Heterogeneity of the Caribbean plateau mantle source: Sr, O and He isotopic compositions of olivine and clinopyroxene from Gorgona Island

Heterogeneity of the Caribbean plateau mantle source:Sr, O and He isotopic compositions of olivine and

clinopyroxene from Gorgona Island

S. Re¤villon a;�, C. Chauvel a;b, N.T. Arndt a;b, R. Pik c, F. Martineau a,S. Fourcade a, B. Marty c

a Ge¤osciences Rennes, Universite¤ de Rennes 1, Campus de Beaulieu, 35042 Rennes Cedex, Franceb Laboratoire de Ge¤odynamique des Cha|“nes Alpines, UMR 5025 CNRS, Universite¤ J. Fourier, BP 53, 38041 Grenoble Cedex, Francec Centre de Recherche Pe¤trographique et Ge¤ochimique, Rue Notre Dames des Pauvres, BP 20, 54501 Vandoeuvre-les-Nancy Cedex,

France

Received 20 August 2001; received in revised form 19 September 2002; accepted 3 October 2002

Abstract

The composition of the mantle plumes that created large oceanic plateaus such as Ontong Java or the Caribbean isstill poorly known. Geochemical and isotopic studies on accreted portions of the Caribbean plateau have shown thatthe plume source was heterogeneous and contained isotopically depleted and relatively enriched portions. Adistinctive feature of samples from the Caribbean plateau is their unusual Sr isotopic compositions, which, at a givenNd isotopic ratio, are far higher than in samples from other oceanic plateaus. Sr, O and He isotopic compositions ofwhole rocks and magmatic minerals (clinopyroxene or olivine) separated from komatiites, gabbros and peridotitesfrom Gorgona Island in Colombia were determined to investigate the origin of these anomalously radiogeniccompositions. Sequentially leached clinopyroxenes have Sr isotopic compositions in the range 87Sr/86Sr = 0.70271^0.70352, systematically lower than those of leached and unleached whole rocks. Oxygen isotopic ratios ofclinopyroxene vary within the range N

18O =5.18^5.35x, similar to that recorded in oceanic island basalts. Heisotopic ratios are high (R/Ra = 8^19). The lower 87Sr/86Sr ratios of most of the clinopyroxenes shift the field of theCaribbean plateau in Nd^Sr isotope diagrams toward more ‘normal’ values, i.e. a position closer to the field definedby mid-ocean ridge basalts and oceanic-island basalts. Three clinopyroxenes have slightly higher 87Sr/86Sr ratios thatcannot be explained by an assimilation model. The high 87Sr/86Sr and variations of 143Nd/144Nd are interpreted as asource characteristic. Trace-element ratios, however, are controlled mainly by fractionation during partial melting. Wecombine these isotopic data in a heterogeneous plume source model that accounts for the diversity of isotopicsignatures recorded on Gorgona Island and throughout the Caribbean plateau. The heterogeneities are related to oldrecycled oceanic lithosphere in the plume source; the high 3He/4He ratios may indicate that the source material onceresided in the lower mantle.@ 2002 Elsevier Science B.V. All rights reserved.

0012-821X / 02 / $ ^ see front matter @ 2002 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 0 2 ) 0 1 0 0 3 - 8

* Corresponding author. Present address: Southampton Oceanography Centre, School of Ocean and Earth Science, EuropeanWay, Southampton SO14 3ZH, UK. Fax: +44-2380-59-3052.E-mail addresses: [email protected] (S. Re¤villon), [email protected] (S. Re¤villon).

EPSL 6445 28-11-02

Earth and Planetary Science Letters 205 (2002) 91^106

www.elsevier.com/locate/epsl

Keywords: Caribbean oceanic plateau; Sr; O; He; clinopyroxenes; plumes

1. Introduction

Oceanic plateaus are thought to form by partialmelting within the heads of large starting mantleplumes [1]. The last pulse of oceanic £ood basaltvolcanism occurred during the Cretaceous andgenerated several giant oceanic plateaus such asOntong Java, Kerguelen and the Caribbean pla-teau [2]. The structure and composition of theplume head is not fully understood, particularlythe nature and origin of the various componentsthat constitute the plume source. Studies of oce-anic plateaus are particularly important in thisrespect because they are placed in an oceanic set-ting, without interaction with the continentalcrust. For this reason they provide informationon the composition of the plume head and themantle source that is unavailable from studies ofcontinental plateaus.

Study of the Caribbean plateau o¡ers severaldistinct advantages compared with investigationsof other oceanic plateaus. First, access is easy dueto the presence of extensive on-land sections thatrepresent tectonically accreted portions of the pla-teau. Some of these sections include stratigraphi-cally lower portions of the plateau that are notexposed in other plateaus. Second, the on-landsections of the Caribbean plateau contain highlymagnesian lavas that may represent the parentalmagmas of the thick overlying sequences ofevolved basalts. Plutonic complexes of ma¢c^ul-trama¢c rocks, such as those on Gorgona Islandand the Be¤ata ridge, may form parts of fossilmagma chambers in which magmatic di¡erentia-tion took place.

Detailed studies of accreted portions of the Ca-ribbean oceanic plateau have led, in the last fewyears, to a better understanding of the composi-tion of mantle sources of oceanic plateaus [3^10].Variations in rock type and geochemical charac-teristics of extrusive and intrusive parts of theprovince indicate derivation from highly heteroge-neous mantle material.

Elevated 87Sr/86Sr ratios in rocks from widelydispersed parts of the region are another distinc-

tive feature of the Caribbean plateau. In a Sr^Ndisotope diagram, the compositions of diverse rocktypes plot well to the right of the ¢eld de¢ned bythe Paci¢c mid-ocean ridge basalts (MORB) andoceanic-island basalts (OIB) (Fig. 1). Aitken andEcheverria [11] report unusually radiogenic Sr iso-topic compositions in komatiites and intrusiverocks from Gorgona Island (see also Table 1);Kerr et al. [3] and Hau¡ et al. [6] report similarvalues in basalts from CuracSao Island, mainlandColombia, the Nicoya Peninsula and in samplesfrom DSDP Leg 15 in the central Caribbean (Fig.1). These isotopic compositions have been attrib-uted to hydrothermal alteration [3,11], to the as-similation of altered oceanic crust [3], and morerecently to an inherent component of the plumesource [6]. These interpretations have very impor-tant implications for the nature and origin of theplume source composition; in particular, theybear on the question of whether this source con-tained a signi¢cant proportion of recycled oceaniccrust or whether the distinctive high-87Sr/86Sr geo-chemical signature was acquired by contamina-tion en route to the surface.

To gain a better understanding of this feature,we analysed the Sr, O and He isotopic composi-tions of magmatic clinopyroxene and olivine sep-arated from komatiites, picrites and ma¢c^ultra-ma¢c intrusive rocks from Gorgona Island.Analyses of these primary minerals provide val-uable information about the initial composition ofthe magma and allow us to evaluate the alterna-tive interpretations of the high-87Sr/86Sr compo-nent. In addition, these data provide new con-straints on the origin of the volcanic andintrusive rocks and on the composition of themantle source of the Caribbean volcanic province.

2. Geological background

The Caribbean Large Igneous Province coversmost of the £oor of the Caribbean basin as well asparts of the Cordillera of Colombia and Ecuador,a total area of about 6U105 km2 [12]. The crust in

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^10692

the Caribbean basin has a composite seismicstructure and varies in thickness from 5 to s 20km, mostly well in excess of that of ‘normal’ oce-anic crust [13,14]. Most authors accept that theprovince formed as an oceanic plateau throughmelting within the head of a large starting mantleplume [1]. Published radiometric ages indicate twoor three magmatic events: the ¢rst from 90 to 88Ma, the second about 76 Ma ago [4,8,15] and apossible third around 55 Ma ago [10]. It is com-monly thought that the plateau formed in the Pa-ci¢c, in part, perhaps, from a plume related to thecurrently active Galapagos hotspot [16^18].Through eastward movement of the Farallonplate, the northern part of the plateau was trans-ported between the two American plates, whilethe southern part was partially accreted aroundthe Caribbean and on the Paci¢c coast of Colom-bia [19].

