| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology | Volume 43 | Number 7 | Pages 1177-1205 | 2002
© Oxford University Press 2002
Mantle Sources and the Highly Variable Role of Continental Lithosphere in Basalt Petrogenesis of the Kerguelen Plateau and Broken Ridge LIP: Results from ODP Leg 183
1DEPARTMENT OF CIVIL ENGINEERING AND GEOLOGICAL SCIENCES, UNIVERSITY OF NOTRE DAME, NOTRE DAME, IN 46556, USA
2SCHOOL OF OCEAN AND EARTH SCIENCE AND TECHNOLOGY, UNIVERSITY OF HAWAII, HONOLULU, HI 96822, USA
Received July 13, 2001; Revised typescript accepted January 31, 2002
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSAMPLE PREPARATION AND ANALYSISRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Ocean Drilling Program (ODP) Leg 183 was designed to investigate the origin and evolution of the large igneous province composed of the Kerguelen Plateau and Broken Ridge. Of the eight sites drilled, basalt was recovered from seven, five on the plateau and two on Broken Ridge. We present results from four of these sites, 1136, 1138, 1141 and 1142. Although this large igneous province is interpreted as being derived from the Kerguelen mantle plume, the geochemical characteristics of basalt from some parts of the province indicate a role for continental lithosphere. The 118–119 Ma basalt flows recovered in the Southern Kerguelen Plateau (Site 1136) have a more subtle continental signature than shown by basalt at Leg 119 Site 738. A continental signature is absent in the 100–101 Ma tholeiitic basalts at Site 1138 in the Central Kerguelen Plateau (CKP); their age-corrected Nd–Sr–Pb isotopic values and incompatible element ratios are similar to those estimated for primitive mantle. These flows may represent a major mantle source in the Kerguelen starting-plume head. The 20 basalt units identified are a product of magma chamber replenishment, fractional crystallization, and resorption of crystallizing phases. The topmost unit, Unit 1, is a dacite that evolved from a basalt magma similar to those represented by Units 3–22; unlike the basalts the dacite magma was probably influenced by continental material. Middle Cretaceous (
95 Ma) lavas of Sites 1141 and 1142 on Broken Ridge (originally part of the CKP) are alkalic, with one exception (a tholeiite at the base of Site 1142). The alkalic lavas may represent a late-stage cap or carapace of relatively low-degree partial melts that overlies a thick tholeiitic lava pile. The tholeiite and pebbles from the top of a probable talus deposit (Unit 2) at Site 1142 have geochemical signatures consistent with a minor contribution from continental material. This signature is absent in the other units from these two sites, which have ocean-island-like incompatible element ratios and age-corrected isotopic characteristics similar (but not identical) to those proposed for the post-30 Ma Kerguelen plume. These alkalic basalts may be the purest representatives of the Cretaceous plume tail composition yet found. KEY WORDS: assimilation; basalt; large igneous province; mantle plume; Ocean Drilling Program
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE PREPARATION AND ANALYSISRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The Kerguelen Plateau and Broken Ridge constitute a large igneous province (LIP) in the southern Indian Ocean that originated after the
132 Ma break-up of India and Australia (e.g. Kent et al., 2002). The consensus view is that this LIP represents the volcanic outpouring from a starting-plume head that also produced magmatism in NE India (Rajmahal Traps; e.g. Mahoney et al., 1983
; Kent et al., 1997, 2002), the Bunbury Basalt of southwestern Australia, and the Naturaliste Plateau off the southwestern coast of Australia (e.g. Storey et al., 1992; Mahoney et al., 1995; Frey et al., 1996). Recent data from Leg 183 suggest several distinct stages of plateau building (Pringle & Duncan, 2000; Coffin et al., 2002; Duncan, 2002). The Kerguelen Plateau is 200–600 km wide, extends for 2300 km from 46°S to 64°S and contains the volcanic islands of Heard and McDonald, as well as the Kerguelen Archipelago (Fig. 1). If a plume model is accepted, these islands represent the most recent volcanic products of a long-lived Kerguelen plume tail that also produced the Ninetyeast Ridge (e.g. Frey et al., 1991; Saunders et al., 1991; Frey & Weis, 1995). Broken Ridge, located 1800 km to the north, was separated from the main plateau by the opening of the Southeast Indian Ridge (SEIR) during the Early Tertiary (e.g. Tikku & Cande, 2000). Located approximately midway between the Kerguelen Archipelago and Broken Ridge on the SEIR are the relatively recent oceanic islands of Amsterdam and St. Paul (Fig. 1), the largest volcanoes of the Amsterdam–St. Paul hotspot.
|
The Kerguelen Plateau is broadly divided into southern, central, and northern domains (Fig. 1), each with a distinct crustal structure and thickness (e.g. Coffin et al., 1990; Schaming & Rotstein, 1990; Frey et al., 2000). The Southern Kerguelen Plateau (SKP) is composed of an igneous crust of 22 km thickness that, on the basis of seismic velocities, may contain fragments of continental material, especially in the northern SKP (Operto & Charvis, 1995, 1996; Charvis et al., 1997). The igneous crust of the Central Kerguelen Plateau (CKP) is between 19 and 21 km thick and does not contain the seismically reflective transition zone suggestive of continental material as seen in the SKP. The CKP includes the volcanic islands of Heard and McDonald, and, originally, Broken Ridge; Elan Bank extends westward from the CKP (Fig. 1). Seismically, the igneous basement of the Northern Kerguelen Plateau (NKP) is divided into an upper crust (8–9·5 km thick) and a lower crust (6–9·5 km thick) (Recq et al., 1990, 1994; Charvis et al., 1995).
