Journal of Petrology Advance Access originally published online on April 28, 2007
Journal of Petrology 2007 48(6):1043-1077; doi:10.1093/petrology/egm010
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Major and Trace Element and Sr, Nd, Hf, and Pb Isotope Compositions of the Karoo Large Igneous Province, Botswana–Zimbabwe: Lithosphere vs Mantle Plume Contribution
1Umr–Cnrs 6526 Géosciences Azur, Université De Nice-Sophia Antipolis, 06108 Nice, France
2Umr–Cnrs 5570, Ecole Normale upérieure De Lyon Et Université Claude Bernard, 69364 Lyon, France
3Berkeley Geochronology Center, 2455 Ridge Road, Berkeley, CA 94709, USA
4Department of Geology, University of Botswana, Gaborone, Botswana
RECEIVED JULY 18, 2006; ACCEPTED FEBRUARY 22, 2007
 |
ABSTRACT |
|---|
We report major and trace element abundances for 147 samples and Sr, Nd, Hf, and Pb isotope compositions for a 36 sample subset of basaltic lava flows, sills, and dykes from the Karoo continental flood basalt (CFB) province in Botswana, Zimbabwe, and northern South Africa. Both low- and high-Ti (TiO2 < 2 wt % and > 2 wt %) rocks are included. MELTS modeling shows that these magmas evolved at low pressure (1 kbar) through fractional crystallization of gabbroic assemblages. Whereas both groups display enrichment in light rare earth elements (LREE) relative to heavy REE (HREE) and high field strength elements, and systematic negative Nb anomalies, they differ in terms of contrasting middle REE (MREE) to HREE fractionation, which is greater for the high-Ti basalts. This reflects different depths of melting of slightly enriched mantle sources: calculations suggest that the low-Ti basalts were generated by melting of a shallow spinel-bearing (2 % spinel) lherzolite, whereas the high-Ti magmas originated from a deeper-seated garnet-bearing (2–7% garnet) lherzolite. In most isotope plots, the high-Ti lavas together with the picrites define a common trend from Bulk Silicate Earth (BSE) to compositions with strongly negative
Ndi and Hfi akin to those of some nephelinites and lamproites. The low-Ti rocks are shifted from BSE-like to more radiogenic Sr isotope ratios, indicative of upper crustal contamination. Trace element and isotope characteristics of the Karoo magmas require a combination of enrichment processes (subduction induced?) and long-term isolation of the mantle sources. We propose two distinct scenarios to explain the origin of the Karoo province. The first calls for polybaric melting of spatially heterogeneous, partially veined, sub-continental lithospheric mantle (SCLM). Calculations show that mixing between SCLM (
BSE) and a strongly Nd–Hf unradiogenic nephelinite-like component (sediment input?) could account for the compositional variations of most of the high-Ti group lavas, whereas the mantle composition responsible for the low-Ti magmas is more likely to be similar to a vein-free, metasomatically enriched SCLM component. The second scenario involves mixing between two end-members represented by the SCLM and its deep-seated alkalic veins and a sub-lithospheric (asthenospheric- or ocean island basalt-like?) mantle plume. In this case, the data are compatible with an increasing mantle plume contribution as the plume rises and expands through the lithosphere. Regardless of which of the two scenarios is invoked, the spatial distribution of the low- and high-Ti magmas matches the relative positioning of the cratons and the Limpopo belt in such a way that strong control of the lithosphere on magma composition and distribution is a mandatory requirement of any petrogenetic model applied to the Karoo CFB. KEY WORDS: Karoo; large igneous province; flood basalts; dyke swarms; major and trace elements; Sr; Nd; Hf; and Pb isotopes
|
INTRODUCTION |
|---|
Continental flood basalts (CFB) are among the most remarkable magmatic events on Earth, consisting of vast outpourings of magma during a relatively brief period of time with potentially dramatic consequences for the biosphere and continental breakup (e.g. Courtillot et al., 1999
; Courtillot & Renne, 2003). The origin of CFBs is a matter of intense controversy and over recent decades a large number of models have been advocated. They range from ‘active’ models, which predict a dominant role for an upwelling deep-seated mantle plume head impinging on the lithosphere (e.g. Campbell & Griffiths, 1990; Hill, 1991), to ‘passive’ models dominated by plate-boundary forces inducing rifting of heterogeneous lithosphere, and/or thermal incubation beneath mega-continents (Anderson et al., 1992; Anderson, 1994). Combined ‘active–passive’ models have also been proposed (Courtillot et al., 1999). Contribution from a deep mantle plume source rooted in the core–mantle boundary layer is advocated in the active models, whereas passive models propose upper mantle sources residing in the sub-continental lithospheric mantle (SCLM) and/or the uppermost enriched part of the asthenosphere [the so-called ‘perisphere’ of Anderson et al. (1992)].
