The Depleted Mantle Component in Kerguelen Archipelago Basalts: Petrogenesis of Tholeiitic Transitional Basalts From the Loranchet Peninsula

Transcription

1 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 PAGES The Depleted Mantle Component in Kerguelen Archipelago Basalts: Petrogenesis of Tholeiitic Transitional Basalts From the Loranchet Peninsula SONIA DOUCET 1,3, DOMINIQUE WEIS 1, JAMES S. SCOATES 1, KIRSTEN NICOLAYSEN 2, FREDERICK A. FREY 2 AND ANDRÉ GIRET 3 1 DÉPARTEMENT DES SCIENCES DE LA TERRE ET DE L ENVIRONNEMENT, UNIVERSITÉ LIBRE DE BRUXELLES, B-1050 BRUSSELS, BELGIUM 2 DEPARTMENT OF EARTH, ATMOSPHERIC AND PLANETARY SCIENCES, MASSACHUSETTS INSTITUTE OF TECHNOLOGY, CAMBRIDGE, MA 02139, USA 3 LABORATOIRE DE GÉOLOGIE PÉTROLOGIE, CNRS UMR 6524, UNIVERSITÉ JEAN MONNET, SAINT-ETIENNE, FRANCE RECEIVED JUNE 20, 2001; REVISED TYPESCRIPT ACCEPTED FEBRUARY 19, 2002 KEY WORDS: Kerguelen Archipelago; basalt; geochemistry; depleted com- ponent; Kerguelen plume; mixing A geochemical study of 28 Ma tholeiitic to transitional basalts from the Kerguelen Archipelago ( Mont des Ruches and Mont Fontaine) indicates that three distinct magma types erupted within >1 Myr. Low-MgO basalts (>4 6 wt %) in both sections are overlain by high-mgo basalts (>7 13 wt %), mostly present in Mont Fontaine. Both high- and low-mgo basalts have nearly identical low 87 Sr/ 86 Sr and high 143 Nd/ 144 Nd and formed from similar parental magmas that represent mixtures between a depleted mantle component and the Kerguelen plume. The third magma type, predominant in Mont des Ruches, is represented by high-mgo basalts that are isotopically heterogeneous with isotopic ratios that are intermediate between those of the stratigraphically lower basalts and the Kerguelen plume compositions; this third magma type may have formed by mixing of similar material, but with a higher contribution from the Kerguelen plume. The depleted component involved in all three magma types is similar to the source for Southeast Indian Ridge basalts and is present in Kerguelen Archipelago basalts older than 26 Ma, which erupted when the ridge axis was <500 km away from the Kerguelen hotspot. Depleted heterogeneities intrinsic to the plume and entrainment of depleted mantle during plume ascent do not explain the marked cut-off in the presence of a depleted component in the archipelago basalts within a time interval of 1 Myr after 26 Ma. Mixing of depleted asthenosphere with the plume in sublithospheric channels during migration of the Southeast Indian Ridge axis away from the Kerguelen hotspot is proposed as a suitable explanation to account for the temporal distribution of the depleted component in basalts from the Northern Kerguelen Plateau and the >26 Ma Kerguelen Archipelago flood basalts; cessation of plume ridge interactions may explain the absence of depleted basalts in the youngest sections that erupted further away from the ridge axis. INTRODUCTION The generation of the >119 Ma to present Kerguelen Large Igneous Province (Duncan, 2002) is related to activity of the Kerguelen mantle plume (e.g. Frey et al., 2000a). The Kerguelen Archipelago and part of the Corresponding author. Present address: Département des Sciences de la Terre et de l Environnement, Université Libre de Bruxelles, B-1050 Brussels, Belgium. Telephone: Fax: sdoucet@ulb.ac.be Present address: Department of Geology, Kansas State University, Manhattan, KS 66506, USA Oxford University Press 2002

2 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 submarine Northern Kerguelen Plateau (Weis & Frey, GEOLOGY OF THE KERGUELEN 2002) represent the last >35 Myr of plume activity. Within this 35 Myr time interval, there is evidence for a ARCHIPELAGO AND THE decreasing degree of melting with decreasing eruption LORANCHET PENINSULA age (Weis et al., 1998b; Frey et al., 2000b). In an effort to The Kerguelen Archipelago (6500 km 2 ) is the emergent evaluate the source components and the petrogenetic part of the Northern Kerguelen Plateau ( Fig. 1). Spreadprocesses responsible for the various isotopic signatures ing rates on the SEIR (Royer & Sandwell, 1989) indicate observed in basalts from the Kerguelen Archipelago, that the archipelago location was along the ridge axis at systematic studies of the spatial and temporal geochemical 40 1 Ma and subsequently evolved toward its present variations in archipelago basalt sections have been undermoved intraplate location, 1400 km to the SW, as the ridge taken (Weis et al., 1998b; Yang et al., 1998; Frey et al., to the NE. Flood basalts (>29 24 Ma; Nicolaysen 2000b, 2002a). These studies complement earlier survey et al., 2000) cover 85% of the surface of the archipelago studies of basalts from diverse locations on the Kerguelen ( Fig. 1a). During emplacement of the flood basalts, the Archipelago (Gautier et al., 1990; Weis et al., 1993). distance between the archipelago and the SEIR increased The proportion of depleted mantle contributing to from >350 to >550 km. The topography of the nearly Kerguelen Archipelago flood basalts has been a matter horizontal flood basalts is locally perturbed by younger of debate (Storey et al., 1988, 1989; Gautier et al., 1990; volcanic plutonic complexes (Giret & Lameyre, 1983; Weis et al., 1993; Yang et al., 1998). The relatively Weis & Giret, 1994). Basaltic sections up to 1000 m in low ( 87 Sr/ 86 Sr) i (down to , where the subscript height (Fig. 1a) have been exposed by glacial erosion i indicates the age-corrected isotopic ratio for each and were systematically sampled during the last 10 years individual section), ( 206 Pb/ 204 Pb) i (down to 18 2) and high of the Kerguelen Archipelago mapping programme ( 143 Nd/ 144 Nd) i (up to >0 5129) in >15% of the exposed (CartoKer programme, Université Jean Monnet). tholeiitic to transitional basalts from the north central Flood basalts cover the entirety of the Loranchet Pen Ma Mont Bureau and 29 Ma Mont Rabouillère insula (Fig. 1) and are exposed in sections of m sections have been interpreted as reflecting assimilation height. NW SE and NE SW oriented fjords and near- of gabbroic oceanic crust (Yang et al., 1998). Mixing vertical dykes cut the basalts and follow the two main between Southeast Indian Ridge (SEIR) and Kerguelen fracture orientations reported on the peninsula (Nougier, plume magmas is documented at Ocean Drilling Program 1970). Dykes and sills of trachyte and rhyolite, and (ODP) Site 1140 (>34 Ma; Duncan, 2002) in the minor gabbroic intrusions also occur on the peninsula. Plagioclase-phyric massive basalt flows, preferentially loc- Northern Kerguelen Plateau, >300 km north of the ated in the lower parts of the volcanic sections, and archipelago (Weis & Frey, 2002). Finally, the nearly olivine-phyric flows are common. The Mont des Ruches isotopically homogeneous basalts from the 26 Ma Mont and Mont Fontaine sections are composed of numerous Tourmente section in the central Kerguelen Archipelago massive basaltic flows of 2 18 m thickness, which are have lower 87 Sr/ 86 Sr and higher 143 Nd/ 144 Nd compared separated by weathered or oxidized horizons ( Fig. 1b). with estimates for the plume (Weis et al., 1998a, 2002); Mont des Ruches is 497 m high and is cut by a dyke of Frey et al. (2002a) have suggested that these basalts may 15 m thickness. A sample of this dyke was not available reflect depleted heterogeneities intrinsic to the Kerguelen for this study but it is described as a zoned dyke with a plume. gabbroic core and a rhyolitic border. Additional field We have studied the Mont des Ruches and Mont descriptions (P. Camps & M. Perrin, personal com- Fontaine basaltic sections, which are located in the north- munication) indicate the presence of numerous sills in western part of the Kerguelen Archipelago on the Lor- the area. Mont Fontaine is a succession of subhorizontal anchet Peninsula ( Fig. 1). These lavas are considered to basaltic flows forming a section of 325 m height, located be amongst the oldest sections exposed on the archipelago >15 km to the east of Mont des Ruches. There are no as a result of a regional 2 5 SE dip of basaltic flows, field relationships that allow us to place relative age which correlates with a general decrease in eruption ages constraints between the two stratigraphic sections. from NW to SE in the archipelago (Nicolaysen et al., 2000). In this paper we use the age and geochemical characteristics of basalts in both sections to (1) constrain the composition and origin of the depleted component, PETROGRAPHY AND MINERAL (2) constrain interactions between the plume and the CHEMISTRY depleted component reservoirs, and (3) evaluate the con- More than 80% of the examined samples from Mont tribution, if any, of contamination by continental material des Ruches and Mont Fontaine are either plagioclasein the petrogenesis of basaltic lavas from the NW Ker- phyric (four of 46 samples) or olivine-phyric (34 of 46 guelen Archipelago. samples). Phenocryst contents and sizes range from 3 to 1342

