Basement sections recovered from 11 drill sites on the Kerguelen Plateau and Broken Ridge (i.e., Leg 119, Site 738; Leg 120, Sites 747, 749, and 750; and Leg 183, Sites 1136, 1137, 1138, 1139, 1140, 1141, and 1142) (Fig. F1) are dominantly tholeiitic basalt (8 of 11 sites) with moderate MgO contents (5-8 wt%) (Figs. F5, F6). Their incompatible element abundance and radiogenic isotopic ratios (Sr, Nd, and Pb) distinguish these LIP lavas from mid-ocean-ridge basalt (Figs. F7, F8, F9). However, at each drill site on the LIP the geochemical characteristics of the basalt require very significant differences in petrogenesis. These differences result largely from (1) changing proportions in source components, namely plume, depleted asthenosphere, and continental lithosphere; (2) geochemical heterogeneity in the source components, especially the plume and continental lithosphere; (3) variable extents of melting; and (4) variable extents and types of postmelting magmatic processes. Following, in order of decreasing eruption age, we discuss the petrogenesis of lavas from each site.

Site 1136: Southern Kerguelen Plateau

Prior to Leg 183, igneous basement of the SKP had been recovered by dredging (Leclaire et al., 1987) and drilling during Legs 119 and 120 (Fig. F1). Basalt from Site 738, the southernmost sampling location, has a tholeiitic composition, but its trace element characteristics and isotopic ratios (Sr, Nd, and Pb) (Figs. F7, F8, F9, F10) clearly reflect a component derived from continental lithosphere, probably crust (Alibert, 1991; Mahoney et al., 1995). Continental crust, especially upper crust, is distinguished by relatively high 87Sr/86Sr, low 143Nd/144Nd, high 207Pb/204Pb and 208Pb/204Pb at a given 206Pb/204Pb, and relative depletions in abundances of Nb and Ta. In particular, Mahoney et al. (1995) used plots of 7/4 and 8/4 vs. La/Ta (or equivalently, La/Nb) to show convincingly that SKP basalt from Site 738 contains a significant proportion of a continental component (Fig. F10). Although recycled continental lithosphere could be intrinsic to a mantle plume, the extreme geochemical characteristics of Site 738 basalt (Figs. F7A, F8, F9, F10) are compatible with an alternative hypothesis; that is plume-derived magma assimilated continental lithosphere, probably crust, that was dispersed as fragments into the Indian Ocean asthenosphere or lithosphere as a result of Gondwana breakup (Mahoney et al., 1995; Frey et al., 2000a; see fig. 14c of Frey et al., 2002b). A continental component is not obvious in basalt from the more northerly Sites 749 and 750 on the SKP (Figs. F7C, F8, F9, F10), but Storey et al. (1989) suggested that the relative depletion in Ta (and Nb) in some dredged basalt from the northern SKP may reflect incorporation of sub-Gondwana continental lithosphere into the asthenosphere beneath the Indian Ocean. Consequently, a specific goal of drilling at Site 1136 in the southeast part of the SKP was to evaluate the areal extent of continental lithosphere components in the uppermost igneous basement of the SKP.

The 33.3 m of basement penetration at Site 1136 (Fig. F2) includes three flow units of tholeiitic basalt that were studied by Neal et al. (2002). 39Ar/40Ar isotope analyses for plagioclase from Units 1 and 2 yield the oldest ages (~119 Ma) found on the Kerguelen Plateau (Duncan, 2002). The field defined by 87Sr/86Sr and 143Nd/144Nd of Units 1 and 2 is slightly offset from the field defined by Cenozoic oceanic island basalt associated with the Kerguelen plume (e.g., the Kerguelen archipelago), and 206Pb/204Pb is clearly lower in Site 1136 basalt than flood basalt from the archipelago (Figs. F8, F9); this offset to lower 206Pb/204Pb is a characteristic difference between Cretaceous Kerguelen Plateau basalt and the lavas forming the Cenozoic Kerguelen archipelago (Fig. F9).