Exposed, accreted parts of the Caribbean pla-teau are mostly composed of basaltic £ows, likeother oceanic plateaus. The Caribbean plateau isdistinguished from the Ontong Java, Kerguelenand other plateaus by the abundance of highMgO lavas such as the komatiites of GorgonaIsland and the picrites of CuracSao Island, main-land Colombia and Costa Rica. Plutonic com-plexes, which probably represent fossil magmachambers and parts of the plumbing that fed thelavas, are found in a few localities. Examples in-clude a sill complex on Be¤ata Ridge (south ofHispaniola) and layered ma¢c^ultrama¢c intru-sions on Gorgona Island and on mainland Co-lombia and Ecuador [9,10,15,20].

The petrographic, geochemical and isotopiccompositions of rocks from the Caribbean plateauare reported by numerous workers [3,6,9,10,15,20^24]. Wide variations in their petrologicaland geochemical characteristics indicate a highlyheterogeneous mantle source (Fig. 1). As seen inFig. 1, at least two distinct components canbe identi¢ed: a ‘depleted component’ with ONd

around +10, and a ‘more enriched’ componentwith ONd around +6 (Fig. 1). The former is mainlyrepresented by komatiites, picrites and certain ba-salts or gabbros from Gorgona Island, Ecuadorand Colombia [9,15,20,23]. The latter is found inpicrites and basalts from CuracSao Island [3,25], in

the Duarte complex in Haiti [7], in samples fromCosta Rica [26] and in the relatively enriched ba-salts from Gorgona Island (the ‘tholeiites’ as op-posed to the ‘komatiitic basalts’ of Echeverria andDupre¤ and Echeverria [23,27] or the ‘E-basalts’ asde¢ned by Kerr et al. [22]).

As seen in Fig. 1, a high proportion of samples,both volcanic and plutonic, have relatively high87Sr/86Sr ratios: they plot to the right (the high87Sr/86Sr side) of the ¢eld de¢ned by Sr and Ndisotopic compositions of Paci¢c MORB and OIB.To investigate this feature, Kerr et al. [3] andHau¡ et al. [6] conducted a series of leaching ex-periments on whole rocks. They found that acidleaching caused a displacement to lower 87Sr/86Srratios, but some strongly leached residues still hadhigh ratios compared to their Nd isotopic compo-sitions. They concluded that part of the high 87Sr/86Sr ratios measured in unleached whole-rocksamples was due to hydrothermal alteration, andthat the still-high ratios of some of the leachedsamples was a magmatic feature, perhaps the re-sult of assimilation of altered oceanic crust by themagmas parental to the erupted basalts.

3. Samples and analytical procedure

3.1. Whole-rock samples

The 87Sr/86Sr ratios reported in Table 1 weremeasured on unleached whole-rock powders. Wealso conducted a series of leaching experiments onsamples from which magmatic minerals were sep-arated for comparison with previous leaching ex-periments [3,6] and to help constrain the interpre-tation of the mineral data. The leaching procedurewas as follows: powders from whole-rock sampleswere leached for 1 h in hot 6 N HCl. The leachatewas decanted and analysed and then the residuewas totally dissolved in a mixture of 3:1 24 N HFand 12 N HNO3 (Table 1 and Fig. 2a,b). To testthe e⁄ciency of this leaching procedure, one ofthe samples (gabbro 94-15; Table 1, Fig. 2a,b)was leached twice. The ¢rst leaching followedthe procedure described above; the second en-tailed leaching for an additional hour in a mixtureof 1:1 6 N HCl and 24 N HF. Once again the

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106 93

leachate was decanted and analysed and the resi-due was totally dissolved in a mixture of 3:1 24 NHF and 12 N HNO3. 25% of the sample wasremoved in the ¢rst leaching step and 82% inthe second. The second step caused only a slightadditional decrease in 87Sr/86Sr, indicating thatthe one-step leaching procedure e¡ectively re-moved the labile high-87Sr/86Sr component fromthe whole rock (Fig. 2 and Table 1).

3.2. Mineral separation, and leaching procedurefor clinopyroxenes

Table 2 contains the locations and petrologicaland geochemical characteristics of the samples

from which minerals were separated. More com-plete information and analyses of whole rocks arefound in Re¤villon et al. [9]. We separated mag-matic clinopyroxene from four relatively coarse-grained volcanic rocks (three komatiites and apicrite) and four plutonic rocks (a dunite, a wehr-lite and two gabbros). Olivine was separated fromthree samples (a dunite, a wehrlite and a gabbro).The samples were chosen to represent the di¡erentlithologies on the island and their range of geo-chemical compositions. Two samples belong tothe ‘low-ONd group’ and six to the ‘high-ONd

group’ (Fig. 1b). The magmatic minerals wereseparated using a Franz-Isodynamic magneticseparator, followed by handpicking under a bin-ocular microscope. A special e¡ort was made toeliminate grains with inclusions or fragments ofsecondary minerals. The separated mineral frac-tion was then leached twice before total dissolu-tion using a four-step procedure adapted fromMachado et al. [28] (Table 1, Fig. 2b). The puri-¢ed mineral fraction was ¢rst leached for 30 minin cold 2 N HF and then for 10 min in hot 2.5 NHCl. The residue was dried and weighed. Thenthe residual sample was leached for 30 min incold 4 N HF followed by 30 min in hot 6 NHCl. Weight losses after the ¢rst and secondstages are reported in Table 1 and Fig. 2c. Cumu-lative weight losses vary from 6 to 19% after the¢rst step and from 16 to 35% after the second step(Fig. 2c), depending on the initial weight of min-erals. Almost no di¡erence was seen in the weightloss of the two leaching stages. Leachates fromeach step were decanted and analysed, and ¢nallythe residues were dissolved in a mixture of 3:124 N HF and 12 N HNO3.

3.3. Analytical techniques

For all measurements (whole rocks, mineralsand leachates) Sr was separated using 0.1 ml Sr-SPEC resin columns. Analyses were performed ona Finnigan MAT 262 at the University of Rennes.Blanks for Sr were 6 100 pg. A 87Sr/86Sr value of0.710251 W 9 was measured for the NBS 987 stan-dard.

For oxygen isotopic measurements, separatedclinopyroxenes were reacted using BrF5 following

Fig. 1. Measured 87Sr/86Sr versus initial 143Nd/144Nd ofunleached whole rocks from (a) the Caribbean Plateau and(b) Gorgona Island. Nd isotopic compositions were recalcu-lated at 88 Ma. Fields for Paci¢c MORB, Paci¢c OIB andGalapagos Islands are also reported. Error bars are smallerthan the size of the symbols. Sources of data: Galapagos[37], Haiti [8,38], Duarte [7], Costa Rica [5,26], Colombia[15], DSDP 15 [8], CuracSao [3,39], Ecuador [20]. Sources ofdata for MORB and OIB [40^50].