3 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Fig. 1. (a) Simplified geological map of the Kerguelen Archipelago, after Nougier (1970), showing the distribution of flood basalts (85% of the surface area), plutonic complexes (5%) and Quaternary deposits (10%). Χ, locations of basaltic sections sampled during the CartoKer mapping programme. (b) The Mont des Ruches and Mont Fontaine sections with sample locations shown (black areas indicate massive parts of flows where sampling was possible). All sample names studied in this paper are prefixed by BY96-. (c) Schematic map of the Indian Ocean sea floor, after Smith & Sandwell (1997). 15 vol. % and >0 3 to 5 mm, respectively ( Table 1). mineral stoichiometry. Representative compositions are The groundmass is doleritic or subophitic and consists given in Table 2. of fine-grained or poikilitic augite, plagioclase laths, oxides (<1 vol. %), olivine, and locally devitrified and/or altered brown glass. Alteration is present in >40% of the Mont Plagioclase-phyric (low-mgo) basalts des Ruches and Mont Fontaine samples (Table 1), and Plagioclase phenocrysts have a nearly constant size of ranges from brown altered and devitrified groundmass, >5 mm. They are typically characterized by a core of to serpentinized and iddingsitized olivine, to secondary An 70 80, a strongly resorbed border and a narrow fringe zeolites (<1 vol. % when present) and rare calcite (<1%, (up to 0 1 mm) of >An (Table 2). They are commonly when present, e.g. sample BY96-31). Alteration is also isolated in a fine-grained groundmass of augite, plagioreflected in variable whole-rock loss on ignition (LOI) clase and altered devitrified glass. Zoning can locally values from 1 2 to 6 7 wt %. be complex: a single phenocryst in BY96-44 exhibits Olivine and plagioclase compositions were analysed oscillatory zoning between An 35 and An 80 around a homowith a Cameca SX100 electon microprobe at the Unigeneous core of An 80. versité Blaise Pascal, Clermont Ferrand, France, using an accelerating voltage of 15 kev, a beam current of 15 na, a beam size of 1 μm, and natural and synthetic standards. Compositions were determined on cores and Olivine-phyric (high-mgo) basalts rims of multiple grains from selected thin sections. Both Olivine is abundant in the high-mgo basalts from the the plagioclase and olivine analyses are consistent with Mont des Ruches and Mont Fontaine sections (Table 1) 1343

4 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Table 1: Petrographic characteristics of Mont des Ruches and Mont Fontaine samples Sample Height Description Groundmass Phenocrysts MgO mg-no. Comment (m) (%) (wt %) Mont des Ruches BY ol-phyric subophitic ol (15) fresh BY ol-phyric subophitic ol (13) little altered BY ol-phyric doleritic fresh BY ol-phyric doleritic ol (12) fresh BY aphyric quenched very altered BY ol-phyric doleritic ol (13) little altered BY ol-phyric doleritic ol (8) little altered BY aphyric microlitic very altered BY ol-phyric doleritic ol (15) fresh BY ol-phyric doleritic ol (5) altered olivine BY ol-phyric doleritic ol (11) fresh BY ol-phyric subophitic ol (5) fresh BY ol-phyric doleritic ol (5) fresh BY ol-phyric doleritic ol (5) fresh BY ol-phyric subophitic fresh BY ol-phyric subophitic ol (5) very altered BY aphyric medium-grained very altered BY aphyric medium-grained very altered BY aphyric medium-grained very altered BY pl-phyric fine-grained pl (15) very altered BY pl-phyric fine-grained pl (16) little altered BY pl-phyric fine-grained pl (10) little altered BY aphyric trachytic fresh Mont Fontaine BY ol-phyric doleritic ol (10) fresh BY aphyric subophitic very altered BY ol-phyric subophitic ol (2) fresh BY ol-phyric subophitic ol (10) fresh BY ol-phyric subophitic ol (10) fresh BY ol-phyric subophitic ol (11) fresh BY ol-phyric subophitic ol (12) fresh BY ol-phyric subophitic ol (5) very altered BY ol-phyric subophitic ol (7) very altered BY ol-phyric subophitic ol (11) altered BY ol-phyric doleritic ol (9) fresh BY ol-phyric doleritic ol (6 5) fresh BY ol-phyric doleritic ol (4) very altered BY ol-phyric doleritic ol (3) very altered BY ol-phyric doleritic ol (3) very altered BY ol-phyric doleritic ol (10) very altered BY ol-phyric subophitic ol (15) very altered BY ol-phyric doleritic ol (5) altered BY ol-phyric doleritic ol (13) little altered BY ol-phyric subophitic ol (13) very altered BY ol-phyric doleritic ol (9 5) fresh BY pl cpx-phyric doleritic cpx (2 5), pl (2) fresh BY pl-phyric microlitic pl (10) little altered ol, olivine; pl, plagioclase; cpx, clinopyroxene. mg-number is calculated on the basis of atomic fractions [Mg 2+ /(Mg 2+ +Fe 2+ )]. 1344

5 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Table 2: Representative microprobe analyses of olivine and plagioclase compositions of Mont des Ruches and Mont Fontaine basalts Mont des Ruches olivines Sample: BY96-25 BY96-25 BY96-25 BY96-29 BY96-29 BY96-30 BY96-30 BY96-32 BY96-32 core rim GM core rim core rim core rim SiO FeO MgO Cr 2 O MnO CaO NiO Total Fa (%) Fo (%) Mont des Ruches plagioclase Sample: BY96-25 BY96-29 BY96-30 BY96-32 BY96-37 BY96-37 BY96-44 BY96-44 BY96-44 GM GM GM GM GM GM core rim GM SiO Al 2 O FeO MgO CaO Na 2 O K 2 O Total An (%) Ab (%) Or (%)