Basement rocks of the Kerguelen Plateau have been sampled via dredging and drilling. Basalt was encountered at one drill site during Ocean Drilling Program (ODP) Leg 119 (Site 738) and four drill sites during ODP Leg 120 (Sites 747–750). Sites 738, 748, 749, and 750 are in the SKP, and Site 747 is in the CKP (Fig. 1). Basalt at Site 748 was an inter-sediment unit, whereas the cored basalt sequences at Sites 738, 747, 749, and 750 were more substantial and were considered to represent igneous basement (e.g. Barron et al., 1989; Schlich et al., 1989; Salters et al., 1992; Storey et al., 1992). The Ninetyeast Ridge was cored at four sites during Deep Sea Drilling Project (DSDP) Legs 22 (Sites 214 and 216) and 26 (Sites 253 and 254), and at three sites during ODP Leg 121 (Sites 756, 757, and 758) (see Frey et al., 1977, 1991; Mahoney et al., 1983; Baksi et al., 1987; Davies et al., 1989; Storey et al., 1989, 1992; Saunders et al., 1991; Frey & Weis, 1995, 1996). Dredged basalts from the Kerguelen Plateau, Naturaliste Plateau and Broken Ridge were studied by Davies et al. (1989), Weis et al. (1989), Storey et al. (1992), and Mahoney et al. (1995). The majority of basalts associated with the Kerguelen hotspot are tholeiitic or transitional between tholeiitic and alkalic (Coffin et al., 2000) as defined by Macdonald & Katsura (1964).
Leg 183 of the Ocean Drilling Program drilled eight sites on the Kerguelen Plateau–Broken Ridge LIP (Coffin et al., 2000). Seven of these sites reached igneous basement. Site 1135 (which cored only sediment) and Site 1136 were drilled in the SKP, Site 1137 in Elan Bank, Site 1138 in the CKP, Sites 1139 and 1140 in the NKP, and Sites 1141 and 1142 in Broken Ridge (CKP; see Fig. 1). The focus of this paper is the basement rocks recovered from Kerguelen Plateau Sites 1136 and 1138, and from Sites 1141 and 1142 on Broken Ridge.
General background
The influences of continental crust and/or lithospheric mantle have been documented in the continental portions of the Kerguelen LIP, the Rajmahal Traps (see Kent et al., 1997, 2002) and Bunbury Basalt (Frey et al., 1996). In the oceanic parts of the LIP, it has also has been recognized that a continental lithospheric mantle and/or crust signature is present in lavas of the SKP and CKP, Broken Ridge and the Naturaliste Plateau (e.g. Storey et al., 1992; Mahoney et al., 1995; Frey et al., 2002), and that at least a small amount of continental lithospheric mantle is present in the NKP (Hassler & Shimizu, 1998). One of the major discoveries of Leg 183 was the presence of garnet–biotite gneiss cobbles in an inter-flow conglomerate from Site 1137 (Frey et al., 2000; Nicolaysen et al., 2001; Weis et al., 2001; Ingle et al. 2002a), which demonstrated the presence of Proterozoic continental crust in Elan Bank.
The longevity of the Kerguelen hotspot (which has been active for at least 120 Myr; Duncan, 2002) affords the opportunity to examine temporal compositional changes in hotspot-derived magmas. However, the highly variable influence and geochemical heterogeneity of continental lithospheric material complicate interpretation of geochemical data. Variations documented from pre-Leg 183 sampling were generally interpreted in terms of a plume-type model. Most workers have interpreted the lavas associated with the Kerguelen hotspot to reflect variable mixtures of a dominant Kerguelen plume component, a second, minor plume-type component (similar isotopically to recent products of the Amsterdam–St. Paul hotspot), plus small amounts of Indian mid-ocean ridge basalt (MORB)-type mantle (Storey et al., 1988; Frey & Weis, 1995, 1996) and isotopically heterogeneous continental crustal and/or lithospheric mantle material (e.g. Weis et al., 1989, 1991, 2001; Storey et al., 1992; Mahoney et al., 1995; Frey et al., 2002).
|
SAMPLE PREPARATION AND ANALYSIS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE PREPARATION AND ANALYSISRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The analytical strategy was to analyze, where possible, one sample for major and trace elements from each unit defined during shipboard description of the core (Coffin et al., 2000
). For the thicker units, at least two samples were analyzed. Unless otherwise stated, the samples analyzed were the least altered available and were chosen on the basis of the following criteria: (1) macroscopic observation; (2) thin section examination; (3) relative density and hardness. Trace element analyses (Table 1) were carried out at the University of Notre Dame (see http://www.nd.edu/icpmslab). All surfaces that had been in contact with the drill bit were removed from samples with a water-cooled rock saw and the interior portions were cut into slabs of 1 cm thickness. All sawn surfaces were then removed using a diamond wheel. The slabs were placed inside a plastic bag, sealed with duct tape, and struck with a duct-tape-covered hammer to reduce the sample to small enough chips (1 cm) to pass through alumina jaw crushers. A portion of these chips was separated for isotope analysis at the University of Hawaii. The remaining chips were lightly leached in dilute (
0·25 N) HCl in an ultrasonic bath and then rinsed several times in high-purity deionized (18 M
cm) water. After drying on a hot plate at 50–75°C, the chips were examined under a binocular microscope and pieces containing veins or vesicles were removed. Remaining pieces were passed through the jaw crushers and then reduced to a powder in an alumina mill. We took great care to remove and avoid metallic contamination as certain trace elements (Pb, for instance) are significantly affected by such contamination. Splits of the powders were dissolved for trace element analysis by inductively coupled plasma mass spectrometry (ICP-MS) using the method described by Neal (2001). The ICP-MS analyses were conducted in six separate batches, resulting in 10 analyses of reference material BHVO-1. Reproducibility is generally better than 3–5% on the basis of duplicate analyses (Fig. 2). All the samples for which data are reported here were analyzed by ICP-MS at least twice. If the data suggested dissolution might have been incomplete (i.e. depletions in Zr and Hf on primitive mantle normalized plots), a LiBO2 fusion method of sample preparation was employed. The latter preparation was conducted to ensure that Zr and Hf were not being retained in phases resistant to conventional inorganic acid attack (e.g. zircon). Major element compositions of the powders were analyzed by X-ray fluorescence (XRF) at the University of Massachusetts [see Rhodes (1996) and http://www.geo.umass.edu/xrf for details]. The mg-number [molar Mg/(Mg + Fe2+)] of the samples was calculated assuming that 15% of the total iron was Fe3+. XRF data quality was monitored by analyzing reference material K-1919 and undertaking replicate analyses.