In contrast to basalt from Site 738, the isotopic characteristics and relative abundance of Nb and Ta show that Site 1136 basalt is not highly contaminated with continental crust (Figs. F7A, F8, F9). Based on a 8/4 vs. La/Nb plot (Fig. F10), Neal et al. (2002) concluded that basalt from Units 1 and 2 of Site 1136 "have been influenced slightly by continental material." Although the limited recovery, 53 cm, of only highly altered rubbly flowtop hinders detailed interpretation, Unit 3 is isotopically distinct from Units 1 and 2 (e.g., Figs. F8, F9); apparently Unit 3 was derived from a different source than Units 1 and 2 (Neal et al., 2002).

Site 1137: Elan Bank

Site 1137 is on the previously unsampled Elan Bank, a large western salient of the main Kerguelen Plateau that is flanked on three sides by oceanic crust of the Enderby Basin (Fig. F1). The ~152-m basement sequence, composed of seven basaltic flow units and three volcaniclastic sedimentary units (Fig. F2), provided several unanticipated results. Basement Units 1-4, 7, 8, and 10 are ~107- to 108-Ma dominantly tholeiitic to transitional basalt (Fig. F5) with no discernible downhole age progression (Duncan, 2002). However, basement Units 5, 6, and 9 are volcaniclastic sedimentary rocks. Unit 9 is ~17 m of altered crystal-vitric tuff containing ~40% 1- to 2-mm angular crystals of sanidine and quartz. Unit 5 is ~4.4 m of sandstone, dominantly lithic volcanic fragments (75%), but also including feldspar (15%), quartz (5%), and garnet (1%). Unit 6 is ~31 m of conglomerate including clasts ranging from granules to boulders of trachyte, rhyolite, basalt with varied phenocryst assemblages (Fig. F5), and, most surprisingly, rounded cobbles of garnet-biotite gneiss and granitoid. The depositional environment of basement Units 5 and 6 appears fluvial, perhaps associated with a braided river.

Nicolaysen et al. (2001) used U-Pb and Pb-Pb dating techniques to obtain a wide range of ages (534-2457 Ma) for zircon and monazite separated from garnet-biotite clasts in Unit 6 and the overlying sandstone of Unit 5. These dates show that old continental crust dispersed during the breakup of Gondwana resides in the shallow crust of the Kerguelen Plateau. Age constraints (Nicolaysen et al., 2001) and isotopic data for the gneiss clasts (Ingle et al., 2002a) are consistent with a source that originated in the Eastern Ghats of eastern India. Isolation of the Elan Bank microcontinent is inferred to have occurred during a northward ridge jump (i.e., toward India) of the early spreading center in the Enderby Basin (see discussion in "The Kerguelen Hotspot and Indian Ocean Plate Reconstructions").

Garnet grains within the gneiss clasts are Fe rich (77-85 mol% almandine) (Nicolaysen et al., 2001), but garnet grains within the sandstone of Unit 5 span a larger range (77-35 mol% almandine). As a result, Reusch and Yates (this volume) suggest that the source area sampled by these sedimentary units included a range of metamorphic facies and variety of bulk compositions, as expected in a continental region composed of pelitic and granitic materials.

Geochemical studies of the conglomerate matrix and diverse clast types by Ingle et al. (2002a) showed that the felsic volcanic clasts (trachyte and rhyolite) are not genetically related to the intercalated basalt. Based on Sr, Nd, and Pb isotopic data, Ingle et al. (2002a) concluded that the felsic volcanic clasts formed by partial melting of evolved upper continental crust.