6 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Table 2: continued Mont Fontaine olivines Sample: BY96-83 BY96-83 BY96-83 BY96-86 BY96-86 BY96-90 BY96-90 BY96-90 BY96-98 BY96-98 BY BY core core rim core rim core rim GM core rim core rim SiO FeO MgO Cr 2 O MnO CaO NiO Total Fa (%) Fo (%) Mont Fontaine plagioclase Sample: BY96-83 BY96-86 BY96-86 BY96-90 BY BY BY BY GM GM in. cpx GM GM core rim GM SiO Al 2 O FeO MgO CaO Na 2 O K 2 O Total An (%) Ab (%) Or (%) GM signifies olivine or plagioclase composition in the groundmass. Fa, Fo, An, Ab and Or means molar fayalite, forsterite, anorthite, albite and orthoclase percent proportions within the mineral. 1346

7 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Table 3: Summary of 40 Ar/ 39 Ar geochronology for Monts des Ruches and Fontaine Sample Locality Plateau 39 Ar p (%) Isochron Cumulative ( 40 Ar/ 36 Ar) i MSWD age (Ma) age (Ma) % 39 Ar BY96-24 Mt. des Ruches 27 95± ± ± BY96-80 Mt. Fontaine 27 92± ± ± BY Mt. Fontaine 27 4± ± ± The plateau ages are defined by at least three consecutive heating steps which overlap within the 2σ uncertainties and the plateau includes at least 50% of the total 39 Ar released (Fleck et al., 1977). All uncertainties are reported at the 2σ confidence limit and the isochron dates are used for the eruption ages discussed in the text. MSWD, mean square weighted deviation. 39 Ar p is the percentage of total K-derived 39 Ar evolved from heating steps on the plateau. and is of variable size (up to 3 mm diameter), abundance concentrations, were determined by inductively coupled and habit. Skeletal or euhedral crystals may occur as plasma mass spectrometry (ICP-MS) on an HP4500 isolated phenocrysts in a fine- or medium-grained do- system at the Katholieke Universiteit Leuven, Belgium. leritic or subophitic groundmass. Rare rounded olivine Samples were digested in sub-boiled 22M HF and 14M grains are observed. Olivine grains have core and rim HNO 3 then dissolved in 10 ml of sub-boiled 14M HNO 3. compositions of Fo> and Fo> 75 50, respectively After digestion, In (6 ppm), Tl (3 25 ppm), Re (8 6 ppm) ( Table 2). One sample (BY96-37) contains an inclusion and Rh (9 8 ppm) were added as internal standards. Li, of fine-grained gabbro. Y, Ce and Tl were used for external HP4500 calibration. One blank was measured for each set of five samples and was used to correct the results for contamination during the dissolution. Except for U, Th, Pb and Ta, ANALYTICAL TECHNIQUES whose blank contribution can exceed 20% of the total A total of 43 samples ( cm 3 ) were analysed for amount in the sample, all blank contributions are major and trace element compositions. Surface alteration p10%, with the majority of blank contributions ranging was removed by a diamond-embedded saw, and then between 0 01 and 2% of the amount in the sample. the cut surface was abraded using sandpaper to remove Repeated measurement of the BE-N (basalt) standard any saw traces and remaining alteration features. The (Geostandards Newsletter, 1995) was used to calculate trace samples were coarsely crushed in a hydraulic piston element concentrations in the samples. The precision of crusher (percussion method) before being reduced to the measurements was controlled by replicate measpowder in an agate shatterbox. Two samples from the Mont Fontaine stratigraphic urements (eight times, on a separately dissolved sample) section and one from Mont des Ruches were selected of a basalt from ODP Leg 183, Site 1140 (31R1- for dating by the 40 Ar/ 39 Ar method. The results are cm; Weis & Frey, 2002; Table 5). summarized in Table 3. The relative paucity of plagioselected to encompass the entire range of compositions Samples for Pb, Sr and Nd isotopic analyses were clase phenocrysts within these basalts required the anaof the least altered samples. The chemical procedure lysis of whole-rock groundmass. Each sample was crushed to Ζ1 mm in diameter and sieved to a mm size used is that described by Weis & Frey (1991). Samples fraction. Using a Frantz magnetic separator, the olivine were acid-leached in HCl 6N (6 8 steps) to remove and clinopyroxene were removed to minimize excess alteration phases. The weight lost by leaching was be- magmatic argon. The samples were acid-leached to refrom separate powders) were also analysed and their tween 28 and 58%. Two complete duplicates (leached move zeolite alteration phases, irradiated, and then anavalues are within error (Table 6). Total blank values lysed in the CLAIR facility at MIT (Nicolaysen et al., 2000). were p1 ng for each isotopic system considered, which Major element oxides and some trace element conin is negligible with respect to the abundance of the elements centrations were determined by X-ray fluorescence at the dissolved samples (i.e. >80 000, 400 and 5000 ng the University of Massachusetts following the analytical in average for Sr, Pb and Nd, respectively). Measurements procedure described by Rhodes (1996). Major element were performed at Université Libre de Bruxelles on compositions are the mean of duplicate analyses ( Table a multicollector thermal ionization mass spectrometer 4) and LOI is the weight loss on ignition after 30 min at (Micromass VG Elemental Sector 54). Sr and Nd com C. Trace elements, including rare earth element positions were measured in the dynamic mode on a single 1347

8 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Table 4: Major and trace element compositions for basaltic lavas from Mont des Ruches and Mont Fontaine (major elements oxides in wt %; trace elements in ppm) Mont des Ruches basalts Sample BY96-n: Type in the study: Height (m): XRF SiO TiO Al 2 O Fe 2 O 3t MnO MgO CaO Na 2 O K 2 O P 2 O Total LOI FeO mg-no AI Rb Sr Ba V Cr Ni Zn Ga Y Zr Nb La Ce ICP-MS La Ce Sc Cs Hf Ta Th U Pb Nd Sm Eu Tb Yb Lu

9 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Mont des Ruches basalts Sample BY96-n: Type in the study: Height (m): XRF SiO TiO Al 2 O Fe 2 O MnO MgO CaO Na 2 O K 2 O P 2 O Total LOI FeO mg-no AI Rb Sr Ba V Cr Ni Zn Ga Y Zr Nb La Ce ICP-MS La Ce Sc Cs Hf Ta Th U Pb Nd Sm Eu Tb Yb Lu

10 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Table 4: continued Mont Fontaine basalts Sample BY96-n: duplicate Type in the study: Height (m): XRF SiO TiO Al 2 O Fe 2 O MnO MgO CaO Na 2 O K 2 O P 2 O Total LOI FeO mg-no AI Rb Sr Ba V Cr Ni Zn Ga Y Zr Nb La Ce ICP-MS La Ce Sc Cs Hf Ta Th U Pb Nd Sm Eu Tb Yb Lu

11 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Mont Fontaine basalts Sample BY96-n: Type in the study: Height (m): XRF SiO TiO Al 2 O Fe 2 O MnO MgO CaO Na 2 O K 2 O P 2 O Total LOI FeO mg-no AI Rb Sr Ba V Cr Ni Zn Ga Y Zr Nb La Ce ICP-MS La Ce Sc Cs Hf Ta Th U Pb La Ce Nd Sm Eu Tb Yb Lu Groups: Group 1 for depleted low-mgo and high-mgo basalt, and Groups 2 and 3 for enriched high-mgo basalt (a local classification explained in the text). LOI is weight loss on ignition after 30 min at 1020 C. FeO and Fe 2 O 3 recalculated on the basis Fe 2+ /(Fe 2+ + Fe 3+ ) = 0 85; mg-number is calculated on the basis of atomic fractions [Mg 2+ /(Mg 2+ + Fe 2+ )]. AI is alkalinity index and represents the distance of the lava from the alkalic tholeiitic boundary defined for Hawaiian lavas (MacDonald & Katsura, 1964). 1351