|
|
Isotopic ratios of Sr, Nd, and Pb (Table 2) were measured following preparation and measurement procedures outlined by Mahoney et al. (1998). As in our previous studies, isotope-dilution measurements of Rb, Sr, U, Th, Pb, Sm, and Nd abundances were made on the same splits of sample as used for isotopic analysis, to age-correct the isotopic ratios. Because these measurements were made on small (2 mm) chips hand-picked from the bulk rock and subsequently subjected to acid-cleaning and, in many cases, to a multi-step acid-leaching procedure before dissolution, the elemental abundances derived by isotope dilution generally do not represent whole-rock values.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE PREPARATION AND ANALYSISRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Site 1136
Basement recovered at Site 1136 was divided into three basaltic units by the shipboard scientific party (Coffin et al., 2000
). Only 53 cm of the highly altered rubbly top of Unit 3 was recovered and the sample had to be carefully hand-picked to obtain enough material suitable for trace element and isotopic work. All three Site 1136 units are plagioclase, clinopyroxene, ±olivine phyric.
Units 1 and 2 exhibit little variation in major element composition (e.g. mg-number = 0·55–0·56; TiO2 = 1·65–1·75 wt %; Table 1), consistent with the shipboard XRF data (Coffin et al., 2000). Except for their higher (207Pb/204Pb)t, Units 1 and 2 are isotopically similar to the Site 1138 basalt units (Table 2, Figs 3 and 4); relative to their 
Nd(t) values, which are near zero, their (206Pb/204Pb)t and (87Sr/86Sr)t are slightly lower than found for the Kerguelen Archipelago, and (206Pb/204Pb)t is much lower than for Heard Island at similar Nd.
|
|
Trace element and isotope data demonstrate that Unit 3 is distinct (Tables 1 and 2; Figs 3–5), with higher REE (rare earth element) and Y abundances, a depletion in the LREE (light REE) relative to the heavy REE (HREE), markedly higher Nd(t), and more radiogenic (208Pb/204Pb)t and (87Sr/86Sr)t than Units 1 and 2. The sample is the most altered of any that we analyzed in this study, as indicated by the marked elevation of normalized U, Ba and, particularly, Rb abundances relative to Th and Nb (Fig. 5a). In contrast to Units 1 and 2, Unit 3 is not anomalously enriched in Pb. High field strength element abundances are broadly similar for all three units, although Units 2 and 3 have primitive mantle normalized (Sun & McDonough, 1989) Nb/Ta ratios [(Nb/Ta)PM] < 1; these signatures, and the marked depletion in Zr and Hf relative to the REE in the Unit 3 pattern, have been replicated through two different dissolutions of the rock powder and one LiBO2 fusion. The age-corrected Nd(t) value (+5·1) of the Unit 3 lava is by far the highest of any among our Leg 183 samples and similar to the highest values measured for basalts from Leg 120 Sites 749 and 750 (Storey et al., 1992; Frey et al., 2002). Its (87Sr/86Sr)t (0·70502) is similar to values for Site 750, which are higher for their Nd(t) than seen for any other Kerguelen Plateau, Broken Ridge, Kerguelen Archipelago, or Heard or McDonald Island lavas. However, we caution that the high degree of alteration of the rock recovered from Unit 3 could mean that a significant amount of alteration-related Sr could have survived aggressive acid-leaching and that the (87Sr/86Sr)t value in Table 2 thus could be higher than the magmatic value.
|
Site 1138
Basement recovered at Site 1138 was divided into 22 igneous units (Coffin et al., 2000
), including cobbles of aphyric to sparsely feldspar-phyric dacite (Unit 1; shipboard XRF major element data for this unit are presented in Table 1) and an interbedded lithic breccia–volcanic ash sequence of 24·3 m thickness (Unit 2) containing completely altered clasts of highly plagioclase–clinopyroxene-phyric basalt. Units 3–22 represent a sequence of tholeiitic basalt flows, some with oxidized flow tops, interpreted as being erupted subaerially (Coffin et al., 2000; Keszthelyi, in preparation). The basalts are aphyric to sparsely plagioclase–clinopyroxene-phyric, with plagioclase present either as solitary crystals or as glomerocrysts. Olivine microphenocrysts (partially to completely replaced by clay) are present in Units 5–16 and 19 and minor titanomagnetite phenocrysts are present throughout.