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^10694

Table 1Nd and Sr isotopic compositions of whole rocks (WR) and clinopyroxene (Cpx)

Sample Rock type ONd T Sr WR87Sr/86Sr

2c 87Sr/86SrWRleachate1

2c 87Sr/86SrWRleachate2

2c 87Sr/86SrWRresidue

2c 87Sr/86SrCpxleachate1

2c Weightloss 1

87Sr/86SrCpxleachate2

2c Weightloss 2

87Sr/86SrCpxresidue

2c Totalweightloss

N18O Cpx

(ppm)

GOR 501 Komatiite +9.82 57.7 0.702993 W 8GOR 502 Komatiite +9.69 79.1 0.703037 W 6 0.703458 W 6 0.702862 W 6 ^ 17 ^ 16 0.702714 W 7 33 5.37GOR 519 Komatiite +9.37 208 0.704976 W 6GOR 520 Komatiite +9.14 73.6 0.703840 W 5GOR 521 Komatiite +9.58 108 0.704157 W 6GOR 525 Komatiite +10.01 115 0.703491 W 5 0.703512 W 5 0.703455 W 8 0.703748 W 8 19 ^ 16 0.703348 W 7 35 5.35GOR 537 Komatiite +10.02 223 0.703467 W 6GOR 538 Komatiite +10.12 48.6 0.703061 W 8GOR 539 Komatiite +9.07 107 0.703550 W 8 0.703660 W 6 0.703463 W 6 0.703881 W 8 15 0.703712 W 7 7 0.702847 W 8 22 5.25GOR 94-07 Komatiite +8.77 51.0 0.704110 W 6GOR 94-08 Komatiite +8.23 1226 0.704274 W 7GOR 94-28 Komatiite +9.35 41.0 0.702977 W 7GOR 94-29 Komatiite +9.87 155 0.703451 W 6GOR 94-30 Komatiite +7.38 51.0 0.703291 W 6GOR 94-41 Komatiite +10.00 50.9 0.703164 W 6GOR 512 Picrite +10.44 30.1 0.704654 W 8GOR 513 Picrite +10.12 21.2 0.703259 W 8GOR 514 Picrite +9.31 44.2 0.704444 W 6GOR 515 Picrite +10.12 19.1 0.703671 W 8 0.703772 W 7 0.703765 W 7 ^ 10 ^ 7 0.703522 W 7 17 5.73GOR 94-35 Picrite +8.78 31.9 0.704654 W 6GOR 510 Gabbro +8.66 121 0.703471 W 5GOR 511 Gabbro +9.27 211 0.703471 W 4 0.704238 W 7 0.703487 W 5 0.705449 W 9 14 0.704100 W 20 6 0.702924 W 7 20 5.18GOR 534 Gabbro +6.71 113 0.703281 W 8GOR 541 Gabbro +10.08 107 0.703463 W 5GOR 94-05 Gabbro +9.78 83.0 0.703507 W 7GOR 94-13 Gabbro +9.89 134 0.703498 W 6GOR 94-13a Gabbro +9.86 134 0.703507 W 7GOR 94-14 Gabbro +9.40 137 0.703182 W 6GOR 94-15 Gabbro +9.39 79.6 0.703663 W 6 0.704172 W 7 0.703601 W 7GOR 94-15a Gabbro +9.39 79.6 0.703663 W 6 0.704183 W 8 0.703624 W 6 0.703596 W 8GOR 94-16 Gabbro +8.58 127 0.703472 W 6GOR 508 Ol-gabbro +8.89 115 0.703304 W 10GOR 509 Ol-gabbro +6.62 11.5 0.704229 W 10 0.704367 W 11 0.703997 W 5 0.704601 W 11 7 ^ 5 0.703498 W 8 21 5.22GOR 535 Ol-gabbro +7.59 50.9 0.703465 W 8GOR 503 Dunite +5.25 53.8 0.703747 W 8 0.703711 W 7 0.703683 W 8 0.703841 W 7 12 0.703678 W 14 9 0.702996 W 9 21 5.07GOR 505 Dunite +9.10 19.5 0.703733 W 10GOR 542 Dunite +9.48 25.4 0.703603 W 5GOR 506 Wehrlite +9.45 27.3 0.702985 W 6GOR 507 Wehrlite +8.94 5.96 0.703733 W 9 0.703850 W 6 0.703570 W 7 0.706330 W 31 6 0.703402 W 18 10 0.702729 W 7 16 5.06

Sr concentrations (ppm) were measured by isotope dilution except samples starting with GOR 94, which were measured by XRF. Nd isotope compositions arefrom [9]. Ol-gabbro=olivine gabbro.a Duplicate analysis. Weight losses are in %.

EPSL

644528-11-02

S.Re ¤villon

etal./E

arthand

Planetary

Science

Letters

205(2002)

91^10695

the method of Clayton and Mayeda [29]. Isotopicanalyses were performed on CO2 using a VGSIRA 10 triple-collector mass spectrometer atthe University of Rennes. Isotopic compositionsare quoted using the N notation with respect toSMOW. The internal reference MORB basalticglass material Circe¤ 93 yielded N

18O = +5.53xW 0.04 (average of 13 values). Analyses were nor-malised to NBS 28 (N18O= +9.60x) and basalticMORB Circe¤ 93 (recommended value N

18O=+5.68x). All analyses were repeated two ormore times and the reproducibility is better than0.1x.

Helium isotopic compositions were measuredon separated olivine at the Centre de RecherchePe¤trographique et Ge¤ochimique (CRPG) ofNancy. Helium was extracted by crushing of ol-ivine after cleaning in distilled water and high-

grade acetone. The mineral fractions were loadedin crushing tubes connected to a high-vacuumrare-gas analytical system and He isotopic com-positions were determined using a sequentialcrushing procedure, as described in Richard etal. [30]. This sequential procedure is used to dis-tinguish magmatic helium trapped in £uid inclu-sions from a potential radiogenic component pro-duced in the mineral matrix.

4. Results

4.1. Strontium isotopic compositions of wholerocks

Sr and Nd isotopic compositions of wholerocks from Gorgona Island are reported in Ta-

Fig. 2. (a) Sr isotopic compositions of leached and unleached samples from Gorgona Island, a basalt (CUR 38) from CuracSao Is-land and a basalt from DSDP Leg 15 Site 152. Data for CuracSao are from Kerr et al. [3] and DSDP Leg 15 Site 152 from Hau¡et al. [6]. (b) Sr isotopic compositions of Gorgona unleached and leached whole rocks with their associated leachates, and se-quentially leached clinopyroxene residues and leachates. (c) Weight losses (%) plotted versus 87Sr/86Sr of whole rock, 87Sr/86Sr ofleachates, and 87Sr/86Sr of clinopyroxenes. The lines indicate analyses from the same sample.

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^10696

ble 1. In Fig. 1, initial 143Nd/144Nd (recalculatedat 88 Ma) is plotted against measured 87Sr/86Sr.We do not report initial ratios for Sr because thevalues calculated using measured Rb/Sr ratios areunlikely to be correct. Because of the depletion ofincompatible elements, Rb/Sr in the unalteredmagmatic rocks (and in clinopyroxene) was lowand the age correction is minor (V0.000015 forsamples with the lowest 87Rb/86Sr). However, inother samples, Rb/Sr is higher, probably becauseof alteration, and the correction is signi¢cant (upto 0.000282). We do not know if the alterationtook place very soon after emplacement (in whichcase the calculated initial ratio would be correct)or some time after (in which case the initial ratiowould be wrong).

Measured 87Sr/86Sr values range between0.70498 and 0.70298. Initial 143Nd/144Nd ratiosvary between two end-member compositions:one highly depleted with ONdV+10, the otherless depleted with ONdV+6 (Fig. 1). These ratiosare comparable to those reported throughout theCaribbean plateau (Fig. 1a) although the mostdepleted component is largely restricted to Gor-gona Island. As in other parts of the province, theSr isotopic compositions of Gorgona Island sam-ples tend to be more radiogenic than expectedfrom their Nd isotopic compositions.

Sr isotopic compositions of leached whole rocksare reported in Figs. 2 and 3 and Table 1, togeth-er with the results of leaching experiments carriedout by Kerr et al. [3] and Hau¡ et al. [6] (Fig. 2a).The displacement to lower 87Sr/86Sr values is

clearly seen in all the leaching experiments, asare the higher ratios of the associated leachates(Table 1). The shifts toward lower values arelike those reported by Hau¡ et al. [6] (Fig. 2b)but ratios in the leached samples are lower thanthose of Kerr et al. [3]. This di¡erence is largelydue to the lower 87Sr/86Sr of the Gorgona sam-ples, but it does have an important bearing onthe interpretation of plume compositions, as dis-cussed below.