12 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Table 5: Average trace element concentration (all in ppm, except P 2 O 5 in wt %) and standard deviations deduced from eight runs (separately dissolved powders) of Leg ODP 183, Site 1140, 31R cm sample mass fractionation by 0 12 ± 0 04% per a.m.u., on the basis of 72 analyses of the NBS 981 Pb standard. 40 Ar/ 39 Ar CHRONOLOGY For all three samples dated by the 40 Ar/ 39 Ar technique, >90% of the released argon defined plateaux (Fleck et al., 1977), which yield ages of >28 Ma (Fig. 2, Table Trace element Average concentration SD 3). Isochronal and plateau ages agree within error (Table 3). The ages for the top and basal samples from Mont P 2 O Fontaine overlap within error, suggesting that this 325 m section of basalt erupted rapidly (<1 Myr). These new Sc data fall within the range of ages previously reported for Ga Kerguelen flood basalts (Nicolaysen et al., 2000) and are Ge comparable with the >29 Ma Mont Bureau and Mont Rb Rabouillère tholeiitic to transitional flood basalts from Sr the north central archipelago (Yang et al., 1998; Fig. 1a). Y This suggests that the emplacement of tholeiitic to alkalic Zr flood basalts exposed on the archipelago occurred within Nb Myr, from >29 Ma in the north central part of the Ba archipelago to >24 25 Ma for mildly alkalic basalts from La the Southeast Province (Weis et al., 1993; Frey et al., Ce b). Pr Nd Sm Eu RESULTS Tb Major element chemistry Dy Basalts from Mont des Ruches and Mont Fontaine are Ho tholeiitic to transitional ( Fig. 3) and overlap in a plot of Er total alkalis vs silica with the >29 Ma Mont Bureau and Tm Mont Rabouillère basalts. Two groups of >29 Ma basalts Yb were defined by Yang et al. (1998): Group D ( depleted ) Lu represents basalts with high MgO contents (6 3 Hf Ta Pb Th U wt %), relatively low incompatible element abundances, and relatively low 87 Sr/ 86 Sr; and Group P ( plume ) represents basalts with low MgO contents ( wt %), relatively high incompatible element abundances, and relatively high 87 Sr/ 86 Sr. In Fig. 3, two groups are clearly shown in the Mont des Ruches and Mont Fontaine samples. As with the >29 Ma basalts, they are also distinguished on the basis of their MgO contents (Figs 4 and 6). The low-mgo basalts (<6 wt %) of Mont des Ta and triple Re Ta filament, respectively. Sr and Nd Ruches and Mont Fontaine overlap with the >29 Ma isotopic ratios were normalized to 86 Sr/ 88 Sr = Group P incompatible element-enriched basalts from and 146 Nd/ 144 Nd = , respectively. The average Mont Bureau and Mont Rabouillère. The high-mgo 87 Sr/ 86 Sr of the NBS 987 and 143 Nd/ 144 Nd of the Rennes basalts (>6 wt %) partly overlap with the >29 Ma Group Nd standards (Chauvel & Blichert-Toft, 2001) during the D basalts of Yang et al. (1998), but have more tholeiitic period of our analyses are ± 7(2σ m on 12 compositions on average. In Mont des Ruches and Mont samples) and ± 10 (2σ m on 27 samples), Fontaine, the low-mgo basalts systematically form the respectively. Pb isotopic compositions were measured in lower flows ( Fig. 4), except for one sample in Mont des the static mode, at temperatures between 1090 and Ruches (BY96-31). The most alkalic sample (BY96-34) 1150 C, on a single Re filament using the H 3 PO 4 silica- is most probably a sill; it is intersected by a dyke in the gel technique. All Pb isotopic ratios were corrected for section ( Fig. 1b). Despite the presence of low-temperature 1352

13 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Table 6: Sr, Nd, and Pb isotopic compositions for Mont des Ruches and Mont Fontaine basalts 87 Sample Type Rb/ ( 87 Sr/ 2σm ( Sr/ Sm/ ( 143 Nd/ 2σm ( 143 Nd/ 238 U/ 235 U/ 232 Th/ ( 206 Pb/ ( 207 Pb/ ( 208 Pb/ ( 206 Pb/ ( 207 Pb/ ( 208 Pb/ 86 Sr Sr)m Sr)i Nd 144 Nd)m 144 Nd)i 204 Pb 204 Pb 204 Pb 204 Pb)m 204 Pb)m 204 Pb)m 204 Pb)i 204 Pb)i 204 Pb)i BY BY BY BY BY BY BY BY BY BY BY BY BY96-43d BY BY BY BY BY BY96-82d BY BY BY BY BY BY BY BY BY BY BY BY BY d, complete duplicate (separate acid-leached). Initial isotopic ratios have been calculated on the basis of the 39 Ar/ 40 Ar eruption age of 28 Ma measured on three basalts from the sections. 1353

14 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Fig. 2. Step-heating plateau and inverse isochron plots for two samples from Mont Fontaine (BY96-80 and BY96-100). Seven and six steps were used to obtain the plateau age for BY96-80 and -100, respectively. All steps were used to define the inverse isochron age. Both the ages for the bottom (BY96-100) and the top (BY96-80) flows are consistent within error with an age of 28 Ma, which indicates that the Mont Fontaine section erupted within 1 Myr. alteration products ( Table 1), the alkali abundances do alteration, such as Ba, Sr and Rb, is correlated with Nb not seem to have been significantly disturbed, as there concentration ( Fig. 7). This suggests that the Mont des is an overall positive correlation between the alkalinity Ruches and Mont Fontaine samples have trace element index (AI) and ratios of incompatible elements such as concentrations that have not been significantly disturbed Nb/Zr ( Fig. 4). by post-magmatic alteration. The Mont des Ruches and Mont Fontaine sections The Mont des Ruches and Mont Fontaine basalts are dominated by relatively high-mgo basalts (6 6 show positive correlations between MgO and Ni and Cr 13 3 wt %) (Fig. 5), in contrast to the 3 5 wt % MgO (Fig. 8) that are consistent with both olivine (± Crbasalts that cover most of the archipelago (Gautier et al., spinel) fractionation to form the low-mgo basalts and 1990; Weis et al., 1993, 1998b; Frey et al., 2000b). For olivine accumulation in high-mgo basalts. many of the Mont des Ruches and Mont Fontaine high- Zr and Nb concentrations in the studied basalts are MgO basalts, the whole-rock mg-number is too high well correlated and show three distinctive groups, which relative to the olivine-basaltic liquid equilibrium curve we call Groups 1, 2 and 3 ( Fig. 9): (Roeder & Emslie, 1970) to be in equilibrium with Group 1: the low-mgo basalts, one high-mgo sample the forsterite content of the olivine cores ( Fig. 6). The (BY96-39) from Mont des Ruches and most of the basalts equilibrium liquid compositions (see arrows in Fig. 6) from Mont Fontaine (samples BY96-80, -85 and -86 are were calculated by systematically removing olivine (5 the exceptions) have Nb/Zr ratios (Nb/Zr > ) 10%) from the whole-rock composition. Relative to the similar to the Ma tholeiitic to transitional basalts whole-rock composition, these equilibrium liquids are from the archipelago ( Fig. 9b; Yang et al., 1998; Frey et 2 wt % lower in MgO, i.e. they change from 9 11 to al., 2002a). 7 9 wt % MgO. This decrease clearly does not bring (2) Group 2: the relatively high-mgo basalts from these samples back to the 3 5 wt % MgO range of most Mont des Ruches (BY96-24 to -38, except BY96-30 and archipelago basalts. the low-mgo sample BY96-31), plus BY96-80 from Mont Fontaine, have higher Nb/Zr (>0 13) that overlaps with the mildly alkalic basaltic lavas from the southern Trace element chemistry archipelago ( Fig. 9; Weis et al., 1993; Frey et al., 2000b). With a few exceptions, such as BY96-31, -46, -84 and (3) Group 3: two high-mgo basalts from Mont Fon- -101, the abundance of elements sensitive to secondary taine (BY96-85 and -86) and three high-mgo basalts 1354