The Site 1138 basalt data appear to be consistent with a sequence of eruptions from a magma chamber undergoing progressive fractional crystallization, except there is a systematic increase in mg-number from the bottom to the top of the lava sequence (Fig. 6), the stratigraphic opposite of what would be expected. The compatible elements Sc, Cr, and Ni correlate positively with mg-number, as do CaO and Al2O3. MgO (4·7–6·4 wt%), Cr (11·1–101 ppm), and Ni (25·5–67·0 ppm) abundances are low in all of these basalts; thus, all are differentiated well beyond any plausible primary magma compositions. The dacite has low MgO, Fe2O3, and CaO, and relatively high alkali and P2O5 contents relative to the basalts (Table 1).
|
Incompatible element abundances, including TiO2 and P2O5, are negatively correlated with mg-number (Fig. 6). Incompatible element ratios Nb/Ta, La/Nb, and Th/Ta, normalized to estimated primitive mantle values (Sun & McDonough, 1989), give an average value of 1·0 ± 0·1 for the entire suite of 20 basalt units. On primitive mantle normalized and chondrite-normalized plots, the basalt data form subparallel patterns (Fig. 7) with a pronounced depletion in Sr and P, and a slight depletion at Ce and Y. The dacite has a similar pattern to the basalts for elements to the right of Sr, except for a marked depletion in Ti (Fig. 7a) and greater LREE enrichment (Fig. 7b). In contrast to the basalts, however, it exhibits an enrichment of La and Th over Nb and Ta (Fig. 7a).
|
The effect of secondary processes on the Site 1138 basalt units is highlighted by the negative Ce anomaly in many of the samples (Fig. 7b). This is unlikely to be a primary igneous feature (e.g. Neal & Taylor, 1989). Rather, it suggests that relatively oxidizing conditions prevailed during alteration, promoting the oxidation of Ce3+ to Ce4+ and preferential removal of Ce. In addition, the highly variable Ba/Rb ratios (7·6–118·8) are distinct from the average ocean island basalt (OIB) and MORB value of 11 (e.g. Hofmann & White, 1983; Sun & McDonough, 1989) and are probably a result of alteration.
Age-corrected Sr, Nd, and Pb isotope ratios for the basalts exhibit very little variation throughout the sequence and the data generally plot close to the field defined by Kerguelen Archipelago lavas (Figs 3 and 4), at slightly lower 206Pb/204Pb. The one isotopically distinctive igneous unit at Site 1138 is the Unit 1 dacite, which has significantly higher (87Sr/86Sr)t (0·7065) and slightly lower Pb isotope ratios than the basalts, but similar Nd(t) (+0·2).
Sites 1141 and 1142
Despite a distance of only 800 m between Sites 1141 and 1142, the basement sections recovered differ considerably. Seismic reflection data revealed that the lower portion of the sedimentary sequence (i.e. below an erosional unconformity associated with the Eocene break-up of Broken Ridge and the CKP) dips northward in this area, suggesting that the basement section encountered at Site 1142 may be stratigraphically deeper than that at Site 1141 (Coffin et al., 2000). Indeed, shipboard elemental data suggested that Unit 6 at Site 1141 and Unit 1 at Site 1142 may be the same unit (Coffin et al., 2000; Tables 1 and 2). Our results support this interpretation.
Each Broken Ridge site contains six basement units, all of which were erupted subaerially, with the possible exception of Unit 6 at Site 1142 (Coffin et al., 2000). At Site 1141, Unit 1 is composed of three small pieces (2–3 cm) of moderately altered plagioclase–clinopyroxene–olivine microgabbro. Units 2–6 are aphyric to plagioclase- and/or olivine-phyric basalts. Basement units at Site 1142 are basalt to basaltic andesite, and are aphyric to olivine and/or plagioclase phyric except for Unit 2, a probable talus deposit that contains pebbles of highly altered quartz-phyric lavas in a volcanic breccia.
All of the Broken Ridge basaltic units are alkalic, except for Unit 6 at Site 1142 (Coffin et al., 2000). The lavas are relatively fractionated, containing 3·0–7·4 wt % MgO (Table 1). Both the alkalic and tholeiitic basalts are LREE enriched, although the former exhibit more extreme enrichments (Fig. 8a). Two samples analyzed from different pebbles of the Site 1142 talus deposit (Unit 2) have distinct major and trace element compositions (Table 1). Generally, the basalts from Sites 1141 and 1142 are enriched in the incompatible elements relative to the central Broken Ridge dredge samples (which were all tholeiitic basalts) reported by Mahoney et al. (1995) (see Fig. 8). The eastern Broken Ridge dredge samples are similar to the tholeiite at Site 1142 (Unit 6; Fig. 8b). The alkalic basalts exhibit a slight depletion in Y and several also show depletions of Zr and Hf relative to Nd and Sm (Fig. 8a). In addition, Nb and Ta are enriched relative to La and Th, and all exhibit a marked depletion in Th (Fig. 8a) as well as U. The Site 1142 tholeiitic basalt is distinguished from the alkalic basalts by the following features: (1) La and Th are enriched relative to Nb and Ta; (2) (Sr/Nd)PM <1 and Eu/Eu* (where Eu* is the estimated Eu abundance assuming a smooth REE profile) is slightly less than unity. The small sample of the Site 1141 microgabbro (Unit 1) is distinct from the basalts only in that it has a suprachondritic Zr/Hf ratio; this has been replicated in a separate analysis, but its origin is enigmatic.