4.2. Strontium isotopic compositions ofclinopyroxenes

The compositions of leached clinopyroxenesand their associated leachates are reported in Ta-ble 1 and plotted in Figs. 2b,c and 3. The leachedclinopyroxenes have 87Sr/86Sr ratios that are sys-tematically lower than both their associated leach-ates (with values of 0.70633^0.70375, Table 1, Fig.2b), and the unleached and leached whole-rockcompositions (Table 1, Figs. 2 and 3).

Three clinopyroxene samples have signi¢cantlyhigher 87Sr/86Sr ratios than the others (Figs. 2 and3). This ‘high-87Sr/86Sr group’, which includes cli-nopyroxene from a komatiite (GOR 525), a pic-rite (GOR 515) and a gabbro (GOR 509), has87Sr/86Sr values from 0.70335 to 0.70352 (Fig. 3,Table 1). Two of these samples display high ONd

(GOR 515 and GOR 525); the other has relativelylow ONd (GOR 509). The ‘low-87Sr/86Sr group’comprises two komatiites, a gabbro, a duniteand a wehrlite, and has 87Sr/86Sr between

Table 2Main petrographical and geochemical characteristics of samples analysed in this study

Sample Rock type Location Texture Mineralogy Chemical features(%)

Ol Plgg CpxWglass MgO La/Sm Gd/Yb

GOR502 Komatiite Playa Pizarro Spinifex 50 25 20 14.53 0.296 1.266GOR525 Komatiite Camaronera Spinifex 50 25 20 20.11 0.291 1.263GOR539 Komatiite La Mancura Spinifex 50 25 20 20.98 0.284 1.316GOR515 Picrite Playa Blanca Porphyritic 50 10 40 27.29 0.203 0.720GOR511 Gabbro Gorgonilla Ophitic 50 45 8.14 0.389 1.261GOR509 Ol-gabbro Huanchinche Coarse-grained 25 35 35 27.20 0.340 0.714GOR507 Wehrlite Huanchinche Coarse-grained 70 6 5 25^30 29.63 0.217 1.034GOR503 Dunite Huanchinche Granular 90 5^10 35.41 0.280 1.024

More details can be found in [9]. Ol: olivine; Plg: plagioclase; Cpx: clinopyroxene. (La/Sm) and (Gd/Yb) are normalised toprimitive mantle [53].

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106 97

0.70271 and 0.70299. All, but one sample, havehigh ONd values (Fig. 3, Table 1).

4.3. Oxygen isotopic compositions ofclinopyroxenes

Oxygen isotopic compositions of clinopyrox-enes are listed in Table 1 and plotted in Fig. 4a.N

18O values range from +5.06x to +5.73x.Most samples have compositions in the rangeN

18O= 5.18^5.35x, except for low values of5.06^5.07x in clinopyroxene fractions from thetwo peridotites (GOR 503 and 507). Another ex-ception is the picrite (GOR 515), which has ahigher N

18O value of 5.73x (Fig. 4a, Table 1).Although no direct correlation can be establishedbetween Sr and O isotopic compositions, the sam-ples with high 87Sr/86Sr tend to have higher N

18Ovalues. An experimentally determined fractiona-tion factor between olivine and clinopyroxenewas used to compare our data with previouslyreported data on olivine from OIB and MORB[31]. A cpx^ol fractionation factor of +0.3x[32] was used to recalculate ¢elds reported by Ei-ler et al. [31]. The ¢ve samples with low 87Sr/86Srplot within the HIMU ¢eld rocks, whereasthe three samples with high 87Sr/86Sr ratios plotwithin or near the ‘high-3He/4He’ ¢eld [31](Fig. 4a,b).

4.4. Helium isotopic compositions of olivines

We analysed He isotopic compositions of oliv-ine from three samples: a dunite (GOR 503), awehrlite (GOR 507) and a gabbro (GOR 509). Allsamples gave relatively high R/Ra values (where Ris the measured ratio and Ra the atmospheric ra-tio; Table 3 and Fig. 4b), from 7.8 to 9.3 in thedunite, from 12.4 to 13.1 in the wehrlite and from13.6 to 18.2 in the gabbro (Table 3, Fig. 4b). Foreach sample the He isotopic composition did notchange during step crushing, except for GOR 509,which gave a signi¢cantly lower value for thehighest number of strokes (Table 3). This suggestsa contribution of radiogenic 4He from the matrix.A second aliquot of olivine was loaded (GOR509-bis) and the high R/Ra value of trapped magmaticgas was con¢rmed (18.2R/Ra). As seen in Fig. 4b,

the three samples plot outside the ¢elds de¢nedfor most OIB and MORB, but near the ¢eld ofIceland picrites and basalts. Particularly interest-ing is the high R/Ra value in the wehrlite (GOR507), which has low 87Sr/86Sr ratios and low con-centrations of incompatible trace elements.

Fig. 3. (a) 87Sr/86Sr versus initial 143Nd/144Nd of Gorgonaclinopyroxenes. Also plotted are representative ¢elds forwhole-rock samples, Paci¢c MORB and Paci¢c OIB. Sourcesof data are identical to Fig. 1. (b) 87Sr/86Sr versus initial143Nd/144Nd of Gorgona clinopyroxenes (black symbols),leached (open symbols) and unleached (grey symbols) wholerocks. The lines indicate clinopyroxene^whole-rock pairs. Er-ror bars are smaller than the data points. Also plotted are¢elds for each unleached whole-rock type (basalt, komatiite,picrite, gabbro, dunite and wehrlite). Source of data: [9].

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^10698

5. Discussion

5.1. Sr isotopic composition of the Caribbeanplateau

As seen in Fig. 1, the 87Sr/86Sr ratios ofunleached whole-rock samples from Gorgona Is-land are high compared to their Nd isotopic com-positions. These Sr isotopic ratios vary indepen-dently of rock type and chemical composition.Because these samples erupted in a submarine en-vironment [23,33], the most likely explanation fortheir radiogenic Sr isotopic compositions is inter-action with seawater. Our whole-rock leaching ex-periments con¢rm and reinforce the results ofKerr et al. [3] and Hau¡ et al. [6], who showedthat acid leaching removes a high-87Sr/86Sr com-ponent from the whole-rock samples and shiftsthe Sr isotopic composition of leached residuestoward lower values. However, as previouslynoted for some CuracSao samples [3,6] and sam-ples from Colombia and DSDP Leg 15 [6], 87Sr/86Sr ratios of leached samples remain high (Fig.3). This result indicates either that acid leachingdoes not remove all secondary Sr, or that theseelevated ratios are a magmatic feature. The leach-ing test conducted on a gabbro sample from Gor-gona (Table 1, Fig. 2) indicates, however, thateven very strong leaching, which dissolved morethan 80% of the sample, gave the same result asthe less aggressive procedure, and a higher 87Sr/86Sr ratio than clinopyroxene from the same sam-ples (Fig. 3b). It appears that even after strongleaching, Sr-rich secondary minerals such as epi-

dote or albite remained in the whole-rock sam-ples.

Magmatic clinopyroxenes provide more robustconstraints on the primary Sr isotopic composi-tion of the magmas and their mantle sources. In-terpretation of our results requires knowledgeof the crystallisation of komatiite magma. It isknown from experimental studies and the inter-pretation of textures that olivine and minorchrome spinel are the ¢rst phases to crystallisein Gorgona komatiites [3]. Clinopyroxene andplagioclase crystallise at lower temperatures, notlong before complete crystallisation of the mag-ma. Strontium is incompatible in olivine andspinel, and concentrates in the residual liquid. Ifthe komatiite had assimilated altered oceaniccrust, most of this assimilation probably tookplace before the crystallisation of clinopyroxene.Magmatic pyroxenes should therefore have Sr iso-topic compositions similar to that of the contam-inated magma. This is not the case, however, be-cause almost all the 87Sr/86Sr ratios determined inclinopyroxene fractions are signi¢cantly lowerthan those of the corresponding whole rocks(Fig. 3). This result provides strong support formodels in which the radiogenic Sr in the whole-rock samples is attributed to post-crystallisationalteration.