15 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Fig. 3. Total alkalis vs SiO 2 (all in weight percent with 85% of total iron calculated as FeO) classification diagram with the alkalic tholeiitic boundary from MacDonald & Katsura (1964). The Mont des Ruches and Mont Fontaine basalts are shown as squares and circles, respectively. High-MgO (>6 wt %) and low-mgo (<6 wt %) basalts are distinguished by filled and open symbols, respectively. The older >29 28 Ma basalts from the northwestern archipelago (Mont Bureau and Mont Rabouillère, Yang et al., 1998; Mont des Ruches and Mont Fontaine, this study) are tholeiitic to transitional, in contrast to the younger Ma mildly alkalic basalts from the SE archipelago (Weis et al., 1993; Frey et al., 2000b). The Group P and D fields for Mont Bureau and Mont Rabouillère basalts (Yang et al., 1998) refer to the plume and depleted signatures, respectively (see text for explanation). The 26 Ma Mont Tourmente basalts are plotted for comparison and straddle the alkalic tholeiitic boundary (Frey et al., 2002a). Fig. 4. Stratigraphic distribution of phenocrysts (vol. %), MgO (wt %), alkalinity index (AI), Nb/Zr, Sr/Ce, ( 87 Sr/ 86 Sr) i,( 143 Nd/ 144 Nd) i,( 206 Pb/ 204 Pb) i and ( 208 Pb/ 204 Pb) i in the Mont des Ruches and Mont Fontaine sections. Alkalinity index [(Na 2 O + K 2 O) 0 37SiO ] is the deviation from the MacDonald & Katsura (1964) line. Initial isotopic ratios are calculated using the 40 Ar/ 39 Ar age of 28 Ma (Fig. 2). Symbols as in Fig

16 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Fig. 5. SiO 2,Fe 2 O 3,TiO 2,Al 2 O 3 /CaO, Al 2 O 3 and CaO vs MgO diagrams (all oxides in wt %) for Mont des Ruches and Mont Fontaine basalts. No clear fractionation trends are observed and the scatter probably reflects differences in parental magma compositions. Symbols as in Fig. 3. from Mont des Ruches (BY96-30, -33 and -34), have 144 Nd) i (< and >0 5126, respectively; Fig. 10). distinctly higher Nb/Zr (> ). Group 2 basalts have higher ( 87 Sr/ 86 Sr) i and lower ( 143 Nd/ 144 Nd) i ( and , respectively) Isotope geochemistry and basalts from Group 3 have much higher ( 87 Sr/ 86 Sr) i and lower ( 143 Nd/ 144 Nd) i values ( ( 87 Sr/ 86 Sr) i values correlate well with ( 143 Nd/ 144 Nd) i and and , respectively). There is an evolution indicate that Sr isotopic ratios of the acid-leached samples upwards in the Mont des Ruches section from ( 87 Sr/ were not affected by alteration (Fig. 10). Consistent with 86 Sr) i > in the lower part (Fig. 4) to relatively their relatively depleted incompatible element abundance high ratios of > at >340 m. The uppermost three (low Nb/Zr), the Group 1 basalts (BY96-31 is the ex- flows have a nearly constant and lower ratio of >0 7050, ception) have relatively low ( 87 Sr/ 86 Sr) i and high ( 143 Nd/ which is closer to the average value of the basalts for 1356

17 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Fig. 6. Whole-rock mg-number vs olivine forsterite content in Mont des Ruches and Mont Fontaine basalts. The mg-number is molar Mg 2+ / (Mg 2+ + Fe 2+ ) calculated for Fe 3+ /Fe 2+ = Most of the whole-rock mg-numbers fall outside the equilibrium field for Fe/Mg exchange between olivine and basaltic melt (0 30 ± 0 03, Roeder & Emslie, 1970) and reflect olivine accumulation. Arrow-ends indicate the equilibrium liquid compositions that have been recalculated. The equilibrium melt composition was obtained by removal of 5 10 wt % olivine, which is consistent with the petrographic observations. The name of the samples is indicated. the section (0 7052). The Mont Fontaine Sr isotopic incompatible element-depleted high-mgo basalts). Also, compositions are much more homogeneous with average the studied basalts define trends in 208 Pb/ 204 Pb vs 206 Pb/ initial ratios of >0 7046, except for three of the seven 204 Pb that lie above the trend for SEIR mid-ocean ridge upper flows, which range up to > basalt ( MORB) (Mahoney et al., 2002). In a Nd vs Sr isotopic ratio plot (Fig. 10), the Group 1 basalts, i.e. all of the Mont Fontaine basalts (except BY96-80, -85 and -86) and the low-mgo basalts (except DISCUSSION BY96-31 from Mont des Ruches upper section), overlap with the >29 Ma tholeiitic to transitional Group D Stratigraphic geochemical evolution in the basalts (Yang et al., 1998). Group 2 basalts, i.e. mostly Mont des Ruches and Mont Fontaine the Mont des Ruches high-mgo basalts, show limited tholeiitic transitional basalts variation and overlap the field of the enriched Ma Three groups of compositionally distinct basalts are present mildly alkalic basalts ratios field (i.e. 87 Sr/ 86 Sr>0 7052; in the Mont des Ruches and Mont Fontaine sections. 143 Nd/ 144 Nd>0 5126), representative of the Kerguelen They do not show simple correlation between radiogenic plume signature (Weis et al., 1993, 2002; Frey et al., isotopic ratios and MgO content. They were erupted 2000b). Excluding BY96-33 and -34 from Group 3, sequentially in the following order: which have extremely high ( 87 Sr/ 86 Sr) i (>0 7056) and (1) Group 1. Low-MgO plagioclase-rich tholeiitic to low ( 143 Nd/ 144 Nd) i (>0 5125), the Mont des Ruches and transitional basalts are characterized by relatively low Mont Fontaine basalts form a field that ranges from that ( 87 Sr/ 86 Sr) i ( ) and Nb/Zr>0 10, except for metagabbro xenoliths found in alkalic dykes and flows the low-mgo sample BY96-31 located higher in Mont on the archipelago (Mattielli et al., 1996) to the field for des Ruches section ( and 0 11, respectively). the Ma mildly alkalic basalts. Tholeiitic to transitional olivine-rich high-mgo basalts, The studied basalts form fairly well-defined trends overlying the low-mgo basalts, have ( 87 Sr/ 86 Sr) i in 208 Pb/ 204 Pb and 207 Pb/ 204 Pb vs 206 Pb/ 204 Pb plots that > and Nb/Zr> , and are relatively overlap with the fields for other Kerguelen Archipelago less enriched in incompatible elements. Only one basalts (Fig. 11). All Group 2 samples, together with the basalt of this type is present in the Mont des Ruches low-mgo sample located higher up in the Mont des section (BY96-39), whereas 18 of the 21 high-mgo basalts Ruches section (BY96-31), have distinctly higher 208 Pb/ in Mont Fontaine section belong to this group. 204 Pb and 207 Pb/ 204 Pb for a given 206 Pb/ 204 Pb than the (2) Groups 2 and 3. Transitional to slightly alkalic, Group 1 samples (i.e. low-mgo basalts and the relatively olivine-rich, high-mgo basalts are slightly more enriched 1357

