal models, J. Geophys.Res., 95, Hofmann, A. W., M.D. Feigenson,and I. Raczek, Kohala revisited, 2727-2752, 1990. Earth Planet. $ci. Lett., 95, 114-122, 1987. Cheatham,M. M., W. F. Sangrey,and W. M. White, Sourcesof error in Jager, E., W. J. Chen, A. J. Hurford, R. X. Liu, J. C. Hunziker, and D. external calibration ICP-MS analysis of geological samplesand an M. Li, BB-6: A Quaternary age standardfor K-Ar dating, Chern. improved non-linear drift correction,Spectrochirn. Acta B, 48, E487- Geol., 52, 275-279, 1985. E506, 1993. Klein, E. M., and C. H. Langmuir, Ocean ridge basaltchemistry, axial Cheminee,J. L., R. Hekinian, J. Talandier, F. Albarede,C. W. Devey, J. depth, crustalthickness and temperaturevariations in the mantle, J. Francheteau, and Y. Lancelot, Geology of an active hot spot: Geophys.Res., 92, 8089-8115, 1987. Teahitia-Mehetiaregion in the southcentral Pacific, Mar. Geophys. Krummenacher,D., and J. Noetzlin, Ages isotopiquesK/Ar de roches Res., 11, 27-50, 1989. prelevees dans les possessionsfrancaises du Pacifique, Bull. Soc. Chen, C.-Y., and F. A. Frey, Trace elementand isotopicgeochemistry of Geol. Fr., 8, 173-175, 1966. lavas from Haleakala Volcano, east Maui, Hawaii: Implications for Lacroix, A., Les roches alkalines de Tahiti, Bull. Soc. Geol. Fr., 10, 91- the origin of Hawaiian basalts,J. Geophys.Res., 90, 8743-8768, 124, 1910. 1985. Lanphere, M. A., and F. A. Frey, Geochemicalevolution of Kohala Cheng,Q. C., J. D. Macdougall,and G. Lugmair,Geochemical studies volcano, Hawaii, Contrib. Mineral. Petrol., 95, 100-113, 1987. of Tahiti, Teahitia and Mehetia, Society Island chain, J. Volcanol. Leotot, C., P.-Y. Gillot, F. Guichard, and R. Brousse, Le volcan de Geotherrn. Res., 55, 155-184, 1993. Taravao (Tahiti): un examplede volcanismepolyphasique associe a DeNeufbourg, G., Notice explicative sur la feuille Tahiti, scale une structure deffondrement. Bull. Soc. Geol. Fr., 8, 951-962, 1990. 1/40,000, Bur. Rech. Geol. Min., Paris, 1965. LeRoy, I., L. Turpin, H. Guillou, S. Chiesa, P.-Y. Gillot, and F. Desonie, D. L., R. A. Duncan, and J. M. Natland, Temporal and Guichard, Temporal evolution of the volcanism of Tahiti, 2, geochemicalvariability of productsof the Marquesashotspot, J. Geochemistry and isotopic compositions, paper presented at Geophys.Res., 98, 17649-17665, 1993. International Workshop on Intraplate Volcanism: The Polynesian Devey, C. W., F. Albarede,J. L. Cheminee,A. Michard, R. Muhe, and Plume Province, August 2-7, Centre National de la Recherche P. Stoffers,Active submarinevolcanism on the Societyhotspot swell Scientifique,Papeete, Tahiti, 1993. DUNCAN ET AL.: GEOCHEMICAL EVOLUTION OF TAHITI 24,357

Levi, S., H. Audunsson,R. A. Duncan,L. Kristjansson,P.-Y. Gillot, and Sleep,N.H., Hotspotsand mantleplumes: Some phenomenology, J. S. P. Jakobsson,Late Pleistocenegeomagnetic excursion in Icelandic Geophys.Res., 95, 6715-6736, 1990. lavas: Confirmationof the Laschampexcursion, Earth Planet. Sci. Stacey, F. D., and D. E. Loper, The thermal boundary layer Lett., 96, 443-457, 1988. interpretationof D" and its role as a plume source,Phys. Earth Macdonald, G. A., and T. Katsura, Chemical compositionof Hawaiian Planet. Inter., 33, 45-55, 1983. lavas, J. Petrol., 8, 82-133, 1964. Sun, S., and W. F. McDonough,Chemical and isotopicsystematics of Marshall,P., The geologyof Tahiti,Trans. Proc. N.Z. Inst.,47, 361-377, oceanicbasalts: Implications for mantlecomposition and processes, 1915. in Magmatismin the OceanBasins, edited