The early Jurassic is marked by the emplacement of one of the largest continental flood basalts (CFBs) on Earth (
3 x 106 km2). The 180 Ma Karoo magmatic province is located in southern Africa, with minor outcrops in Antarctica, and consists of tholeiitic lava flows, sills, and giant radiating dyke swarms emplaced prior to the breakup of southern Gondwana and the opening of the SW Indian Ocean and the Southern Ocean (e.g. Cox, 1988). It was emplaced between 174 and 185 Ma (Fig. 1; Duncan et al., 1997; Jourdan et al., 2005). It consists, at present, of widespread remnants of basalts (lava flows and sills) and giant dyke swarms (Fig. 1).
|
The so-called ‘triple junction’ formed by rift structures and pseudo-radiating dyke swarms (Fig. 1) is a key indicator of the widely supported plume origin for the Karoo CFB province (Burke & Dewey, 1972
; Campbell & Griffiths, 1990; Ernst & Buchan, 2001). However, the role of a plume head impact on the Karoo triple junction has recently been questioned by the structural inheritance of two and possibly three branches of the triple junction (Watkeys, 2002; Jourdan et al., 2004, 2006), and all geochemical investigations concur that enriched SCLM contributed substantially to the Karoo CFB genesis. Whereas some workers maintain that only the SCLM melted (Duncan et al., 1984; Hawkesworth et al. 1984; Ellam & Cox, 1989; Elburg & Goldberg, 2000), others envisage that the SCLM may have mixed with asthenospherically (plume?) derived magmas (Cox, 1988; Sweeney & Watkeys, 1990; Ellam & Cox, 1991; Sweeney et al., 1991, 1994) with an ocean island basalt (OIB)-like signature (Ellam et al., 1992). Therefore, the various scenarios currently proposed for the origin of the Karoo CFB (passive or active rifting models) are conflicting and incomplete, as they rest on insufficient geochemical data obtained for only a few localized areas, ignoring crucial regions such as the northwestern part of the Karoo CFB and the northern branches of the giant dyke system. Here we attempt to bridge this gap by presenting major and trace element and Sr, Nd, Hf, and Pb isotopic data for the extensive lava flows and sills from Botswana and the dykes from the two northern giant dyke swarms; that is, the Okavango and the Save–Limpopo dyke swarms (Fig. 1). This study is the first comprehensive investigation of the northwestern Karoo. In addition to providing an extended database for this area, our work also aims at (1) constraining the origin of the low- and high-Ti magmatism characterizing the Karoo CFB province and (2) tentatively outlining two possible scenarios for the origin of the Karoo CFB.
|
OVERVIEW OF THE KAROO MAGMATIC PROVINCE |
|---|
The Karoo province in southern Africa
The Karoo magmatism in southern Africa consists of a vast cover of lava flows and sills, giant dyke swarms, and more localized intrusive centers (Fig. 1), intruding the Archaean Kaapvaal and Zimbabwe cratons, the Paleo-Proterozoic Limpopo belt, and the Permian–Jurassic Karoo sedimentary cover (Eales et al., 1984
; Carney et al., 1994; Key & Ayres, 2000).
The main dyke systems are the N110° Okavango dyke swarm, the N70° Save–Limpopo dyke swarm, and the north–souths north Lebombo and Rooi Rand dyke swarms, all of them dominantly basaltic. The Okavango and Save–Limpopo dyke swarms, mainly exposed in Botswana (and partly in Zimbabwe), have been dated recently by the 40Ar/39Ar method at 179 Ma (n = 18; Le Gall et al., 2002; Jourdan et al., 2004, 2005) and are thus inferred to represent a brief tectono-magmatic event. The Olifants River dyke swarm, previously proposed to be part of the Karoo province (e.g. Ernst & Buchan, 2001), can no longer be considered as a Karoo swarm because of its Proterozoic–Archaean age (Marsh, 2002; Jourdan et al., 2006).
Karoo lava remnants are present over most of southern Africa and include he following.