18 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Fig. 7. Ce, Rb, Ba and Sr vs Nb concentrations (all in ppm). Except for some of the low-mgo basalts that have relatively low Rb, Sr and Ba (BY96-31, -46 and -101), or high Rb and Ba (BY96-84), the Ba/Rb and K/Rb variations (from 4 6 to 42 and from 202 to 776, respectively) are limited. The good correlation with Nb of these elements sensitive to secondary alteration suggests that the Mont des Ruches and Mont Fontaine samples have trace element concentrations that have not been significantly disturbed by post-magmatic alteration. Symbols and legend as in Fig. 3. in incompatible elements than high-mgo basalts from Temporal evolution in tholeiitic to Group 1 ( Fig. 9). Their isotopic composition is variable transitional basalts from the Kerguelen [e.g. ( 87 Sr/ 86 Sr) i > and Nb/Zr>0 13 Archipelago 0 19]. These groups contain 13 of the 14 high-mgo A general temporal decrease in the proportion of a samples from Mont des Ruches but only three of the 21 depleted component in the source of the Kerguelen high-mgo samples from Mont Fontaine. Together with Archipelago basalts was interpreted by Gautier et al. the low-mgo sample from Mont des Ruches located (1990) and Weis et al. (1993) as reflecting the decreasing higher in the section ( Fig. 4), these groups have more contribution of a MORB source-related asthenospheric radiogenic 207 Pb/ 204 Pb and 208 Pb/ 204 Pb for a given 206 Pb/ component with time, as the ridge moved away from the 204 Pb than Group 1 basalts. The lead isotopic stratigraphic archipelago. In contrast, the coeval eruption at 29 Ma variations ( Fig. 4) are consistent with Sr and Nd isotopic of depleted (Group D) and enriched (Group P) basalts, variations and show enrichment upwards in the Mont the latter with isotopic ratios similar to those proposed des Ruches and Mont Fontaine sections. The trace for the Kerguelen plume, led Yang et al. (1998) to argue element and isotopic compositions of Group 1 basalts that there was no systematic temporal geochemical trend suggest that both low-mgo and high-mgo basalts from in the archipelago. On the basis of high Ba/Th and Sr/ this group had similar parental magmas. The Group 2 Nd, reflecting a plagioclase-rich component despite the and 3 basalts are slightly more enriched in alkalis ( Figs absence of plagioclase phenocrysts, Yang et al. (1998) 3 and 4) and may have formed by a lower extent inferred that Group D basalts had interacted with a of melting as supported by their higher La/Yb (5 18 gabbroic SEIR crust component. The positive correlation compared with 3 10, respectively) on average. of 143 Nd/ 144 Nd with MgO content in high-mgo basalts 1358

19 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Fig. 8. Ni and Cr abundance (in ppm) vs MgO (wt %) in Mont des Ruches and Mont Fontaine basalts. The positive correlations between MgO and Ni and Cr are consistent with olivine (± Cr-spinel) fracof lavas in Mont des Ruches and Mont Fontaine. Group 1 represents Fig. 9. (a) Abundance of Zr vs Nb (all in ppm) showing three groups tionation. Fields for Groups 1, 2 and 3, which are defined in the text, are shown. Symbols and legend as in Fig. 3. the low-mgo basalts (Nb/Zr >0 10), the high-mgo BY96-39 from Mont des Ruches and most of the basalts (samples BY96-80, -85 and -86 are the exceptions) from Mont Fontaine (Nb/Zr > ). Group 1 basalts have Nb/Zr ratios similar to the Ma tholeiitic from Mont Fontaine Group 1 suggests interaction of the to transitional basalts from the archipelago (pale grey field; Yang et al., 1998; Frey et al., 2002a). Group 2 represents the relatively high-mgo ascending magmas with the lithosphere. In contrast, the basalts from Mont des Ruches (BY96-24 to -38, except BY96-30 and Pb and Sr isotopic composition enrichment upwards in the low-mgo BY96-31), plus BY96-80 from Mont Fontaine, and have 28 Ma Mont des Ruches basalts is independent of MgO higher Nb/Zr >0 13 that overlap with the mildly alkalic basaltic lavas content and is not consistent with combined assimilation from the southern archipelago (Weis et al., 1993; Frey et al., 2000b). Group 3 represents two high-mgo basalts from Mont Fontaine (BY96- and fractional crystallization. 85 and -86) and three high-mgo basalts from Mont des Ruches (BY96- A MORB source-like component was present only 30, -33 and -34), which have distinctly higher Nb/Zr > during the formation of the oldest >29 28 Ma flood The differences in the slope of the three trends reflect different parental magma compositions and do not reflect different fractionating phases. basalts and Yang et al. (1998) noted that a plume com- (b) Histogram of Nb/Zr in the Kerguelen Archipelago flood basalts ponent was present throughout the growth of the Ker- showing (in black) the three populations in Mont des Ruches and Mont guelen Archipelago. The Sr and Nd isotopic data for the Fontaine basalts. The fields for SEIR MORB are from D. Christie (personal communication, 2002). The fields for Northern Kerguelen 28 Ma Mont des Ruches and Mont Fontaine basalts are Plateau, Site 1140, are from Weis & Frey (2002). The fields for consistent with these observations. A plot of Δ8/4 vs Kerguelen Archipelago basalts are from Weis et al. (1993), Yang et al. ( 87 Sr/ 86 Sr) i (Fig. 12) shows the important variations of (1998) and Frey et al. (2000b). Symbols as in Fig