The Sr isotopic compositions of clinopyroxenesfrom the ‘low-87Sr/86Sr group’ of samples plotwell within the ‘normal’ ¢eld for plume-relatedmaterial (Fig. 3). For this set of samples, we canassume that the Sr isotopic compositions of theclinopyroxenes represent those of the parental

Table 3He isotopic composition of olivines from Gorgona samples

Sample Number of strokes Weight R/Ra 2c 4He(g) (mol/g)

GOR503 50 0.4206 7.77 W 0.18 5.37U10314

GOR503 100 0.4206 8.82 W 0.18 7.7U10314

GOR503 500 0.4206 9.34 W 0.16 2.1U10313

GOR507 50 0.4353 13.07 W 0.47 2.25U10314

GOR507 100 0.4353 12.54 W 0.6 2.77U10314

GOR507 500 0.4353 12.39 W 0.25 7.57U10314

GOR509 50 0.4315 14.73 W 0.98 1.01U10314

GOR509 100 0.4315 16.65 W 0.94 1.06U10314

GOR509bis 100 0.5045 18.21 W 0.7 1.94U10314

GOR509 500 0.4315 13.64 W 0.4 2.69U10314

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106 99

magmas and of their mantle source. On a globalscale, we can use our new data to rede¢ne the Srisotopic compositions of Caribbean plateau mag-mas. Only the three ‘high-87Sr/86Sr’ samples retainslightly anomalous compositions.

5.2. Variations in 87Sr/86Sr of clinopyroxenes

An explanation must now be found for theslightly elevated 87Sr/86Sr values (s 0.703) of thethree unusual samples (Fig. 3a,b). The di¡erenceis not large (87Sr/86Sr = 0.703348 vs 0.702996) butwell above analytical error (2c6 W 0.000009, Ta-ble 1).

For the following reasons we suggest that theslightly elevated ratios re£ect original magmaticcompositions. (a) Careful examination of the sep-arated minerals under a binocular microscope re-vealed no di¡erence between samples with high87Sr/86Sr and the other samples. The high-87Sr/86Sr clinopyroxenes are not more altered andthey contain no visible inclusions of secondaryminerals that could have carried a more radiogen-ic Sr signature. (b) All samples were processedusing an identical chemical procedure and analyt-ical method. Moreover, they were all prepared ina single batch in the same clean laboratory andwere run together on the mass spectrometer. (c)Weight losses after leaching are identical for thetwo groups (Table 1, Fig. 2c), ruling out the pos-sibility of bias in the leaching procedure.

The elevated 87Sr/86Sr of the three samples maybe due to (a) a relatively enriched component ina heterogeneous mantle source, or (b) a minoramount of assimilation of altered crust. The ¢rstexplanation has several implications. As can beseen from Fig. 3, samples with high Sr isotopiccompositions are found within both the ‘high-ONd’and ‘low-ONd’ groups of Gorgona rocks [9,21,22].If the high Sr isotopic ratios are a source feature,then Nd and Sr isotopes were decoupled in themantle source of Gorgona rocks. Moreover, asshown in Table 2, the high 87Sr/86Sr values arefound in three di¡erent rock types (a picrite, akomatiite and an olivine-gabbro), each of whichcame from liquids with di¡erent trace-elementcharacteristics [9].

Rather than immediately accepting such a com-

plicated situation, we evaluated the possibilitythat the isotopic compositions of the three anom-alous samples resulted from assimilation of asmall amount of altered oceanic crust (Table 4,Fig. 5). To do this, we assumed that the initialmagma had an isotopic composition like that ofthe low-87Sr/86Sr Gorgona clinopyroxenes andthat this magma had assimilated oceanic crust.To represent the contaminant, we used the Srand O isotopic compositions of the Trinity ophio-lite in California [34]. We chose this ophiolite be-cause its Sr and O isotopic compositions are well-documented and comparable to reported valuesfor extrusive and cumulate sections of the oceaniccrust ([35] and GERM database, http://earthref.org/GERM/main.htm). In this ophiolite, N

18Ovalues correlate with 87Sr/86Sr and range fromhigh values in the basalts to lower values in thecumulate layers. Two mixing models were used(Table 4, Fig. 5). In the ¢rst, cumulate layers ofoceanic crust were adopted as the contaminant.This material has a moderately high 87Sr/86Sr, aSr content of 100 ppm and N

18O slightly higherthan the average value of Gorgona rocks. In thesecond model, the contaminant was the alteredupper basaltic layer, which has a higher 87Sr/86Sr, 150 ppm Sr and a high N

18O. In both mod-els, the Gorgona end-member has an 87Sr/86Srratio of 0.7027, a value that corresponds to thelowest value measured in clinopyroxene separates.Given the mobility of Sr during alteration pro-cesses, we cannot use the whole rock Sr content.Instead we used a Sr content of 30 ppm, whichcorresponds to the average parental liquid com-position calculated for Gorgona Island rocks [9].The N

18O of this end-member is 5.2x. This mod-elling depends only slightly on the Sr content ofthe contaminant, as shown by the dashed line inFig. 5. This curve was calculated using a contam-inant with the same characteristics as the cumu-late layers but with a Sr content of 40 ppm in-stead of 100 ppm. From Fig. 5, it can be seen thatboth models ¢t the Sr and O isotopic data. Inmodel 1, the isotopic compositions of the high-87Sr/86Sr group are reproduced through the incor-poration of about 20% of material from cumulatelayers of the oceanic crust. In model 2, the samplewith high 87Sr/86Sr and high N

18O is reproduced

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106100

through contamination with about 10% of alteredbasalts. The e¡ect on Nd isotopic composition issmall, because the Nd isotopic compositions ofoceanic crust and Gorgona rocks are similar.

Despite the success of this part of the model-ling, the contamination hypothesis cannot explainall geochemical aspects of the samples. Both thekomatiites and picrites of Gorgona have very lowconcentrations of incompatible elements: V0.5ppm La and 1.15 ppm Sm in the komatiiteGOR 525, and 0.15 ppm La and 0.45 ppm Smin the picrite GOR 515. These values are only asmall fraction of the concentrations in basaltsfrom the oceanic crust (2.5^6 ppm La and 1^4ppm Sm). In Fig. 6, we plot the trace-elementcontents of Gorgona Island samples as well asestimates for the extrusive and cumulate parts ofthe oceanic crust. In the same manner as for Srand O isotopes, we calculated mixing curves be-tween samples and these end-members. In thismodelling, if the high-87Sr/86Sr group sampleshad been contaminated by oceanic crust (basaltsor cumulates), they should display distinctivetrace-element characteristics (i.e. higher La andLa/Sm). The potentially ‘contaminated’ samples(the three samples from the high-87Sr/86Sr group)are represented by open symbols whereas ‘non-contaminated’ samples (samples from the low-87Sr/86Sr group) are represented by ¢lled symbols.As seen in Fig. 6, two samples from the high-87Sr/86Sr group (GOR 525 and 515) plot very close tosamples from the other group. Contamination bya small amount of altered basalts should haveincreased both La contents and La/Sm ratios. Inthe same manner, the trace-element contents of

the high-87Sr/86Sr samples cannot be reproducedwith a contamination by cumulates. These threesamples cannot therefore have resulted from 10^20% contamination, the amount necessary to ex-plain the shift in Sr isotopic composition.

A further argument against contaminationcomes from the He isotopic compositions. Asshown in Fig. 4, the supposedly contaminatedsample GOR 509 from the high-87Sr/86Sr grouphas the highest R/Ra ratio of the three analysedsamples (Table 3). If the magma had been con-

Fig. 4. (a) 87Sr/86Sr versus N18O of Gorgona clinopyroxenes.