- The Lesotho basaltic lava pile (Marsh et al., 1997), which has yielded five plagioclase 40Ar/39Ar ages ranging between 182·3 ± 1·8 and 181·2 ± 1·6 Ma (2
; Jourdan et al., in press) and one (outlier?) plagioclase age at 183·9 ± 1·4 Ma (Duncan et al., 1997).
- The 10 km thick Lebombo lava pile, forming a 700 km long monoclinal rifted margin at the eastern edge of the Kaapvaal craton (Watkeys, 2002) and comprising a succession of nephelinitic, picritic, basaltic, and overlying rhyolitic units (e.g. Sweeney et al., 1994, and references therein). The basalts have yielded one 40Ar/39Ar age of 184·2 ± 1·2 Ma on plagioclase separates (Duncan et al., 1997), whereas two north–south dykes cross-cutting the basaltic sequence have provided 40Ar/39Ar ages of 181· 4 ± 0·7 and 182·3 ± 1·7 Ma (Jourdan et al., 2005). The rhyolites gave a sensitive high-resolution ion microprobe (SHRIMP) U–Pb age of 180·2 ± 2·7 (Riley et al., 2004) and a plagioclase 40Ar/39Ar age of 177·8 ± 0·7 Ma.
- Picrites and basalts capping the Karoo sedimentary sequence in the Mwenezi (= Nuanetsi) basin (Ellam & Cox, 1989, 1991; Ellam et al., 1992).
- Extensive outcrops and subcrops of basaltic lava flows in Botswana (Wigley, 1995) and adjacent Zimbabwe–Zambia (Jones et al., 2001). These formations have been dated recently to be between 177 and 182 Ma by 40Ar/39Ar on plagioclase separates (n = 24; Jones et al., 2001; Jourdan et al., 2005).
- Basaltic lava flows in central Namibia at 183·0 ± 1·2 Ma (Duncan et al., 1997).
Vast doleritic sills are particularly striking in South Africa and constitute the main outcrops in the Karoo sedimentary basin. In spite of their widespread distribution, they have so far not been dated, except for one zircon U–Pb age of 183·7 ± 0·6 Ma obtained on a granophyric differentiate (Encarnacion et al., 1996). Sills are also documented in southern and eastern Namibia and (as subcrops) in western Botswana, and have yielded 40Ar/39Ar ages on plagioclase separates ranging from 180·5 ± 1· 4 to 184·7 ± 1· 0 Ma (n = 3; Duncan et al., 1997) and 181· 0 ± 0·7 to 181· 8 ± 1· 6 Ma (n = 2; Jourdan et al., 2005), respectively. Two dykes from the minor Underberg dyke swarm (Fig. 1), near the Lesotho basalts, have yielded two younger ages at 176·4 ± 1·2 Ma and 176·1 ± 1·2 Ma (Riley et al., 2006). Several gabbroic, syenitic, and granitic intrusive complexes were emplaced in the Mwenezi trough and have yielded 40Ar/39Ar ages ranging from 178·2 ± 1·7 to 174·4 ± 0·7 Ma (n = 5; Jourdan et al., 2007). The north–south-striking Rooi Rand dykes emplaced along the Lebombo monocline show younger ages ranging between 172·1 ± 2·3 and 173·9 ± 0·7 Ma (n = 3; Jourdan et al., 2007).
Geochemical studies conducted so far have focused mainly on the lava flows and a few dykes from the eastern and southern parts of the Karoo province, and provide constraints on the origin and genesis of these magmas (see references hereafter). The rocks can be subdivided into five groups.
- The Mashikiri nephelinites (MgO 2·6–12 wt %) are located in the Mwenezi district and display significant incompatible trace element (ITE) enrichment and extremely low Ndi (–9·8 to –20·9) for fairly radiogenic Sr (Harmer et al., 1998). They have been interpreted as reflecting ancient metasomatically enriched SCLM (Hawkesworth et al., 1984; Ellam & Cox, 1991; Harmer et al., 1998).
- The picrites (MgO 10–24 wt %) of the Letaba Formation are mostly restricted to the Mwenezi area, but are of greater volume and extent than the nephelinites. An origin from ancient enriched SCLM akin, but not identical, to the source of the nephelinites was first proposed (e.g. Bristow et al. 1984; Hawkesworth et al., 1984; Ellam & Cox, 1989). Later, the pricrites were reinterpreted in terms of mixing between either ambient asthenospheric mantle and SCLM (Ellam & Cox, 1991; Sweeney et al., 1991) or, based on Os isotopes, SCLM and a mantle plume (Ellam & Cox, 1992). Most recently, based on new Pb and Hf isotope data, Ellam (2006) proposed a derivation from either a heterogeneous lithospheric mantle source or a more complex mixture of source components.