20 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Fig. 10. ( 87 Sr/ 86 Sr) i vs ( 143 Nd/ 144 Nd) i in Mont des Ruches and Mont Fontaine basalts. Labelled fields are SEIR MORB (Mahoney et al., 2002); Northern Kerguelen Plateau (NKP) Site 1140 (Weis & Frey, 2002); Mont Tourmente tholeiitic to transitional basalts (Frey et al., 2002a); Group D and P Mont Bureau and Mont Rabouillère tholeiitic to transitional basalts (Yang et al., 1998); Ma mildly alkalic basalts (Weis et al., 1993; Frey et al., 2000b); metagabbro xenoliths (Mattielli et al., 1996). Groups 1, 2 and 3 (dashed fields) are labelled. Crosses represent binary mixtures (space between crosses represents 20% fraction) between average SEIR N-MORB and the Kerguelen plume end-members, the fields for which are highlighted in pale grey. Sr concentrations of 122 and 559 ppm and Nd concentrations of 10 and 37 ppm have been used for the compositions of the SEIR and plume end-members, respectively. Symbols as in Fig. 3. Is a depleted component intrinsic to the Kerguelen mantle plume? There has been much discussion about the origin of a depleted component in the petrogenesis of Icelandic plume-related basalts (e.g. Fitton et al., 1997; Hanan et al., 2000; Kempton et al., 2000) and the need for heterogeneities intrinsic to the Icelandic plume has be- come widely accepted. The heterogeneity of the Ga- lápagos plume has been related to both shallow mixing with local asthenosphere and intrinsic heterogeneity within the plume (White et al., 1993; Kurz & Geist, 1999; Harpp & White, 2001). In general, the Kerguelen Archipelago flood basalts, especially the younger alkalic flood basalts, have nearly uniform Sr and Nd isotopic ratios and these have been attributed to the Kerguelen plume (Weis et al., 1998b, 2002; Frey et al., 2000b) without any contribution from a depleted component. It has been suggested that the Ma Kerguelen Archipelago flood basalts with lower 87 Sr/ 86 Sr and higher 143 Nd/ 144 Nd contain a proportion of an SEIR-related component (Yang et al., 1998; and this study). A MORB-source depleted mantle component appears to be required from the Hf and Nd isotopic systematics of the Kerguelen Archipelago basalts (Mattielli et al., 2002). Frey et al. (2002a) have noted that the 26 Ma basalts from Mont Tourmente are nearly homogeneous in Sr and Nd isotopic compositions, but that they have ratios distinct from those proposed for the Kerguelen plume (Figs 10 and 12). They considered the possibility that these different isotopic ratios reflect plume heterogeneity. As the Mont Pb isotopic compositions in these tholeiitic to transitional sections. In such a diagram, where Δ8/4 represents the deviation of 208 Pb/ 204 Pb for a given 206 Pb/ 204 Pb from the Northern Hemisphere Reference Line (Hart, 1984), Mont des Ruches and Mont Fontaine basalts define a mixing trend between the inferred composition of the Kerguelen plume (Weis et al., 1993, 1998b, 2002) and the SEIR N- MORB field (Mahoney et al., 2002). We suggest that the compositions observed in Mont des Ruches and Mont Fontaine basalts are consistent with simple binary mixing between an SEIR-like component and the Kerguelen plume component, and infer that the absence of a de- pleted SEIR-like component in the <26 Ma basalts in the archipelago reflects decreasing contribution of an SEIR-like source component. Nature of the depleted component in tholeiitic to transitional basalts from the Kerguelen Archipelago A Nb/Y vs Zr/Y diagram has been useful in dis- criminating between source components contributing to Icelandic plume-related magmatism (Fitton et al., 1997). In such a diagram, Group 1 basalts from both Mont des Ruches and Mont Fontaine are distributed along a mixing trend (Fig. 13), located along the lower boundary of the Iceland Neovolcanic Zone basalt field (Fitton et al., 1997), between the Kerguelen plume inferred composition (Weis et al., 1993, 2002; Frey et al., 2000b) and SEIR N-MORB. The high-mgo basalts in Group 1 have lower Nb/Y and Zr/Y than the low-mgo basalts and may represent higher extents of partial melting of a similar source. Group 2 and 3 basalts plot mostly within the Iceland plume array (Fig. 13). The temporal geochemical trend observed from 29 to 24 Ma in the archipelago flood basalts shows a decreasing role for a depleted component as the SEIR moved from 350 to 550 km away from the archipelago. There are several possible origins for such a component, such as entrained asthenosphere at the periphery of the plume, migration of MORB material from the SEIR axis, and a depleted component within the plume itself. Below we address each of these possibilities using isotopic and geochemical systematics of the Kerguelen Archipelago flood basalts and we use Fig. 14 to illustrate a possible scenario. 1360

21 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Fig. 11. Lead isotopic compositions for Mont des Ruches and Mont Fontaine basalts, showing that the studied basalts form well-defined trends in 208 Pb/ 204 Pb and 207 Pb/ 204 Pb vs 206 Pb/ 204 Pb plots that overlap with the fields for other Kerguelen Archipelago basalts. All Group 2 samples together with the low-mgo sample located higher in the Mont des Ruches section (BY96-31), have distinctly higher 208 Pb/ 204 Pb and 207 Pb/ 204 Pb for a given 206 Pb/ 204 Pb than the Group 1 samples. Also, the Mont des Ruches and Mont Fontaine basalts define trends in 208 Pb/ 204 Pb vs 206 Pb/ 204 Pb that lie above the trend for SEIR MORB. Labelled fields and crosses as in Fig. 10. Pb concentrations of 0 5 and 4 6 ppm have been used for the compositions of the SEIR and plume end-members, respectively. Symbols as in Fig. 3. Tourmente basalts straddle the isotopic gap that exists between the Mont Fontaine samples ( Figs 10 12), an alternative explanation is that the Mont Tourmente basalts reflect a homogeneous source formed by efficient mixing of plume and MORB-source components. If this interpretation is valid, an explanation is required for the absence of such efficient mixing in the Mont Bureau, Rabouillère, Ruches and Fontaine sections. We have demonstrated a stratigraphic evolution in the 28 Ma Mont des Ruches and Mont Fontaine basaltic compositions with enrichment upwards in the sections (Fig. 4). This is also true for the 29 Ma Mont Bureau section, where most enriched basalts (Group P) are preferentially located at the top of the section (Yang et al., 1998). These stratigraphic changes in the Ma basaltic sections may reflect replenishment in the magma plumbing system, with replenishing magmas resulting from a higher contribution of the Kerguelen plume component and providing more heterogeneous compositions than those observed in the 26 Ma Mont Tourmente basalts. In summary, we propose that depleted heterogeneities intrinsic to the plume are not necessary to account for the abrupt decrease in the role of a depleted component at 25 Ma (see Weis & Frey, 2002, fig. 15); the cessation of sampling of depleted heterogeneities within 1 Myr seems unlikely as the increase in lithosphere thickness within 1361

22 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 which formed by melting of the head of the Kerguelen plume, where most of the entrained depleted mantle may have mingled with the upwelling plume (e.g. Campbell & Griffiths, 1990). Only one ODP Leg 120 site shows the important involvement of a depleted component (Site 749; Frey et al., 2002b). Alternatively, melting of plume and entrained depleted mantle may occur in ponded plume material areas, which can reach lateral extensions of thousands of kilometres (Nataf, 2000), and where mingling is more efficient. On the basis of the geochemical data we are unable to determine if the depleted component was depleted mantle entrained during plume ascent or depleted material associated with the SEIR. Is the depleted component the upwelling asthenospheric source for SEIR basalts? Rare earth element patterns of lower-crustal metagabbro xenoliths found in the alkalic dykes and flows on the archipelago are similar to those of gabbroic dykes related to MORB and to plagioclase-rich cumulates sampled at Indian Ocean ridges (Grégoire et al., 1998). Although the isotopic compositions of the Kerguelen metagabbro xenoliths, with 143 Nd/ 144 Nd > and 87 Sr/ 86 Sr > (Mattielli et al., 1996), do not overlap with SEIR MORB, they overlap with the Ma depleted basalts from the Kerguelen Archipelago [e.g. low ( 87 Sr/ 86 Sr) i of ; Fig. 10]. Hence an SEIR component may be present in these xenoliths. Plume ridge interactions involving asymmetric spreading along the south Mid-Atlantic Ridge and subsequent sublithospheric channelling of plume material to the ridge axis have been proposed, based on geochemical observations of short-wavelength variations in the ratios of incompatible elements, such as La/Sm and Nb/Zr (Schilling et al., 1985). Locally, the isotopic compositions of lavas from the Easter seamount chain have been used to argue for bidirectional flow between the spreading ridge axis and the Easter plume (Haase, 1996). Geo- chemical evidence of mixing between plume- and ridge- derived material is observed as far as 400 km away from the ridge axis in the Foundation chain, where the Pacific Antarctic Ridge moves towards the Foundation hotspot (Maia et al., 2000). In the opposite case, when a ridge moves away from a hotspot location, as is the case for the SEIR relative to the Kerguelen hotspot, numerical experiments (e.g. Ito et al., 1997) show that hotspot ridge interaction will be enhanced and active at larger distances than when the ridge moves towards the hotspot. Numerical experiments from Yale & Phipps Morgan (1998) show the important effect of asymmetric spreading on flow of plume-related material toward a ridge. Also, seismicity along the E fracture zone on the Antarctic plate between the Kerguelen Plateau and the Fig. 12. Δ8/4 Pb vs ( 87 Sr/ 86 Sr) i and Nb/Zr. Δ8/4 represents the deviation of 208 Pb/ 204 Pb for a given 206 Pb/ 204 Pb from the Northern Hemisphere Reference Line (Hart, 1984). Mont des Ruches and Mont Fontaine basalts clearly belong to the Dupal anomaly (Hart, 1984) and define a general mixing trend between the Kerguelen plume inferred composition (Weis et al., 1993, 1998a; Frey et al., 2000b) and SEIR N- MORB-like compositions. UCC and the associated arrow indicate the effect of upper continental crust contamination on magma compositions. Labelled fields and crosses as in Fig. 10. Nb concentrations of 3 and 47 ppm and Zr concentrations of 95 and 285 ppm have been used for the compositions of the SEIR and plume end-members, respectively. Symbols as in Fig. 3. this short time interval would not be important enough to abruptly limit the extent of partial melting. Is the depleted component entrained, depleted upper mantle? An entrained, depleted upper mantle component should be mostly present in the Kerguelen plateau lithosphere, 1362