|
Isotopically, the microgabbro is equivalent to the alkalic basalts from both sites, but the tholeiitic basalt of Unit 6 at Site 1142 is distinct in that it has a slightly lower (206Pb/204Pb)t ratio (Figs 3 and 4 and Table 2). In terms of Sr and Nd isotopes, the Site 1141 and 1142 samples are nearly indistinguishable (Table 2), and plot roughly in the middle of the range defined by the dredged basalts from the central and eastern Broken Ridge. Their Nd(t) values are slightly positive, similar to those of the Site 1138 basalts and the upper two lava units at Site 1136; however, their (87Sr/86Sr)t values are significantly higher (0·7053–0·7056 vs 0·7046). Lead and Sr isotope ratios are similar to those reported for Site 1137 (Ingle et al., 2002b), but the Site 1141 and 1142 basalts exhibit higher (slightly positive) Nd values (Figs 3 and 4). Compared with the Site 1138 lavas and the upper two units of Site 1136, the Site 1141 and 1142 rocks have similar (206Pb/204Pb)t but higher (207Pb/204Pb)t and (208Pb/204Pb)t.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE PREPARATION AND ANALYSISRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Elemental data for basalt lavas erupted over a 24 Myr time span (Site 1136 at 118–119 Ma, Site 1138 at 100–101 Ma, and Sites 1141 and 1142 at
95 Ma; Duncan, 2002) show geochemical heterogeneity typical of the Kerguelen Plateau–Broken Ridge LIP (see Davies et al., 1989; Weis et al., 1989; Salters et al., 1992; Mahoney et al., 1995). This heterogeneity is generally interpreted in terms of magma generated from a heterogeneous plume source and which occasionally assimilated small amounts of crustal material. However, the isotope data reported here exhibit surprisingly little overall variation in (206Pb/204Pb)t and Nd(t) and only modest variation in (87Sr/86Sr)t.
Southern Kerguelen Plateau (SKP)
Four ODP sites drilled into the SKP (Fig. 1) recovered basaltic basement. The basalt units recovered exhibit a wide range of incompatible element and isotopic compositions (Figs 3, 4, and 5a and b; Salters et al., 1992; Storey et al., 1992; Frey et al., 2002). Basement from Site 738 is characterized by a pronounced continental crustal or lithospheric mantle signature (e.g. Storey et al., 1992; Mahoney et al., 1995; Fig. 5b), very different from anything seen at Sites 749 and 750 (see, e.g. Salters et al., 1992; Frey et al., 2002). Frey et al. (2002) have invoked a plagioclase-rich lower-crustal end-member to explain the compositions of the Site 750 lavas, including their relatively non-radiogenic Pb isotopes and [Sr/Nd]PM >1. Primitive mantle normalized patterns show that unlike the basalts from Sites 749 and 750, those from Site 1136 (and Site 738) lack peaks at Sr. The three Site 1136 units exhibit Nb depletions, but smaller than those of Site 738 lavas. Unit 3 at Site 1136 exhibits a smooth decrease from Ce to Nb, is enriched in the REE and Y relative to the other two units, and also is distinct isotopically (Figs 3 and 4). Tantalum is enriched relative to Nb in Units 2 and 3 (i.e. Nb/Ta = 12·1–12·6, versus a chondritic value of 17·4 in Unit 1), qualitatively similar to six of the 11 Site 749 and Site 750 lavas studied by Frey et al. (2002), which have Nb/Ta = 12·1–14·8 and overlap with values estimated for average continental crust (e.g. Rudnick & Fountain, 1995; Barth et al., 2000; Fig. 5c).
Although seismologic data suggest that some continental crust is present in the northern part of the SKP (Operto & Charvis, 1996), the geochemical data now available indicate that both the influence and types of continental material affecting magmas in the SKP are highly variable. The influence of continental crust and perhaps lithospheric mantle is emphasized when (La/Nb)PM and (La/Ta)PM ratios are plotted against 
7/4 and 8/4 (Fig. 9). Although highly positive 7/4 and 8/4 values do not by themselves indicate the involvement of continental material, they are strongly suggestive of such an influence if accompanied by (La/Nb)PM and (La/Ta)PM values significantly greater than one (e.g. Mahoney et al., 1995; Frey et al., 1996).
|
This is highlighted by the comparison of basalts from each of the SKP drill sites. Site 738 basalts have the highest 7/4 and 8/4 values and the highest (La/Nb)PM and (La/Ta)PM values of 2. Site 749 basalts have the lowest 7/4 and 8/4 values (although they are still positive) and (La/Nb)PM 1·1, with (La/Ta)PM 1. Basalts from Sites 750 and 1136 have similar 7/4 and 8/4 values, but the former have (La/Nb)PM and (La/Ta)PM values 1. Site 1136 basalts all have (La/Nb)PM between 1·3 and 1·6 and (La/Ta)PM between 1·0 and 1·3 (Fig. 9). The Site 1136 (La/Nb)PM and 7/4 values are similar to those of the Casuarina group of Bunbury Basalt lavas, which have been interpreted as being slightly influenced by continental material (Frey et al., 1996). Units 1 and 2 lavas have elevated (207Pb/204Pb)t relative to values for the Site 1138 basalts (see below); their relatively high (La/Nb)PM and the marked Pb peaks in their incompatible element patterns (Fig. 5a) are also consistent with minor involvement of continental material of some sort. We conclude that the compositions of Site 1136 basalts likewise have been influenced slightly by continental material.
The highly altered nature and poor recovery of Unit 3 hinder the interpretation of its composition. The elevated REE and Y in this sample are not a result of secondary apatite or calcite because this sample has the same P2O5 and Sr contents as other Site 1136 basalts (Table 1; Fig. 5a). Unit 3 has the most positive Nd(t) value (+5·1) of any sample analyzed in this study (Table 2). The high Nd is unlikely to be caused by alteration; seawater has negative Nd values, and Nd isotopes are resistant to even relatively high levels of alteration. Similarly, the normalized incompatible element profile is difficult to reconcile with alteration. We suggest that Unit 3 at Site 1136 represents a unique component in the source region for this area of the Kerguelen LIP.