- The tholeiitic basalts and dolerites, which constitute the overwhelming majority of Karoo rocks, have been classified into two subgroups, the low- and high-Ti groups, on the basis of their TiO2, P2O5, and ITE contents, with the limit set at 2·0–2·5 wt % TiO2 [first recognized by Cox et al. (1965)]. The two groups show strong geographic provinciality (Figs 1 and 2). The high-Ti basalts occur in northern Lebombo, in the Mwenezi and Tuli basins (Cox et al., 1967; Duncan et al., 1984), and in northern Botswana along the Okavango dyke swarm (Elburg & Goldberg, 2000; Jourdan et al., 2004), as well as in the uppermost lava flows capping the low-Ti lava pile (Wigley, 1995). The low-Ti group is located in southern Lebombo (Sweeney et al., 1994), in Lesotho and southern South Africa (Duncan et al., 1984; Marsh et al., 1997; Riley et al., 2005, 2006), in Namibia (Duncan et al., 1984), and in Botswana, except for the Okavango dyke swarm (Wigley, 1995). Beyond the effects of crustal contamination, the origins of the high- and low-Ti sub-provinces have yet to be agreed upon, as they have so far been ascribed to either heterogeneous SCLM (Duncan et al., 1984; Hawkesworth et al., 1984; Sweeney & Watkeys, 1990; Elburg & Goldberg, 2000) or mixing between asthenospheric mantle and SCLM (Sweeney et al., 1994).
- The rhyolites capping the basaltic lava pile in Lebombo and Mwenezi encompass a wide range of isotopic values from mantle-like to extreme crustal signatures (Betton, 1979; Harris & Erlank, 1992).
- The mid-ocean ridge basalt (MORB)-like Rooi Rand dyke swarm (Fig. 1) emplaced along the southern part of the Lebombo monocline is interpreted as the final stage of Karoo magmatism just prior to the onset of ocean-floor spreading (Hawkesworth et al., 1984; Duncan et al., 1990; Watkeys, 2002).
|
The Karoo province in Antarctica
In Antarctica, the Karoo magmatic rocks consist of dykes and lava flows cropping out in western Dronning Maud Land (DM), from Vestfjellan to H.U. Svedrupfjella. Detailed descriptions have been given by Luttinen & Furnes (2000
) and Riley et al. (2005) and thus will be only briefly summarized here. Karoo rocks in Antarctica cover or intrude the Archaean Grunehogna craton, the Mesoproterozoic Maud Belt, and Neoproterozoic rocks of the Ritscherlflya supergroup.
The geochronology of the Karoo province in Antarctica consists of seven published 40Ar/39Ar mineral ages ranging from 172 to 183 Ma (Brewer et al., 1996; Duncan et al., 1997; Zhang et al., 2003). We note that these ages have been obtained using different, not necessarily intercalibrated, standards, rendering their comparison not straightforward.
The rocks have been classified into several groups based on their geographical locations and chemical compositions, with the CT (chemical type) groups being from the Vestfjella area (Luttinen & Furnes, 2000) and the DM groups from various areas of Dronning Maud Land (Riley et al., 2005). The principal groups are for the most part (with the exception of DM3) equivalent to existing rocks in southern Africa and are summarized below.
- The CT1 and DM1 groups have low MgO (<9 wt %) and TiO2 (<2·2 wt %) contents and moderately enriched Sr isotopic compositions (87Sr/86Sr 0·705–0.710) and low Ndi values (–1 to –15). These groups have been interpreted as either the products of mixing between a depleted end-member (DM3) and 12% partial melts of the SCLM or a depleted end-member contaminated by the upper crust (Riley et al., 2005). They correspond to the low-Ti basalts of southern Africa.
- The CT2 and DM2 groups have variable TiO2 contents ranging from 2 to 4 wt% and trace element abundances and Sr and Nd isotopic compositions similar to the MORB-like Rooi Rand dyke swarm from southern Africa.