23 DOUCET et al. DEPLETED MANTLE IN KERGUELEN ARCHIPELAGO BASALTS Fig. 13. Log (Nb/Y) vs log (Zr/Y) diagram showing upper and lower boundaries defined for the Iceland Neovolcanic Zone basalts (continuous lines) (Fitton et al., 1997). Samples from Mont des Ruches and Mont Fontaine are distributed along distinct mixing lines between the Kerguelen plume (Weis et al., 1993; Frey et al., 2000b) and the SEIR N-MORB compositions. This reflects distinct source components and variable degrees of partial melting as shown with inset arrows. A line of constant Nb/Zr (0 5) is shown for reference. Labelled field for SEIR MORB is from D. Christie (personal communication, 2002). Labelled fields and crosses as in Fig. 10. Nb and Zr concentrations used for the compositions of the two end-members as in Fig. 12. Y concentrations of 32 and 28 ppm have been used for the compositions of the SEIR and plume end-members, respectively. Symbols as in Fig. 3. Amsterdam Saint Paul Plateau ( Fig. 1) has been in- in the plume stem ( Fig. 14) to form the heterogeneous terpreted as thermal and bending stresses in the lithosphere coeval depleted and enriched compositions observed in overlying a thermal anomaly resulting from the Ma flood basalts from the northwestern Ker- channelled flow between the Kerguelen hotspot and the guelen Archipelago. This scenario best resolves the tem- SEIR axis (Bergman et al., 1984). poral decreasing contribution of SEIR mantle source in The 34 Ma basalts from the Northern Kerguelen Plateau the Northern Kerguelen Plateau and the Kerguelen at Site 1140, which was >200 km away from the Archipelago, as the ridge moved from 200 km to 560 km SEIR at its formation (Fig. 14), show evidence of plume away from the Kerguelen hotspot. This interpretation and upwelling depleted mantle magma mixing (Weis & could also explain the presence of enigmatic topographic Frey, 2002). The Kerguelen hotspot was > km highs observed on the ocean floor between the Northern away from the ridge axis when the compositionally Kerguelen Plateau and the SEIR axis [see global topoheterogeneous Ma flood basalts erupted ( Fig. 14); graphic map of Smith & Sandwell (1997)]. these Ma basalt compositions also reflect mixing of plume and SEIR material, but with a significantly higher contribution of the Kerguelen plume component than for depleted basalts from Site 1140 ( Figs 10 13). These differences may indicate that mixtures of plume Did continental crust contaminate Mont and upwelling asthenosphere material formed during des Ruches and Mont Fontaine basalts? asymmetric spreading of the SEIR (oval labelled 1+4 Contrary to the Cretaceous Kerguelen Plateau, which in Fig. 14). We propose that plume and ridge source records significant continental contamination in the early material mingled below the lithosphere and that these products of Kerguelen plume activity (Mahoney et al., heterogeneous mixtures of upwelling plume and depleted 1995; Weis et al., 2001; Frey et al., 2002b; Ingle et al., mantle source for SEIR basalts have subsequently melted 2002), the Mont des Ruches and Mont Fontaine basalts 1363

24 JOURNAL OF PETROLOGY VOLUME 43 NUMBER 7 JULY 2002 Fig. 14. Schematic diagram illustrating the temporal evolution of the Kerguelen plume, the Kerguelen Plateau lithosphere and the Kerguelen Archipelago relative to the SEIR, from 40 Ma, when the ridge and the Kerguelen hotspot were coincident, to 24 Ma, when basalts with isotopic characteristics of the Kerguelen plume erupted. In this figure the position of the SEIR is fixed, with the location of the plume relative to the ridge changing with time. We used a crust thickness of >20 km according to Charvis et al. (1995), a minimum lithospheric thickness of >120 km, and a minimum 100 km plume stem diameter, according to Wolfe et al. (1997). The triangle below the SEIR axis represents upwelling asthenosphere, with melt segregating at the top of the triangle (black area). The thick black and white opposing arrows between 40 and 34 Ma represent movement of MORB-related asthenosphere and plume material, respectively. The oval labelled 1+4 represents a mixture of the two materials, which are subsequently melted to form basalts in the Kerguelen Archipelago. Different proposals for the nature of the depleted component involved in the Kerguelen Archipelago basalts are represented (2 5) and discussed in the text. We favour an interpretation where physical mingling of horizontally migrating depleted asthenosphere and the upwelling Kerguelen mantle plume form a mixed source (labelled 1+4) in a sublithospheric channel that partially melts to form Kerguelen Archipelago basalts. Although this process occurs at a distance of 450 km between the SEIR and plume (location of Mont des Ruches and Mont Fontaine), the absence of depleted material in 24 Ma Mont Crozier basalts indicates that the flux of depleted asthenosphere decreases with increasing distance from the ridge. 28 Ma Kerguelen Archipelago basalts [e.g. ( 87 Sr/ 86 Sr) have isotopic compositions that do not reflect con- >0 7052, Weis et al., 1993; Frey et al., 2000b) are contamination by lower or upper continental crust (e.g. Fig. sistent with variable amounts of mixing between a upwel- 12). The positive correlation of their Sr isotopic ratios ling depleted mantle asthenospheric component and the with Nb/Zr ( Fig. 4) is inconsistent with the potential Kerguelen plume source. Evidence for mixing between involvement of continental crust, which has high 87 Sr/ plume and ridge-related material from 34 Ma in the 86 Sr and low Nb/Zr. The relative enrichment in Nb Northern Kerguelen Plateau (ODP Site 1140) to at least compared with Kerguelen Plateau basalts argues against 28 Ma, i.e. from >200 km to >400 km away from the continental contamination because continental crust is ridge axis, suggest that plume ridge interactions occurred depleted in Nb (e.g. Sun & McDonough, 1989). On a as the SEIR migrated away from the Kerguelen hotspot larger scale, there is also no continental contamination (as far as 450 km away from the SEIR axis). This is documented at ODP Site 1140, >300 km north of the consistent with previous interpretation of seismicity on archipelago, in the Northern Kerguelen Plateau (Weis & the oceanic floor between the Northern Kerguelen Plat- Frey, 2002). eau and the SEIR that could reflect the existence of a sublithospheric channel flow between the Northern Kerguelen Plateau and the SEIR. We propose that during asymmetric spreading along the SEIR, a two-way channel CONCLUSIONS flow permitted an efficient mingling of the upwelling Strong geochemical and isotopic heterogeneities in the depleted mantle with the Kerguelen plume. Mixtures of the enriched and depleted sources, which are still present i > ], compared with the relative homoand in the top of the plume stem, were subsequently melted geneity and enriched composition of Ma mildly the resulting magmas ascended through the Ker- alkaline basalts on the archipelago (e.g. 87 Sr/ 86 Sr guelen Plateau lithosphere. Entrained depleted mantle 1364