Representative ¢elds for MORB and samples with HIMU,EM1 and EM2 mantle source a⁄nity are also plotted. Fieldswere plotted using data from olivine phenocrysts [31]. Equiv-alent N

18O for clinopyroxenes were recalculated using a frac-tionation factor between olivine and clinopyroxene of 0.3[32]. (b) 87Sr/86Sr of clinopyroxene versus R/Ra from olivinein Gorgona samples. Fields of MORB, Iceland picrites,HIMU and high 3He/4He OIB are also reported [31,51,52].

Table 4End-member compositions used in the mixing modelling

Model 1 End-member 1(Gorgona)

End-member 2(cumulate section)

87Sr/86Sr 0.7027 0.7045Sr (ppm) 30 100N

18O 5.2 5.4

Model 2 End-member 1(Gorgona)

End-member 2(altered basalts)

87Sr/86Sr 0.7027 0.705Sr (ppm) 30 150N

18O 5.2 10

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106 101

taminated by material with high 87Sr/86Sr en routeto the surface, this process should have a¡ected itsHe isotopic composition and the samples shouldhave a lower R/Ra.

We conclude that of two processes that mayaccount for the elevated Sr isotopic compositionsof rocks from Gorgona Island (contamination ofmagma by altered oceanic crust and heteroge-neous mantle source), only the second is plausible.Such an interpretation implies that the mantlethat gave rise to the volcanic rocks of GorgonaIsland was highly heterogeneous, being composedof several isotopically distinct components. Thepresence of components with high and relativelylow ONd had been known from earlier studies, ashad variations in Pb isotopic compositions [27].Our work has demonstrated additional variationin Sr isotopic compositions.

5.3. Implication for the Caribbean LIP mantlesource

Three distinct groups of rocks can be identi¢ed

on Gorgona Island, using Nd and (clinopyroxene-derived) Sr isotopic compositions (Fig. 3). Thebasic distinction is made using the Nd isotopiccomposition, which distinguishes a high- and alow-ONd group. Our new data show that, withthe exception of two samples, the Sr isotopic com-positions of the Gorgona rocks fall in the ¢eld ofnormal Paci¢c MORB and OIB. Despite the smallsize of Gorgona Island, the isotopic characteris-tics of samples from the island cover a range sim-ilar to that from the rest of the province (Fig. 1).It is probable that if the Sr isotopic compositionsof clinopyroxenes from other parts of the plateauwere measured, they too would plot within orclose to the ¢elds of normal mantle-derived mag-mas. A possible exception is CuracSao, which iscomposed of rocks with distinctly higher 87Sr/86Sr ratios.

The high- and low-ONd groups correspond tothe ‘enriched’ and ‘depleted’ magma types recog-nised by Echeverria [9,21,23] from Gorgona Is-land and more recently by Hau¡ et al. [6] fromthroughout the Caribbean plateau. The latter au-thors have suggested that the two groups repre-sent di¡erent parts of old oceanic lithosphere that

Fig. 6. La versus La/Sm of Gorgona samples. The supposed‘contaminated’ samples are represented by open symbols andthe ‘non-contaminated’ samples by ¢lled symbols. The N-MORB value is from [53]; the cumulate value is from theGERM database (http://earthref.org/GERM/main.htm). Thecurved lines represent mixing between two non-contaminatedsamples and the two contaminants. Numbers indicateamount of contamination, in percent.

Fig. 5. 87Sr/86Sr versus N18O of Gorgona clinopyroxenes. The

two mixing curves were calculated using values listed in Ta-ble 4. The mixing model is only slightly dependent on the Srcontent of the contaminant. As an example, the dashed linerepresents a mixing curve calculated with 40 ppm Sr, whereasthe solid line is calculated with a contaminant containing 100ppm Sr. Only the percentage of assimilant changes, from20% for sample GOR 509 using 100 ppm Sr, to about 40%using 40 ppm Sr.

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106102

cycled through the mantle to reappear in theplume source of Caribbean magmas. The ‘en-riched’ component is equated with recycled upperoceanic crust, the ‘depleted’ component with low-er oceanic crust or oceanic lithosphere. By assum-ing that trace-element ratios in the two groups ofrocks were similar to those of their sources, Hau¡et al. [6] calculated that the time taken to recyclelithosphere through the mantle was between 300and 500 Ma. Although our new oxygen and Srisotopic data could be cited as additional evidencefor a model involving recycling of oceanic crust,other aspects of our data are less accommodating.We note, for example, that Sr isotopic composi-tions in the high-ONd group show signi¢cant var-iation (the separation into low and high 87Sr/86Srtypes). This behaviour is unlikely in the lower oce-anic crust or lithosphere that is less subject toalteration by hydrothermal £uid.

More telling is the behaviour of trace elementsin the Gorgona rocks, which are readily explainedin terms of fractionation during melting but notas the compositions of old, recycled components.In Fig. 7a we show that the D-basalts, komatiitesand picrites from Gorgona have a large range of147Sm/144Nd but similar 143Nd/144Nd. The ¢gurealso shows that 147Sm/144Nd ratios in komatiitesand picrites are far higher than in normal MORB.In Fig. 7b,c, the komatiites are seen to plot in twogroups, both with similar and high 147Sm/144Ndand 143Nd/144Nd, one with near-chondritic Gd/Yb, the other with signi¢cantly higher Gd/Yb.This complex decoupling of trace-element and iso-topic compositions is incompatible with a litho-sphere-recycling model but is readily explained interms of fractionation during melting [9,21]. Thehigh Sm/Nd ratios are a consequence of fraction-ation during melting; the di¡erences in Gd/Yb areexplained by the presence of garnet in the residueof melting for one set of komatiites, but not forthe other [9].

We do not reject the theory that the presence inthe source of components from di¡erent parts ofrecycled oceanic lithosphere contributed to theheterogeneities in the source of rocks from Gor-gona (and throughout the Caribbean plateau).Certainly, the di¡erence between the isotopically‘depleted’ and ‘enriched’ components must re£ect

Fig. 7. (a) 147Sm/144Nd versus 143Nd/144Ndi of Gorgona sam-ples. Also shown are ¢elds for MORB and Gorgona E-ba-salts. (b) 147Sm/144Nd versus Gd/Yb normalised to primitivemantle [53]. (c) Initial 143Nd/144Nd versus Gd/Yb normalisedto primitive mantle [53]. Dark grey ¢elds highlight the twogroups of komatiites. Source of data: Gorgona basalts[21,23], other Gorgona rocks [9], MORB as in Fig. 1.

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106 103

the presence of old materials with di¡erent com-positions, and these components could have comefrom recycled lithosphere. We know that this ma-terial resided in a reservoir with high 3He/4Heratios, probably the lower mantle. However, be-cause of the trace-element fractionation that ac-companied melting, we do not believe that the ageof this material can be calculated.

The high 3He/4He ratios measured in the oliv-ine fractions probably indicate that the source re-sided in the lower mantle. The high 3He/4He ra-tios measured in the two samples with depletedNd and Sr signatures (low 87Sr/86Sr and high143Nd/144Nd) illustrate again the well-known de-coupling of He from the other isotope systems. Italso provides further evidence that plumes containdepleted as well as enriched components. Thisconclusion has important implications for the in-terpretation of the high 3He/4He ratios measuredin Archean komatiites [30,36] because we havefew ¢rm constraints on the He isotopic composi-tion of the Archean upper mantle. The couplingof high 3He/4He ratios with depleted Nd and Srsignatures, in Gorgona komatiites as well as inIceland picrites (Fig. 3b), suggests that these iso-topic features are indeed characteristic of a mantleplume component.

6. Conclusions

The Sr isotopic compositions of most of theprimary clinopyroxenes from Gorgona Islandare signi¢cantly lower than those reported forwhole rocks from the Caribbean plateau. Evenstrongly leached whole-rock samples display ele-vated 87Sr/86Sr relative to their Nd isotopic com-positions and compared to the clinopyroxenecompositions. Post-emplacement alteration is thelikely explanation of the abnormally high 87Sr/86Sr signature, and this raises questions aboutthe e⁄ciency of acid leaching of whole-rock pow-ders as a means of removing the e¡ects of suchalteration.