Prior to drilling at Site 1137, Mahoney et al. (1995) inferred that the continental component in basalt from Site 738 was introduced into plume-derived magmas at shallow depths rather than deeply recycled crust intrinsic to the Kerguelen plume. This inference is supported by the presence of continental crust clasts in the conglomerate at Site 1137. Moreover, at Site 1137 the uppermost basaltic units are largely plume derived, but the plume-derived lowermost basaltic units apparently assimilated continental lithosphere (Figs. F7A, F10) (Weis et al., 2001; Ingle et al., 2002b) What is the proportion of continental component in these contaminated lavas? Ingle et al. (2002b) calculated that the isotopic trends of Site 1137 basalt can be explained by 5% to 7% assimilation of material similar to the garnet biotite gneiss. In detail, these clasts do not have Pb isotopic ratios that are suitable for explaining the trend in Pb isotopic ratios defined by Site 1137 basalt (Fig. F9). Also, the Pb isotopic data show that the crustal components contributing to Site 738 and 1137 basalt are geochemically different and that both differ in Pb isotopic ratios from the clasts of garnet biotite gneiss (Fig. F9) (Weis et al., 2001). Given the heterogeneity of continental crust, this result is not surprising. Ingle et al. (2002b), however, emphasized the striking similarities in Pb isotopic trends between basalt erupted at Site 1137 and the continental flood basalt forming the Rajmahal Traps in northeast India and Bunbury Basalt in Southwest Australia. Although these locations are now widely disparate (Fig. F1), they were closer (~1000 km apart) at their time of formation (Fig. F4), and Ingle et al. (2002b) argued that in each case magmas derived from the Kerguelen plume interacted with Gondwana crust having similar Pb-Pb isotopic characteristics.

Site 1138: Central Kerguelen Plateau

A major objective at this drill site was to determine the basement age of the CKP; no reliable age information is available for basalt from Site 747, the only previous drill site on the CKP (Figs. F1, F2). At Site 1138, 22 units of igneous basement were recognized in the 144 m of recovered rock. Particularly surprising were two units of felsic igneous rock overlying tholeiitic basalt (Fig. F2), that is Unit 1, cobbles of flow-banded dacite, and Unit 2, an ~20-m-thick volcaniclastic section interpreted as subaerial pyroclastic flow deposits. No evidence of compositionally evolved, explosively erupted volcanic deposits, such as those at Sites 1137 and 1138, was found during previous drilling on the Kerguelen Plateau. Although the dacite at Site 1138 (Unit 1) may have largely formed by extensive fractional crystallization of magmas similar to the underlying basalt, differences in Sr and Pb (but not Nd) isotopic ratios suggest the possible influence of continental material (Neal et al., 2002). Processes involved in creating these SiO2-rich igneous units in uppermost igneous basement range from partial melting of continental crust fragments within the oceanic lithosphere (e.g., at Site 1137 the felsic clasts in Unit 6 and the sanidine-rich felsic tuff of Unit 9) (Weis et al., 2001; Ingle et al., 2002a) to combined fractional crystallization and assimilation of continental crust fragments (Unit 1 of Site 1138) (Neal et al., 2002).

Underlying the nonbasaltic rocks at Site 1138 are 20 units of ~5-m-thick tholeiitic to transitional basalt (Fig. F5), some with oxidized flow tops, which are interpreted as subaerial eruptives. 40Ar/39Ar data for plagioclase and whole-rock basalt from two flow units indicate an age of 100-101 Ma (Duncan, 2002). These basaltic units have 87Sr/86Sr, 143Nd/144Nd, and 206Pb/204Pb ratios that overlap those of lavas from Units 1 and 2 at Site 1136 on the SKP (Figs. F8, F9), suggesting that basalt sources were similar over a large area and over an ~20-m.y. time span. In contrast with basalts from Sites 738, 1136, and 1137, the Site 1138 lavas are not relatively depleted in Nb (Figs. F7E, F10). Consequently, Neal et al. (2002) concluded that there is no significant influence of continental crust in Site 1138 basalt.