Central Kerguelen Plateau (CKP)
Tholeiitic to transitional basalts have been recovered from two drill sites on the CKP at Site 747 and Site 1138, both to the SE of Heard Island (Fig. 1; Storey et al., 1992; Frey et al., 2002). Unlike the Site 747 lavas, those from Site 1138 do not have significant relative depletions in Nb, Ta or Th, and the Site 747 basalts have much lower Nd(t) and (206Pb/204Pb)t and higher (87Sr/86Sr)t (Figs 3 and 4). Frey et al. (2002) argued that the Site 747 magmas assimilated appreciable amounts of old, lower continental crust, to account for the unusual isotopic characteristics (Figs 3 and 4), the relatively low abundances of Th, Nb, and Ta, and the combination of positive 7/4 and 8/4 values in conjunction with (La/Nb)PM and (La/Ta)PM > 1 in these basalts (Figs 7a and 9). In view of the isotopic characteristics and the lack of Nb, Ta, or Th depletions in Site 1138 lavas (Figs 3, 4, 7a, and 9), we infer that there is no significant influence of continental crust in these basalts.
The Site 1138 basalts appear to have experienced substantial plagioclase (Sr and Eu depletions), clinopyroxene (Y depletion), and olivine (25–67 ppm Ni; Table 1) fractionation, consistent with petrographic observations. The resultant geochemical variations at first sight suggest an evolving magma in a closed system. This is indicated by the combination of: (1) the very small amount of Sr, Nd, and Pb isotopic variation (Figs 3 and 4 and Table 2); (2) the sub-parallel incompatible element profiles (Fig. 7); (3) a general increase in incompatible element and decrease in compatible element abundances going through the sequence from Unit 3 to Unit 22 (Fig. 6); (4) the lack of any obvious continental lithospheric signature (Figs 7a and 9). However, the range in incompatible element abundances cannot be generated by fractional crystallization of a single magma and, as noted above, the sequence is the stratigraphic opposite of what would be expected. The basalt sequence can be successfully modeled by periodic replenishment of a fractionating basaltic magma (e.g. RFC: O’Hara & Matthews, 1981). The low MgO values of the basalts suggest that the incoming magma was probably not primary, also suggesting that there were multiple magma chambers beneath Site 1138. As noted above, petrography indicates that plagioclase and clinopyroxene are the main phenocryst phases (see Fig. 10) in the sparsely phyric basalts; olivine is a minor phase. Detailed petrography of the opaque phases indicates that titanomagnetite was also a minor yet persistent fractionating phase. Toplis & Carroll (1995) demonstrated that the crystallization of titanomagnetite was dependent on ferric iron abundance requiring an oxygen fugacity at FMQ or higher. The experiments of Toplis & Carroll (1995, 1996) are particularly relevant here as starting compositions are similar to the Site 1138 basalts [i.e. compare SC-1 of Toplis & Carroll (1995) with Unit 7 of Site 1138]. The negative correlation between total Fe and SiO2 (Fig. 11a) has a similar, although slightly steeper, gradient to that proposed by Toplis & Carroll (1995) for the onset of magnetite saturation in basaltic magmas. However, this would require that the highest SiO2 magmas would be the most evolved, yet they have the highest mg-number. Conversely, there is a positive correlation between total Fe and TiO2 (Fig. 11b), and TiO2 increases with decreasing mg-number (Fig. 6). Therefore, the proportion of titanomagnetite on the liquidus was not large enough (i.e. <10%) to promote a decrease in TiO2 or total Fe contents of the magma. Rather, the crystallization of this phase tempered the increase of Fe (see Toplis & Carroll, 1996, fig. 2a) and TiO2. Finally, as phenocrysts exhibit resorption features, it suggests titanomagnetite fractionation was inefficient.
|
|
Although the downhole correlations are coherent in the basalt lava sequence at Site 1138, element–element correlations are diffuse, suggesting processes other than fractional crystallization are occurring. It may be that this negative correlation between total Fe and SiO2 represents olivine resorption coupled with the onset of titanomagnetite crystallization (see Toplis & Carroll, 1995; see fig. 12a of Toplis & Carroll, 1996), which produces a steeper gradient on iron–silica correlation plots such as that observed with our data. Such a process would explain the fact that the highest SiO2 magmas have the highest mg-number. Indeed, olivine phenocrysts exhibit resorption features, which we interpret as a result of magmatic evolution that resulted in olivine instability such that it reacted with the magma. At high oxygen fugacities, magnetite is a crystalline product (see Toplis & Carroll, 1995). However, petrography also demonstrates resorption features on clinopyroxene and titanomagnetite phenocrysts, especially in Units 3–10. These we interpret as indicating an influx of new magma into the chamber.