The new clinopyroxene data shift the Sr iso-topic composition of the Caribbean plateau to-ward values of fresh volcanic rocks from oceanicsettings (MORB and OIB). However, three of the

analysed clinopyroxene mineral separates displayslightly higher 87Sr/86Sr values than the others. Acontamination model by assimilation of oceaniccrust is unable to explain the isotope and trace-element characteristics of these samples and weinterpret these ratios as a source feature. Theplume source of Gorgona rocks was heteroge-neous in terms of both its Nd and Sr isotopiccompositions. This heterogeneity of the plumesource is further indicated by oxygen isotopiccompositions. Oxygen isotopic compositions ofall the samples are relatively constant, with anaverage value of N

18O= 5.2x, which is lowerthan the N

18O= 5.7x currently reported in man-tle-related material. This result, together with thehigh 87Sr/86Sr in three samples, could indicate thepresence of recycled oceanic crust in the Caribbe-an plume. Finally, He isotopic ratios are veryhigh, even in samples with low 87Sr/86Sr valuesand high ONd. This may indicate that materialfrom which the magma originated once residedin the lower mantle.

All these results point toward a highly hetero-geneous mantle source for Gorgona Island andprobably for the rest of the Caribbean plateau.These heterogeneities can be related to the pres-ence of old recycled oceanic lithosphere in theplume source. However, the complex decouplingof trace-element and isotopic compositions inGorgona rocks favours fractionation during melt-ing.

Acknowledgements

The authors are grateful to FrancSois Se¤nebierfor help in mineral separation, and to NicoleMorin and Joe«l Mace¤ for technical support withSr isotopic analyses. The authors are also gratefulto Gustavo Garzon (Ministerio del Medio Am-biante, Cali, Colombia) for his assistance in ar-ranging permission to visit Gorgona, and to Clau-dia Isabel Acevedo and Gustavo Mayor for helpand logistic support in the ¢eld. Critical and con-structive reviews by D. Weis, A.C. Kerr and C.R.Neal are gratefully acknowledged. The FrenchCNRS supported the research through a grantin the IT programme to N.T.A.[AH]

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106104

References

[1] M.A. Richards, R.A. Duncan, V.E. Courtillot, Flood ba-salts and hot-spot tracks: plume heads and tails, Science246 (1989) 103^107.

[2] R.L. Larson, Latest pulse of Earth: Evidence for a mid-Cretaceous superplume, Geology 19 (1991) 547^550.

[3] A.C. Kerr, J. Tarney, G.F. Marriner, G.T. Klaver, A.D.Saunders, M.F. Thirlwall, The geochemistry and petro-genesis of the late Cretaceous picrites and basalts fromCuracSao, Netherlands Antilles: a remnant of an oceanicplateau, Contrib. Mineral. Petrol. 124 (1996) 29^43.

[4] A.C. Kerr, J. Tarney, A. Nivia, G.F. Marriner, A.D.Saunders, The internal structure of oceanic plateau infer-ences from obducted cretaceous terranes in western Co-lombia and the Caribbean, Tectonophysics 292 (1998)173^188.

[5] F.K. Hau¡, H. Hoernle, U. Schminke, R.A. Werner,Mid-Cretaceous origin for the Galapagos hotspot: volca-nological, petrological and geochemical evidence fromCosta Rican oceanic crustal fragments, Geol. Rundsch.86 (1997) 141^155.

[6] F. Hau¡, K. Hoernle, G. Tilton, D.W. Graham, A.C.Kerr, Large volume recycling of oceanic lithosphereover short time scale: geochemical constraints from theCaribbean Large Igneous Province, Earth Planet. Sci.Lett. 174 (2000) 247^263.

[7] H. Lapierre, V. Dupuis, B. Mercier de Le¤pinay, M.Tardy, J. Ruiz, R.C. Maury, J. Hernandez, M. Loubet,Is the lower Duarte igneous complex (Hispaniola) a rem-nant of the Caribbean plume-generated oceanic plateau?,J. Geol. 105 (1997) 111^120.

[8] C.W. Sinton, R.A. Duncan, M. Storey, J. Lewis, J.J. Es-trada, An oceanic £ood basalt province within the Carib-bean plate, Earth Planet. Sci. Lett. 155 (1998) 221^235.

[9] S. Re¤villon, N.T. Arndt, C. Chauvel, E. Hallot, Geochem-ical studies of ultrama¢c volcanic and plutonic rocks fromGorgona island, Colombia: the plumbing system of anoceanic plateau, J. Petrol. 41 (2000) 1127^1153.

[10] S. Re¤villon, E. Hallot, N.T. Arndt, C. Chauvel, R.A.Duncan, A complex history for the Caribbean plateau:petrology, geochemistry and geochronology of the Be¤ataridge, south Hispaniola, J. Geol. 108 (2000) 641^661.

[11] B.G. Aitken, L.M. Echeverria, Petrology and geochemis-try of komatiites and tholeiites from Gorgona Island, Co-lumbia, Contrib. Mineral. Petrol. 86 (1984) 94^105.

[12] M.F. Co⁄n, O. Eldholm, Large igneous provinces: crus-tal structure, dimensions, and external consequences, Rev.Geophys. 32 (1994) 1^36.

[13] S. Leroy, Structure et origine de la plaque Cara|«be. Im-plications ge¤odynamiques, Ph.D., University of Paris 6,1995.

[14] A. Mau¡ret, S. Leroy, Seismic stratigraphy and structureof the Caribbean igneous province, Tectonophysics 283(1997) 61^104.

[15] A.C. Kerr, G.F. Marriner, J. Tarney, A. Nivia, A.D.Saunders, M.F. Thirlwall, C.W. Sinton, Cretaceous basal-

tic terranes in Western Colombia: elemental, chronolog-ical and Sr-Nd isotopic constraints on petrogenesis,J. Petrol. 38 (1997) 677^702.

[16] R.A. Duncan, R.B. Hargraves, Plate tectonic evolution ofthe Caribbean region in the mantle reference frame, in:W.E. Bonini, R.B. Hargraves, R. Shagam (Eds.), The Ca-ribbean-South American Plate Boundary and RegionalTectonics, Geological Society of America Memoir 162,Boulder, CO, 1984, pp. 81^84.

[17] J.L. Pindell, S.C. Cande, W.C. PitmanIII, D.B. Rowley,J.F. Dewey, J. LaBrecque, W. Haxby, A plate-kinematicframework for models of Caribbean evolution, Tectono-physics 155 (1988) 121^138.

[18] J.L. Pindell, S.F. Barrett, Geological evolution of the Ca-ribbean region; a plate tectonic perspective, in: G. Dengo,J.E. Case (Eds.), The Caribbean Region, The Geology ofNorth America H, Geological Society of America, Bould-er, CO, 1990, pp. 405^432.

[19] K. Burke, Tectonic evolution of the Caribbean, Ann. Rev.Earth Planet. Sci. 16 (1988) 201^230.

[20] H. Lapierre, D. Bosch, V. Dupuis, M. Polve¤, R.C. Maury,J. Hernandez, P. Monie¤, D. Yeghicheyan, E. Jaillard, M.Tardy, B. Mercier de Le¤pinay, M. Mamberti, A. Desmet,F. Keller, F. Se¤nebier, Multiple plume events in the gen-esis of the peri-Caribbean Cretaceous oceanic plateauprovince, J. Geophys. Res. 105 (2000) 8403^8421.

[21] N.T. Arndt, A.C. Kerr, J. Tarney, Dynamic melting inplume heads: the formation of Gorgona komatiites andbasalts, Earth Planet. Sci. Lett. 146 (1997) 289^301.