The basalt units at Site 1138 increase in Mg/Fe from the lowermost to uppermost unit (i.e., Mg/Fe increases with decreasing eruption age) (Shipboard Scientific Party, 2000; Neal et al., 2002). With decreasing Mg/Fe, the abundance of incompatible elements increases (e.g., TiO2) (Fig. F6) and the abundance of compatible elements decreases. Neal et al. (2002) used major and trace element abundances in these basaltic units to show that the data are inconsistent with closed-system fractional crystallization. They concluded that the compositional variations of Site 1138 lavas can be explained by an open-system process involving periodic replenishment of a fractionating basaltic magma with plagioclase and clinopyroxene as major fractionating phases; for example, the relative depletion in Sr (Fig. F7E) reflects plagioclase fractionation. Clearly, during the final growth stage of the CKP, the magma flux from the mantle was sufficiently low to enable extensive fractional crystallization of basaltic magma and eventually the formation of dacitic magma by combined fractional crystallization and assimilation.

Sites 1141 and 1142: Broken Ridge

Broken Ridge and the CKP formed as a single entity during Cretaceous time, but they were separated at ~40 Ma by the newly formed SEIR (e.g., McKenzie and Sclater, 1971; Houtz et al., 1977; Mutter and Cande, 1983). The igneous basement of central and eastern Broken Ridge has been sampled at three locations by dredging (Fig. F1). The dredged rocks are tholeiitic basalt that yielded ages from ~61 to 89 Ma (Duncan, 1991); these ages are no longer considered reliable (Duncan, pers. comm., 2002). They are quite variable in their geochemical characteristics, but basalt from eastern Broken Ridge (Dr 8) contains a continental component (Fig. F10) (Mahoney et al., 1995).

At Sites 1141 and 1142 our goal was to obtain in situ basement in order to document more fully the age of Broken Ridge and to document further the areal extent of basalt with isotopic and trace element characteristics that indicate the presence of a continental component. During operations at Site 1141 the drill string became stuck, and the hole was abandoned after only ~72 m of basement penetration, which included six basement units (Fig. F2). At Site 1142, only ~800 m away from Site 1141, ~51 m of basement penetration recovered six basement units (Fig. F2).

39Ar/40Ar isotope data of whole-rock samples from Sites 1141 and 1142 yield ages of 94-95 Ma (Duncan, 2002), significantly older than the ages inferred for the dredged basalt. In contrast to the dredges on Broken Ridge, which recovered only tholeiitic basalt (Mahoney et al., 1995), all of the Site 1141 and 1142, basement units are slightly alkalic, except for the lowermost Unit 6 at Site 1142, which is tholeiitic basaltic andesite (Fig. F5). This alkalic-tholeiitic classification was initially made using a total alkalis-SiO2 plot, and it is confirmed by the generally lower abundance of incompatible elements in the tholeiitic basaltic andesite (Neal et al., 2002). An important result is that only Unit 6 is relatively depleted in Nb and Ta (Fig. F7B).

Also in contrast with the dredged tholeiitic basalt from Broken Ridge, basalts from Sites 1141 and 1142 span only a small range in 87Sr/86Sr and 143Nd/144Nd, which is at the enriched end (i.e., high 87Sr/86Sr and low 143Nd/144Nd) of the field for Kerguelen archipelago flood basalt (Fig. F8). These ratios overlap with the field proposed for the Cenozoic Kerguelen plume and may represent the major component in the plume tail (Neal et al., 2002). Relative to the alkalic basalt units at Site 1141 and 1142, the tholeiitic basalt of Unit 6 at Site 1142 has the lowest 206Pb/204Pb and highest 207Pb/204Pb and 208Pb/204Pb (Fig. F9). In Pb-Pb isotopic plots, Unit 6 lavas are within the field for the lowermost basaltic units at Site 1137 (Fig. F9).