|
, parental magma composition, which is the composition that periodically replenishes the system (Cr 85 ppm; Sc 56 ppm; Ti 12 000 ppm; Zr 100 ppm; La 9 ppm). The continuous line linking the open squares indicates how the actual compositions of successive basaltic units from Site 1138 change. The open circles linked by a continuous line represent individual replenishments after periods of fractional crystallization during which there is no resorption of crystallized phases. The fractionating assemblage is plagioclase, clinopyroxene, and titanomagnetite crystallizing in the proportions 32:60:8. The mass crystallized during each RFC stage is 15%, the mass of magma erupted is 8% and the mass of magma added is 10%. The dashed lines emanating from each open circle indicate how the magma composition is changed as crystalline phases (olivine, clinopyroxene, and titanomagnetite; see text for discussion) are resorbed in the proportions 1:3:6. The furthest end of each of these dashed lines from the open circle represents 5% resorption. Using this model, the Site 1138 basalt sequence can be approximated by eight replenishment cycles followed by variable (0–5%) resorption of the crystallized phases. Partition coefficients were taken from the compilations of Rollinson (1993)
In our RFC model (Fig. 12), we have used petrography as a guide in establishing a fractionating assemblage of plagioclase, clinopyroxene, and titanomagnetite crystallizing in the proportions 32:60:8, which is consistent with the behavior of Eu, Sc, Ni, and TiO2 abundances as well as with Sc/Sm and (Sr/Nd)PM ratios in the lava sequence (Figs 6 and 10; Table 3). The mass crystallized during each RFC stage is 15%, the mass of magma erupted is 8%, and the mass of magma added is 10%; this combination of parameters represents a shrinking or ‘dying’ magma chamber, consistent with the drilled sequence being the final set of eruptions in this area of the CKP. The model parameters, including the parental magma composition, have been chosen to best fit the data. We envisage that equilibrium would be disturbed by an influx of parental magma into the chamber, changing the magma composition and causing clinopyroxene and titanomagnetite to be resorbed. The scarcity of olivine would be a result of it being carried up from a lower magma chamber and then resorbed through reaction with the more evolved magma and promoting titanomagnetite crystallization (see Toplis & Carroll, 1995). Therefore, it is not included as a fractionating phase in our model but is treated as a resorbed phase; the phases being resorbed are olivine, clinopyroxene and titanomagnetite in the proportions 1:3:6 (Table 3) and for simplicity we model resorption as a single-stage process. The resorption process accounts for the modest change in Sc (Fig. 6), even though clinopyroxene forms 60% of the fractionating sequence. What is immediately noticeable is that when the Site 1138 basalt data are represented on element–element plots, they do not define a smooth trend from Unit 3 to Unit 22 (Fig. 12). This disjointed sequence of elemental variations is reproduced in our model by varying the amount of phenocryst resorption. Although these results are clearly model dependent, they illustrate that RFC processes can generate the basalt sequence at Site 1138.
|
The model does not, however, necessarily account for the fact that the entire Site 1138 basalt sequence is the stratigraphic opposite of what would be expected in that the least differentiated units were erupted last (i.e. are at the top; Fig. 6). The criteria used for defining unit boundaries and the orientation of vesicle trains indicate that the sequence is not overturned (Coffin et al., 2000; Keszthelyi, in preparation). A possible explanation could be that the lowermost unit recovered at Site 1138 represents a more highly evolved basalt formed after an extended period of fractional crystallization without replenishment. Later flows would then represent successive influxes of comparatively primitive parental magma, the amount of which would exceed the amount crystallized during the preceding stage (i.e. the reverse of the sequence modeled above). Consecutively more primitive-looking basaltic magmas would be erupted even while phenocrysts continued to be resorbed. This can be modeled by using the most evolved basalt from Site 1138 as the initial magma composition, with the assumed parent from the model defined in Fig. 12 as the incoming magma. Resorption of phenocryst phases and a decrease in the amount of fractional crystallization are required to generate the observed basalt compositions using this approach. Whatever the details of the process(es), it is clear that the system was isotopically essentially homogeneous and that multiple stages of differentiation were required.
The Unit 1 dacite is isotopically distinct from the basalts (Figs 3 and 4). The horizontal displacement of the dacite from the basalts on an Nd(t) vs (87Sr/86Sr)t diagram (Fig. 3) is unlikely to be caused solely by seawater alteration that was not removed by acid leaching, because the dacite has a high Sr concentration (179·5 ppm in the leached split; 187 ppm in the whole rock) relative to that of seawater (8 ppm) and the Sr isotope ratio of 100 Myr old seawater was only 0·7073 (e.g. Burke et al., 1982), close to the 0·7065 value of the dacite itself. The trace element data (Table 1; Fig. 7) show that the dacite is more LREE enriched than the basalts, exhibits a larger Sr depletion, and exhibits depletions in Nb and Ta relative to La. We speculate that the dacite could be the product of crystal fractionation of a basaltic magma like those represented by Units 3–22, as oxide fractionation can potentially generate magmas with (La/Nb)PM and (La/Ta)PM ratios slightly above unity (Rollinson, 1993; Green, 1994). However, the magnitude of the LREE enrichment in the dacite is slightly greater than can be achieved from fractional crystallization alone. Its high (87Sr/86Sr)t and lower Pb isotope ratios suggest open-system crystal fractionation and possibly reflect the influence of some continental lithospheric material.
Broken Ridge
The presence of alkalic lavas overlying a tholeiitic basalt at Site 1142, coupled with the fact that all of the igneous rocks dredged from Broken Ridge are tholeiitic basalts, suggests that a main tholeiitic phase of volcanism may have been capped by a late-stage alkalic carapace in the Broken Ridge portion of the original CKP. The sequence of volcanism at Broken Ridge would thus be similar to that observed in some ocean islands. For example, late-stage alkalic lavas commonly cap main-stage tholeiitic (shield-building) basalts on Hawaiian volcanoes (e.g. Clague & Dalrymple, 1987). At Broken Ridge, the small amount of variation in, for example, Zr/Y, Sc/Sm, and La/Yb (i.e. Fig. 13), is interpreted to represent a change in degree of partial melting of a spinel peridotite source, rather than a change to a deeper, garnet-bearing source with a concomitant decrease in degree of partial melting. One sample with distinctly high (La/Yb)PM and Zr/Y (Fig. 13) is from the Site 1142 talus deposit (Unit 2).
|
The alkalic basalts recovered from Broken Ridge during Leg 183 are distinct from the dredged tholeiites in that they contain higher abundances of Nb and Ta and also a marked Th depletion (Fig. 8). However, tholeiitic Unit 6 and a pebble from the top sample of Unit 2 from Site 1142 are similar to the dredge samples in that they have (La/Nb)PM > 1 and (Nb/Th)PM < 1 (Fig. 8). The dredge samples exhibit a broader range of Sr, Nd, and Pb isotopic compositions [e.g. Nd(t) = -2·7 to +3·4] that encompasses the small range defined by the Site 1141 and 1142 samples [which all have, for example, Nd(t) = +0·3 to +0·7; Fig. 3].