[22] A.C. Kerr, G.F. Marriner, N.T. Arndt, J. Tarney, A.Nivia, A.D. Saunders, R. Duncan, The petrogenesis ofGorgona komatiites, picrites and basalts: New ¢eld, pet-rographic and geochemical constraints, Lithos 37 (1996)245^260.

[23] L.M. Echeverria, Tertiary or Mesozoic komatiites fromGorgona Island, Columbia: Field relations and geochem-istry, Contrib. Mineral. Petrol. 73 (1980) 253^266.

[24] L.M. Echeverria, B.G. Aitken, Pyroclastic rocks: anothermanifestation of ultrama¢c volcanism on Gorgona Island,Columbia, Contrib. Mineral. Petrol. 92 (1986) 428^436.

[25] G.T. Klaver, The Curacao lava formation: an ophioliticanalogue of the anomalously thick layer 2B of the mid-Cretaceous oceanic plateau in the western Paci¢c and cen-tral Caribbean, Ph.D., University of Amsterdam, 1987.

[26] C.W. Sinton, R.A. Duncan, P. Denyer, Nicoya Peninsula,Costa Rica, a single suite of Caribbean oceanic plateau,J. Geophy. Res. 102 (1997) 15507^15520.

[27] B. Dupre¤, L.M. Echeverria, Pb isotopes of Gorgona is-land (Colombia): isotopic variations correlated with mag-ma type, Earth Planet. Sci. Lett. 67 (1984) 186^190.

[28] N. Machado, C. Brooks, S.R. Hart, Determination ofinitial 87Sr/86Sr and 143Nd/144Nd in primary mineralsfrom ultrama¢c rocks: experimental procedure and impli-cations for the isotopic characteristics of the Archeanmantle under the Abitibi greenstone belt, Canada, Geo-chim. Cosmochim. Acta 50 (1986) 2335^2348.

[29] R.N. Clayton, T.K. Mayeda, The use of bromine penta-

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106 105

£uorine in the extraction of oxygen from oxydes and sil-icates for isotopic analysis, Geochim. Cosmochim. Acta27 (1963) 43^52.

[30] D. Richard, B. Marty, M. Chaussidon, N.T. Arndt, He-lium isotopic evidence for a lower mantle component indepleted archean komatiite, Science 273 (1996) 93^95.

[31] J.M. Eiler, K.A. Farley, J.W. Valley, E. Hauri, H. Craig,S.R. Hart, E.M. Stolper, Oxygen isotope variations inocean island basalt phenocrysts, Geochim. Cosmochim.Acta 61 (1997) 2281^2293.

[32] Y.F. Zheng, Calculation of oxygen isotope fractionationin anhydrous silicate minerals, Geochim. Cosmochim.Acta 57 (1993) 1079^1091.

[33] A. Gansser, V.J. Dietrich, W.E. Cameron, Paleogene ko-matiites from Gorgona island, Nature 278 (1979) 545^546.

[34] C. Le¤cuyer, S. Fourcade, Oxygen isotope evidence for amulti-stage hydrothermal alteration at a fossil slow-spreading center: the Silurian Trinity ophiolite (Califor-nia, USA), Chem. Geol. 87 (1991) 231^246.

[35] H. Staudigel, T. Plank, W. White, H.U. Schmincke, Geo-chemical £uxes during sea£oor alteration and the basalticupper oceanic crust: DSDP Sites 417 and 418, in: G.E.Bebout, D.W. Scholl, S.H. Kirby, J.P. Platt (Eds.), Sub-duction Top to Bottom, AGU Monograph 96, 1996.

[36] T. Matsumoto, A. Seta, J. Matsuda, M. Takebe, Y. Chen,S. Arai, Helium in the Archean komatiites revisited: sig-ni¢cantly high 3He/4He ratios revealed by fractionalcrushing gas extraction, Earth Planet. Sci. Lett. 196(2002) 213^225.

[37] W.M. White, A.R. McBirney, R.A. Duncan, Petrologyand geochemistry of the Gala¤pagos Islands: portrait ofa pathological mantle plume, J. Geophys. Res. 98 (1993)19533^19563.

[38] G. Sen, R. Hickey-Vargas, D.G. Waggoner, F. Maur-rasse, Geochemistry of basalts from the Dumisseau For-mation, southern Haiti: implications for the origin of theCaribbean Sea crust, Earth Planet. Sci. Lett. 87 (1988)423^437.

[39] S. Re¤villon, N.T. Arndt, E. Hallot, A.C. Kerr, J. Tarney,Petrogenesis of picrites from the Caribbean plateau andthe North Atlantic magmatic province, Lithos 49 (1999)1^21.

[40] R.S. Cohen, N.M. Evensen, P.J. Hamilton, R.K.O’Nions, U-Pb, Sm-Nd and Rb-Sr systematics of mid-ocean ridge basalt glasses, Nature 283 (1980) 149^153.

[41] R.S. Cohen, R.K. O’Nions, The lead, neodymium andstrontium isotopic structure of ocean ridge basalts, J. Pet-rol. 23 (1982) 299^324.

[42] J.D. MacDougall, G.W. Lugmair, Sr and Nd isotopes inbasalts from the East Paci¢c Rise: signi¢cance for mantleheterogeneity, Earth Planet. Sci. Lett. 77 (1986) 273^284.

[43] W.M. White, A.W. Hofmann, H. Puchelt, Isotope geo-chemistry of Paci¢c mid-ocean ridge basalts, J. Geophys.Res. 92 (1987) 4881^4893.

[44] C.W. Devey, F. Albarede, J-L. Chemine¤e, A. Michard, R.Mu«he, P. Sto¡ers, Active submarine volcanism on theSociety hotspot swell (West Paci¢c): a geochemical study,J. Geophys. Res. 95 (1990) 5049^5066.

[45] R.A. Duncan, M.T. McCulloch, H.G. Barczus, D.R. Nel-son, Plume versus lithospheric sources for the melts at UaPou, Marquesas Island, Nature 303 (1986) 142^146.

[46] C. Dupuy, P. Vidal, H.G. Barsczus, C. Chauvel, Origin ofbasalts from the Marquesas Archipelago (south centralPaci¢c Ocean): isotope and trace element constraints,Earth Planet. Sci. Lett. 82 (1987) 145^152.

[47] D.C. Gerlach, J.C.J. Stormer, P.A. Mueller, Isotopic geo-chemistry of Fernado de Noronha, Earth Planet. Sci.Lett. 85 (1987) 129^144.

[48] Y. Nakamura, M. Tatsumoto, Pb, Nd, and Sr isotopicevidence for a multicomponent source for rocks ofCook-Austral Islands and heterogeneities of mantleplumes, Geochim. Cosmochim. Acta 52 (1988) 2909^2924.

[49] Z.A. Palacz, A.D. Saunders, Coupled trace element andisotope enrichment in the Cook-Austral-Samoa islands,Southwest Paci¢c, Earth Planet. Sci. Lett. 79 (1986)270^280.

[50] W.M. White, A.W. Hofmann, Sr and Nd isotope geo-chemistry of oceanic basalts and mantle geochemistry,Nature 296 (1982) 821^825.

[51] D. Harrison, P. Burnard, G. Turner, Noble gas behaviourand composition in the manle: constraints from the Ice-land plume, Earth Planet. Sci. Lett. 171 (1999) 199^207.

[52] D.R. Hilton, M.F. Thirlwall, R.N. Taylor, B.J. Murton,A. Nichols, Controls on magmatic degassing along theReykjanes Ridge with implications for the helium para-dox, Earth Planet. Sci. Let. 183 (2000) 43^50.

[53] A.W. Hofmann, Chemical di¡erentiation of the Earth:the relationship between mantle, continental crust, andoceanic crust, Earth Planet. Sci. Lett. 90 (1988) 297^314.

EPSL 6445 28-11-02

S. Re¤villon et al. / Earth and Planetary Science Letters 205 (2002) 91^106106