- The DM3 group is the only group not represented in southern Africa. It consists of basalts, picrites, and ferro-picrites with humpback-shaped rare earth element (REE) and normal MORB (N-MORB)-like incompatible trace element (ITE) patterns. Their isotopic compositions are characterized by remarkably high Ndi (+5 to +9) and unradiogenic 87S/86Sr (0·7035–0·7062) consistent with derivation from a depleted mantle (plume?) source (Riley et al., 2005).
- Group 4 includes basaltic and picritic rocks that have high TiO2 contents (3·8–5·2 wt %) and enriched REE and ITE patterns. These rocks show large variations in Nd and moderate variations in Sr isotopic compositions, and are therefore roughly similar to high-Ti rocks from southern Africa. They have been interpreted as mixing between a MORB source and variable degrees (20–30%) of SCLM melts (Riley et al., 2005).
|
GEOLOGICAL SETTING AND SAMPLING STRATEGY |
|---|
 TOPABSTRACTINTRODUCTIONOVERVIEW OF THE KAROO... GEOLOGICAL SETTING AND SAMPLING... PETROGRAPHY OF THE SAMPLESANALYTICAL PROCEDURESGEOCHEMISTRYDISCUSSIONCONCLUSIONSSUPPLEMENTARY DATAREFERENCES |
|---|
We focused our sampling efforts on the northern Karoo province, and in particular on the Okavango and Save–Limpopo dyke swarms, as well as on the extensive lava flows and sills of Botswana and southernmost Zimbabwe.
The N110°-trending Okavango dyke swarm extends from northern Namibia to southern Zimbabwe through Botswana (Fig. 2a). Whereas it is hidden by Kalahari sands to the west, it is well exposed to the east, where it intrudes the Zimbabwe Archaean craton, the metamorphic Limpopo–Shashe belt, and the Karoo Permian–Jurassic sedimentary sequence deposited in the Tuli half-graben. The present sample suite, from 77 dykes, was collected primarily along the Shashe river (65 dykes), representing a 100 km long cross-section of the dyke swarm (Le Gall et al., 2002; Jourdan et al., 2004). Additional sampling was carried out in the Tuli half-graben, both in eastern Botswana (Thune river: 10 dykes) and in southern Zimbabwe (one dyke). A further dyke (Bo52), the orientation of which is poorly constrained, was sampled in western Botswana, close to the western extremity of the swarm. The mean width of individual subvertical dykes differs as a function of whether the dykes intrude basement or sedimentary rocks (on average c. 18 m and 1·3 m wide, respectively; Le Gall et al., 2005).
The N70°-trending Save–Limpopo dyke swarm is mainly (though poorly) exposed in the SW part of the Tuli basin and surrounding region. It consists of a c. 50 km wide swarm including vertical to subvertical dykes with a mean thickness of 27 m (Le Gall et al., 2005). These dykes intrude the metamorphic basement of the Limpopo belt and also crop out in Zimbabwe along and within the Save–Limpopo monocline (Chavez Gomez, 2001). A few dykes with poorly constrained orientations were sampled in northern and eastern Botswana (Bo5, Bo6, Bo4, Fig. 2a).
Lava flows and sills cover most of Botswana and the bordering part of Zimbabwe. Only limited remnants, fortuitously preserved from erosion, are exposed (mainly to the east of the 25°E meridian). Large volumes of basalt are covered by the Kalahari sands over the entire country and are documented by several boreholes (Fig. 2a). The main lava flows are exposed in the Tuli–Mwenezi–North Lebombo area, which is filled in by a 1 km thick volcanic sequence overlying the Karoo sedimentary deposits (Eales et al., 1984; Key & Ayres, 2000). From this sequence, we sampled picritic basalts (nine, four, and two samples from Tuli, Mwenezi, and Lebombo, respectively), the distribution of which is restricted to the lower part of the lava flow. We further collected seven overlying basalts from the Tuli area (Fig. 2a), as well as basalts from several scattered lava flow outcrops in northern Botswana (Bo10, Bo11 and Bo16) and Serowe (Bo1, Bo2, Bo3), where basalts are associated with doleritic plugs (Bo31, Bo32, Bo33) (Fig. 2a).