by A.D. Saundersand M. McBirney,A. R., and K. Aoki, Petrologyof the islandof Tahiti, in J. Norry, GeologicalSociety of LondonSpec Pub. 42, London,pp. Studiesin Volcanology:A Memoir in Honor of Howell Williams, 313-345, 1989. editedby R. R. Coats,R. L. Hay, andC. A. Anderson,Mere. Geol. Talandier, J., and E. Okal, The volcanoseismicswarms of 1981-1983 in Soc. Am., 116, 523-556, 1968. the Tahiti-Mehetia area, French Polynesia, J. Geophys.Res., 89, McNutt, M. K., and K. M. Fischer, The South Pacific superswell,in 11,216-11,234, 1984. Seamounts,Islands, and Atolls, Geophys. Monogr. Ser., vol. 43, edited Tracy, R. J., Petrology and genetic significanceof an ultramafic by B. H. Keating,P. Fryer,R. Batiza,and G. W. Boehlert,pp. 25-34, xenolith suite from Tahiti, Earth Planet. Sci. Lett., 48,80-96, 1980. AGU, Washington,D.C., 1987. White, W. M., The sources of ocean basalts: Radiogenic isotopic Morgan,W. J., Convectionplumes in the lowermantle, Nature, 230, 42- evidence,Geology, 13, 115-118, 1985. 43, 1971. White, W. M., and A. W. Hofmann, Sr and Nd isotopegeochemistry of Morgan,W. J., Deep mantleconvection plumes and plate motions,Am. oceanic basalts and mantle geochemistry,Nature, 296, 821-825, Assoc. Pet. Geol. Bull., 56, 203-213, 1972. 1982. Nakamura, N., Determination of REE, Ba, Fe, Mg, Na and K in White, W. M., M. M. Cheatham, and R. A. Duncan, Isotope carbonaceousand ordinary chondrites,Geochim. Cosmochim.Acta, geochemistryof Leg 115 basaltsand inferences on the historyof the 38, 757-775, 1974. Reunionmantle plume,Proc. OceanDrill Program,Sci Results,v. Natland, J. H., and D. L. Turner, Age progressionand petrological 115, 53-61, 1990. developmentof Samoanshield volcanoes: Evidence from K-Ar ages, White, W. M., J. M. Bird, K. S. Harpp, R. A. Duncan,R. Brousse,and lava compositions,and mineral studies, in GeologicalInvestigations of T. Gisbert, Volcanic evolution of Tahaa: Does erosion causepost- the Northern Melanesioan Borderland, Circum-Pacific Council for erosionalvolcanism?, paper presented at InternationalWorkshop on Energyand Mineral Resources,Earth Sci. Ser., vol. 3, editedby T. Intraplatevolcanism: The PolynesianPlume Province,August 2-7, M. Brocher,pp. 139-171,Houston, Tex., 1985. Centre National de la RechercheScientifique, Papeete,Tahiti, 1993. Newsom, H. E., W. M. White, K. P. Jochum, and A. W. Hofmann, Williams, H., Geologyof Tahiti, Mooreaand Maiao, Bull. 105, 89 pp., Siderophileand chalcophileelement abundances in oceanicbasalts, B.P. BishopMus., Honolulu,HI, 1933. Pb isotopeevolution and growthof the Earth's core, Earth Planet. Sci. Lett., 80, 299-313, 1986. Nielsen, R. L., Simulation of igneous differentiation processes,in R. A. Duncan, M. R. Fisk, and R. L. Nielsen, College of Oceanicand Modern Methods of IgneousPetrology: Understanding Magmatic AtmosphericSciences, Ocean AdministrationBuilding 104, Oregon Processes,edited by J. Nichollsand J. K. Russell,Rev. Mineral., 24, StateUniversity, Corvallis, OR 97331.(email: [email protected]) 65-105,Mineralogical Society of America,Washington, D.C., 1990. W. M. White, Departmentof GeologicalSciences, Cornell University, Richards,M. A., B. H. Hager,and N.H. Sleep,Dynamically supported Ithaca, NY 14853. geoidhighs over hotspots: Observation and theory, J. Geophys.Res., 93, 7690-7708, 1988. Roeder, P. L., and R. F. Emslie, Olivine-liquid equilibrium, Contrib. (ReceivedNovember 22, 1993; revisedApril 5, 1994; Mineral. Petrol., 29, 275-289, 1970. acceptedApril 12, 1994.)