At 206Pb/204Pb of ~18, basalt from several sites define a near-vertical trend in plots of 207Pb/204Pb and 208Pb/204Pb vs. 206Pb/204Pb (Fig. F9) (i.e., 208Pb/204Pb increases from Site 749 to Site 1138 to Site 1136 to Sites 1141/42 and Site 1137) (Fig. F9). This increase in 208Pb/204Pb, as measured by 8/4, is correlated with a relative deficiency in Nb, as measured by (La/Nb)PM (Fig. F10); relatively high 8/4 and (La/Nb)PM indicate an increasing role for a continental component (Mahoney et al., 1995; Ingle et al., 2002b; Neal et al., 2002). Among the Site 1141 and 1142 basement units, only Unit 6 from Site 1142 is characterized by high 8/4 and (La/Nb)PM > 1 (Fig. F10). This unit, like Drو samples from eastern Broken Ridge ~100 km to the east, may contain a continental component (Neal et al., 2002). In summary, the geochemical variability of Broken Ridge basaltic basement documented by basalt from three dredge sites and two drill sites shows that the uppermost basement at Broken Ridge is highly heterogeneous, much like the igneous basement of the CKP and SKP.

Site 1139: Skiff Bank

Skiff Bank is a bathymetric and gravimetric high on the NKP, ~350 km west-southwest of the Kerguelen archipelago, that is bathymetrically continuous with the NKP (Fig. F1). Prior to Leg 183, Skiff Bank was interpreted to be related to the NKP. Also prior to Leg 183, the submarine basement of the NKP was believed to be of Cenozoic age (<40 Ma) and Skiff Bank had been proposed to be the present-day location of the Kerguelen plume (e.g., Duncan and Storey, 1992; Müller et al., 1993). During a predrilling survey cruise for locating Site 1139, Skiff Bank was dredged and a wide range of rock types coated with Fe-Mn crust were recovered (Weis et al., 2002). These rocks have not been studied, but they provide no evidence for recent volcanism.

As at Site 1137 on Elan Bank, the basement recovery at Site 1139 had many surprises. Basement penetration reached ~232 m, and 19 igneous units were identified (Fig. F2). All of the igneous units have <4.5 wt% MgO (Fig. F6). Despite complications arising from intense postmagmatic alteration there is no doubt that the magmas were alkalic in composition (Fig. F5) (Kieffer et al., 2002); for example, excluding elements controlled by fractionating phases, such as Sr by plagioclase, Site 1139 lavas have higher abundances of incompatible elements than all other mafic lavas recovered during Leg 183 (Fig. F7). The basement sequence is bimodal, with a 73-m-thick series of trachybasaltic lava flows sandwiched between felsic (trachyte and rhyolite) volcanic rocks (Figs. F2, F5) (Kieffer et al., 2002). No tholeiitic basalt, the dominant basalt type at all other Kerguelen Plateau drill sites, is present at Site 1139 (Fig. F5). Ar-Ar geochronology of whole-rock samples and feldspar separates from these felsic rocks yield ages of ~68-69 Ma (Duncan, 2002). Clearly, Skiff Bank is not a site of recent volcanism and this part of the NKP is not Cenozoic in age. Furthermore, the magmatism at Skiff Bank was alkalic in composition, like basalt erupted on Cenozoic islands constructed on the plateau, but unlike sampled regions of the CKP and SKP.

Kieffer et al. (2002) inferred that the subaerially erupted lava units at Site 1139 represent part of a shield volcano built on a volcanic plateau of unknown age; shield volcanoes constructed on the Ethiopian plateau may be an analogy. The felsic and mafic lavas at Site 1139 have similar Nd and Pb isotopic ratios (Sr is perturbed by alteration), and Kieffer et al. (2002) infer that felsic lavas formed as partial melts of mafic rocks. In contrast to the intermediate 206Pb/204Pb (~18) and near-vertical 207Pb/204Pb and 208Pb/204Pb vs. 206Pb/204Pb trends defined by basalt from Sites 749, 1136, 1137, 1138, 1141, and 1142, the mafic lavas at Site 1139 have low 206Pb/204Pb, ~17.5, much like basalt from Sites 747 (CKP) and 750 (SKP) (Fig. F9). Site 747 basalt is also depleted in Nb (Fig. F10); therefore, Frey et al. (2002b) inferred that such low 206Pb/204Pb ratios reflect a component derived from lower continental crust. However, like basalt from Site 750, the mafic lavas from Site 1139 are not highly depleted in Nb (Figs. F7E, F10).