The sill subcrops, 16–170 m thick, are located in southwestern Botswana and were sampled from boreholes Ckp8C1, Ckp6 etc. In addition, samples of widespread lava flow subcrops were collected from eight boreholes that recovered a 90 to >835 m thick lava sequence. A 21 lava flow sample suite was collected systematically from a northern borehole (P11 lava flow), whereas only the least altered samples from the other drill cores were selected.
|
PETROGRAPHY OF THE SAMPLES |
|---|
The northwestern Karoo doleritic dykes and sills studied here consist of fine- to medium-grained rocks with variable amounts of phenocrysts composed of plagioclase (the most abundant phase), augitic clinopyroxene, occasionally associated with pigeonite, and opaque minerals (titanomagnetite and in some cases ilmenite). Minor olivine phenocrysts occur in some samples. The groundmass comprises a variable mixture of glass and cryptocrystalline plagioclase, pyroxene, Ti-magnetite, and olivine. The N110° and N70° dykes are petrographically indistinguishable. The basaltic lava flows have a similar modal mineralogy to the dykes, but their textures are more fine-grained with larger amounts of glass and cryptocrystalline mesostasis. Picritic and basaltic picritic dykes and lava flows consist of predominant euhedral olivine phenocrysts (fresh or partially iddingsitized) occurring in a groundmass made of olivine ± clinopyroxene ± plagioclase ± oxides ± interstitial glass.
Alteration phases are variably developed, depending on the individual sample and are mainly represented by sericite (after plagioclase) and bowlingite or iddingsite (after olivine) along with voids filled by zeolite in the upper part of some of the lava flows.
|
ANALYTICAL PROCEDURES |
|---|
A suite of 147 samples were selected for major and trace element analysis. They were crushed and powdered in an agate mill and analyzed by X-ray fluorescence (XRF) using a Philips PW 1404 spectrometer at the University of Lyon. Analytical uncertainties are 1–2% and 10–15%, respectively, for major and trace elements. A total of 133 rocks were also analysed for REE and additional trace elements by inductively coupled plasma mass spectrometry (ICP-MS) at the Chemex Laboratories (Canada), with errors varying from 5 to 10% depending on the concentration of any given element.
Isotope analyses were carried out on 34 rocks and two plagioclase separates selected to be representative of the whole sample suite. Lead and Sr isotope analyses were conducted at the University of Nice. Samples were broken into small chips and carefully separated (in a clean laboratory) to avoid saw marks. Chips were leached with HF and dissolved successively in HF and HCl. Purification and elution of Pb were performed following a modified HBr procedure (Manhès et al. 1978; Schärer, 1991). Lead isotopic compositions were measured on an EMI electron multiplier system of a Thomson-206 mass spectrometer, using single zone-refined Re filaments onto which Pb was loaded with H3PO4 and silica-gel. Mass fractionation was 0·1 ± 0·05%/a.m.u. for ionization temperatures between 1300 and 1400°C, controlled by the NBS-981 Pb standard, which yielded average ratios of 16·941 ± 0·004 (2) for 206Pb/204Pb, 15·501 ± 0·004 for 207Pb/204Pb, and 36·728 ± 0·009 for 208Pb/204Pb (eight runs). Total procedural Pb blanks were 60–80 pg. Strontium was separated from Rb and purified on Sr-spec resin and deposited on Re filaments with TaF5. The Sr isotope measurements were carried out in dynamic double-collector mode using 86Sr/88Sr = 0·1194 for linear mass fractionation correction. Strontium measurement accuracy was controlled using the NBS-987 standard, which yielded a mean value of 0·71026 ± 0·00005 (2-STERR) for 87Sr/86Sr. Total procedural Sr blanks ranged between 60 and 100 pg. Both Pb and Sr blanks were negligible compared with the amounts of these elements analyzed.
Hafnium and Nd isotope compositions were determined by multi-collector (MC) ICP-MS using the VG Plasma 54 at the Ecole Normale Supérieure in Lyon (ENSL). Hafnium and Nd chemical separation was carried out at ENSL following the procedures described by Blichert-Toft et al. (1997, 2005) and Blichert-Toft (2001). The JMC-475 and La Jolla standards were measured systematically after every two sample analyses to monitor machine performance and gave 176Hf/177Hf = 0·282160 ± 10 (2) and 143Nd/144Nd = 0·511858 ± 18 (2), respectively. Instrumental mass fractionation was corrected relative to 179Hf/177Hf = 0·7325 and 146Nd/144Nd = 0·7219 using an exponential law. Hafnium and Nd total procedural blanks were better than 25 pg. Age corrections of 179 Ma were applied to all four isotope systems using their respective parent isotope decay constants, which in the controversial case of 176Lu was chosen to be the so-called ‘terrestrial’ decay constant of 1·865 x 10–11 per year (Scherer et al., 2001).