Mahoney et al. (1995) interpreted the low-Nd(t) dredge samples, particularly from Dredge 8 on the eastern Broken Ridge, to contain continent-derived material on the basis of combined Th/Ta, La/Nb, La/Ta, and 7/4 and 8/4. The similarity in incompatible element profiles of the Unit 6 tholeiite from Site 1142 to the eastern Broken Ridge dredge samples (Fig. 8b) suggests it may also contain a small amount of continent-derived material. The two samples analyzed from this unit have (La/Ta)PM and (La/Nb)PM > 1, similar to the dredge samples, the Bunbury Basalt, and some Naturaliste Plateau lavas (Fig. 9). The Broken Ridge lavas recovered during Leg 183 have similar (206Pb/204Pb)t to those from Site 1137 (Figs 3 and 4), which are considered to contain a component of continental crust (Ingle et al., 2002a, 2002b). However, except for the tholeiitic Unit 6 of Site 1142, the lavas from Sites 1141 and 1142 have lower (207Pb/204Pb)t and (208Pb/204Pb)t (Fig. 4 and Table 2). The lack of a continental influence in the alkalic basalts is emphasized by trace element ratios (Fig. 9). The alkalic samples from Sites 1141 and 1142 have (La/Ta)PM and (La/Nb)PM 1, but like Kerguelen Archipelago lavas, still have elevated 7/4 and 8/4. Such features are likely to be intrinsic to the Kerguelen plume (see Weis et al., 1989, 1991; Frey & Weis, 1995, 1996).
End-members in Kerguelen Plateau development
Most recent workers have argued that the Kerguelen Plateau was constructed during the early stages of activity of a long-lived mantle plume (e.g. Davies et al., 1989; Weis et al., 1989, 1991; Coffin, 1992; Storey et al., 1992; Frey & Weis, 1995; Mahoney et al., 1995). As noted above, the geochemistry of the magmatic products of the Cretaceous–Recent Kerguelen hotspot system generally has been described in terms of variable mixtures of a dominant plume component (with Nd near zero), a secondary plume component (isotopically similar to Amsterdam–St. Paul), possibly a small amount of Indian MORB-type mantle, and the incorporation of variable amounts of heterogeneous continental lithospheric material. The isotope data reported here are from rocks that span 24 Myr yet exhibit little variation in (206Pb/204Pb)t and Nd(t) and, except for the Site 1138 dacite, only modest variation in (87Sr/86Sr)t. A much wider range of Nd and Sr isotopic values is documented for other Kerguelen Plateau and Broken Ridge rocks, as well as Naturaliste Plateau and Rajmahal Traps lavas. Together with our Leg 183 rocks, these same lavas also encompass a wide range of 208Pb/204Pb and 207Pb/204Pb values. Although not all values are age-corrected, SKP lavas from Sites 738, 749, and 1136, CKP lavas from Sites 1137 and 1138, Broken Ridge lavas from Sites 1141 and 1142, and many of the dredged samples, Rajmahal basalts, and Naturaliste Plateau lavas are characterized by surprisingly limited variation in 206Pb/204Pb, between about 17·8 and 18·1. This range probably represents the 120–90 Ma composition of the dominant material of the Kerguelen hotspot mantle source. Moreover, much of the continental-type material that affected many of these magmas, such as those at Sites 738 (Mahoney et al., 1995) and 1137 (Ingle et al., 2002b) must have had 206Pb/204Pb in broadly this same range.
To further evaluate the different components involved in magmatism related to the Kerguelen hotspot, we use incompatible element ratios in conjunction with isotope data. The element ratios used are relatively insensitive to moderate to large degrees of partial melting and to fractional crystallization processes, and thus should approximately represent source compositions or, in the case of mixed magmas, should represent mixtures of end-member source ratios. In general, Nb/Y vs Zr/Y data for the basalts from Sites 1136, 1138, 1141, and 1142 plot in the Iceland field in Fig. 14a (see Fitton et al., 1997). The CKP basalts have generally higher Zr/Y and Nb/Y than basalts from the SKP basement sites. Probable basaltic products of the Kerguelen hotspot known to reflect a continental lithospheric influence, such as the Bunbury Basalt (Frey et al., 1996), Rajmahal Traps (Kent et al., 1997), Naturaliste Plateau, and Site 738 (Storey et al., 1992; Mahoney et al., 1995) define sub-parallel trends extending below the Iceland array. Data for hawaiites from Afanasy Nikitin Rise (a small plateau 3560 km NNW of Site 1141), which record a large continental lithospheric mantle or lower-crustal influence (Mahoney et al., 1996; Borisova et al., 2001), fall slightly beneath the Iceland array. Data for the alkalic lavas of Broken Ridge Sites 1141 and 1142 plot near those for St. Paul Island (Frey & Weis, 1995).
|