Site 1140: Northern Kerguelen Plateau

Site 1140 is on the northernmost part of the NKP, 270 km north of the Kerguelen archipelago (Fig. F1). The boundary between the NKP and Australia-Antarctic Basin is only 5 km north of Site 1140. Major objectives at this site were to test the hypothesis that the uppermost igneous basement of the NKP formed at <40 Ma and to compare the igneous basement with the <30-Ma flood basalt forming the Kerguelen archipelago.

Drilling at Site 1140 penetrated 88 m of basement rocks, which included five units of tholeiitic pillow basalt (Fig. F5) and an interbedded 1-m-thick carbonate unit (Fig. F2). These are the only sampled basement lavas in the LIP that erupted in a submarine environment. Earliest Oligocene (32.8-34.3 Ma) nannofossil and foraminifer oozes overlie Unit 1 (Persico et al., this volume). A magnetic reversal between Units 1 and 2 is inferred to be C13n/C13r (33.6 Ma). These age constraints are consistent with ages of ~34 Ma determined by Ar-Ar geochronology on plagioclase and whole rocks (Duncan, 2002).

At 34 Ma the location of Site 1140 was within 50 km of the SEIR (Weis and Frey, 2002). Each of the five pillow basalt units is geochemically distinct, but they broadly divide into two groups, (1) Units 2 and 3, which are relatively enriched in P2O5 and TiO2 and highly incompatible elements, and (2) Units 1, 5, and 6, which have incompatible element abundances similar to mid-ocean-ridge basalt (MORB) (Figs. F6, F7D). The latter group has lower 87Sr/86Sr, higher 143Nd/144Nd, and lower Pb isotopic ratios (Figs. F8, F9). The isotopic differences between each unit can be explained by mixing between plume-derived magmas (represented by lavas erupted in the Kerguelen archipelago) and SEIR MORB. Consistent with the proximity of this site to the SEIR at the time of eruption (34 Ma), the SEIR MORB component is dominant (60%-99%); the isotopic ratios (Sr, Nd, and Pb) of the youngest lavas (Unit 1) overlap with SEIR MORB (Figs. F8, F9) (Weis and Frey, 2002). In contrast to all other igneous basement recovered from the Kerguelen Plateau, which has 206Pb/204Pb < 18.2, the Site 1140 lavas range to higher 206Pb/204Pb; those with the largest proportion of plume component have Pb isotopic ratios overlapping with basalt forming the Kerguelen archipelago (Fig. F9).

Four of the Site 1140 pillow basalt units have sufficient glass for geochemical studies. Ion microprobe analyses of these glasses by Wallace (2002) also showed that the compositions of these basaltic units can be explained by mixing of MORB and plume-derived magmas. Importantly, these glasses provide information not available from whole-rock studies. In particular, Wallace (2002) found that units with MORB-like Sr, Nd, and Pb isotopic ratios have volatile (H2O, S, and Cl) contents similar to MORB, but the two units with a higher proportion of a plume-derived component have higher water contents (0.44-0.69 wt%). Despite their higher contents of H2O, these units have H2O/Ce ratios less than those of MORB. Apparently, the Kerguelen plume was not a wet spot (Schilling et al., 1980; Bonatti, 1990), requiring that melting in the plume was caused by relatively high mantle temperature and not due to anomalously high mantle H2O contents.