|
GEOCHEMISTRY |
|---|
Major elements
The Karoo lavas and dykes from Botswana and western Zimbabwe include a large range of rock types (from picrites to basaltic andesites) and compositions; for example, the Mg-number [atomic ratio of 100Mg/(Mg + 0·85 Fetot)] ranges from 82 to 31. Following earlier investigators (e.g. Cox et al., 1967
; Cox, 1988; Sweeney et al., 1994), the rocks can be subdivided into two major groups on the basis of their TiO2 and P2O5 contents (Table 1 and Electronic Appendix 1, available for downloading at http://petrology.oxfordjournals.org): the high-Ti group (TiO2 > 2 wt %) includes basaltic dykes from the Okavango (78 samples) and Save–Limpopo (four samples) swarms, basaltic lava flows and sills (18 samples), and picrites (13 samples), whereas the low-Ti group (TiO2 < 2 wt %) comprises only basaltic lava flows (32 samples) and two picrites.
|
The high-Ti dykes, lava flows, and sills share similar major element compositions characterized in particular by high TiO2 (2·16–4·42 wt %) and P2O5 (0·23–1·00 wt %) contents. They are quartz- to olivine-normative tholeiites and their SiO2 contents vary from 47·18 to 52·40 wt %, with the exception of one slightly more differentiated 57·08 wt % SiO2). On the total alkali–silica diagram (TAS; Le Bas et al., 1986
; Fig. 3), the dyke rocks define a narrow range of compositions, which lie in the field of basalts straddling the boundary with basaltic andesites (except for Bot0049, which is classified as a trachyandesite). In contrast, the high-Ti lava flows display a slightly wider range of compositions extending up to the field of basaltic trachyandesites. MgO and Mg-number vary from 2·6 to 7·2 wt % and from 31 to 57, respectively, indicating that these samples are moderately to highly evolved. Mg-number exhibits a negative covariation with TiO2, P2O5, and FeOt (FeOt not shown) and a positive covariation with CaO and Al2O3 (Fig. 4a), suggesting that these magmas have undergone low-pressure differentiation involving largely plagioclase and pyroxene.
|
|
The low-Ti basaltic lava flows (n = 32) are classified as basalts with a few basaltic andesites, just as for the high-Ti dykes. Overall, the low-Ti basalts tend to be shifted towards higher Mg-number compared with the high-Ti basalts (Fig. 4a), indicating less evolved compositions. The two groups differ primarily in terms of their TiO2 and P2O5 contents, and, for the low-Ti group, by the more limited range of these two elements for a given Mg-number.
The last group consists of mafic dykes and lava flows from the Letaba Formation, corresponding to the so-called picrites (Eales et al., 1984; Ellam & Cox; 1989), which are restricted to the Tuli–Mwenezi–North Lebombo area. Although these rocks classify as basalts on the TAS diagram (Fig. 3), they will hereafter be referred to as ‘picrites’ according to previous studies and in agreement with their high MgO (12·7–17·6 wt %) and olivine contents. As for the basaltic lava flows, the picrites can also be divided into two subgroups defined on the basis of their TiO2 contents, which range from 2·5 to 3·5 wt % for the widely dominant high-Ti group (nine samples from Tuli, two samples from Mwenezi, and two samples from Lebombo) and from 1·7 to 1·8 wt % for the less abundant low-Ti group (two samples deriving from rare intrusions within the Mwenezi high-Ti basalts). These rocks are characterized by high Mg-number (ranging from 71 to 77 and 79 to 82 for the high- and low-Ti groups, respectively). In the CaO and Al2O3 vs Mg-number plots (Fig. 4a), the high- and low-Ti picrites define a negative covariation, indicative of either olivine fractionation or accumulation in the absence of clinopyroxene (or plagioclase) fractionation (as illustrated by the crystallization vectors in Fig. 4a; e.g. Albarède et al., 1997; Peate et al., 2003).
Trace elements
The high- and low-Ti groups also display contrasting trace element behaviours. The high-Ti basaltic lavas, sills, and dykes show large variations defining the following trends as a function of decreasing Mg-number (Fig 4b): (1) decreasing concentrations of the compatible elements Ni and Cr, whereas V and Sc are more scattered (not shown); (2) increasing concentrations of the incompatible elements Th, La, Nb, Zr, Y, Ba, Rb, and Sr with more scatter among the large ion lithophile elements (LILE) such as Ba and Rb. The few N70° dykes analyzed lie within the compositional range of the N110° dykes. For very incompatible trace elements [VICE; e.g. light REE (LREE), Th, Nb, Zr], the high-Ti lava flows are subdivided into two groups forming two distinct trends (except for Yb and the more compatible trace elements) generally encompassing the range of the high-Ti dykes. These two groups will hereafter be referred to as the low incompatible trace element (low-ITE) and the high incompatible trace element (high-ITE) groups (Fig. 4b).
Chondrite-normalized REE patterns of the high-Ti dykes are variably LREE-enriched [Lan ranges from 48 to 211 and (La/Yb)n from 3·3 to 12·3], and show slight negative Eu anomalies [Eu/Eu* = 0·80–0·93 with Eu* = 
(Sm x Gd)], except for the more differentiated sample Bot0049, which has Lan = 291 and Eu/Eu* = 0·72 (Fig. 5a). We cautiously note, however, that the small Eu anomalies could be due to overcorrection of the isobaric interference of BaO on Eu during ICP-MS analysis.
|
The high-Ti rocks are characterized by a significant fractionation of heavy REE (HREE) relative to middle REE (MREE), with the Sm/Ybn ratio varying from 2·1 to 5·1. The primitive mantle-normalized trace element patterns also show variable incompatible element enrichment and negative Nb and Sr anomalies, except for sample Bo24, which exhibits a strong positive Sr anomaly as a result of plagioclase glomerocryst accumulation (Fig. 5b). Among the high-Ti lava flows, the low-ITE and high-ITE groups differ in terms of their REE patterns (Fig. 5a) and their normalized trace element patterns (Fig. 5b). The low-ITE group fits the less enriched patterns of the high-Ti dykes, whereas the high-ITE group displays more enriched patterns and stronger HREE fractionation [(Sm/Yb)n = 4·5–5·2]. Both lava groups have larger negative Nb anomalies [Nb/Nb* = 0·43–0·57, with Nb* =
(K x La)] than the high-Ti dykes (Nb/Nb* = 0·48–1·54; Figs 4b and 5b). Compared with the high-Ti group, the low-Ti basaltic lava flows and sills have much less variable ITE abundances. They display a similar decrease and increase, respectively, in the abundances of compatible trace elements (Ni, Cr), HREE and Y, but a considerably smaller increase in VICE (e.g. Th, La, Nb, Zr), with decreasing Mg-number (Fig. 4b). Ratios such as Zr/Y are therefore much lower for the low-Ti (2·7–4·2) than for the high-Ti (5·7–10·7) group. Rubidium and, to a lesser extent, Ba show more scattered concentrations, whereas Sr contents decrease slightly with decreasing Mg-number, contrary to the high-Ti group. The REE patterns and normalized trace element patterns of the low-Ti group are more homogeneous and less enriched in LREE and other VICE [e.g. (La/Yb)n = 2·0–3·4 compared with 4·8–13· 1 for the high-Ti group]. The low-Ti group also differs from the high-Ti group by having flatter MREE–HREE patterns [(Sm/Yb)n = 1· 2–1· 6]. The normalized trace element patterns display marked negative Nb anomalies (Nb/Nb* = 0·22–0·81) broadly similar to those observed for the high-Ti lava flows, small negative Ti and Eu anomalies (Ti/Ti* = 0·66–0·93 and Eu/Eu* = 0·82–1· 02), and variable Sr anomalies ranging from negative to positive (Sr/Sr* = 0·48–1· 38) depending on plagioclase removal or accumulation (Fig. 5b).
All the picrites have high contents of compatible trace elements, which decrease with decreasing Mg-number; nevertheless, the low-Ti picrites have the highest concentrations (e.g. Ni, Fig. 4b). VICE contents (except for Nb) for low- and high-Ti picrites are about the same as for low- and high-Ti basalts with larger variations observed within the high-Ti groups, but no clear increase with decreasing Mg-number (Fig. 4b). The range for less incompatible elements such as Y (Fig. 4b) and HREE (not shown) is more restricted and lower for the high- and low-Ti picrites (22–28 and 14–15 ppm Y) than for the high- and low-Ti basalts (22–52 and 19–41 ppm Y). As a result, the picrites display REE and multi-trace element patterns that are more fractionated than those of the basalts with higher MREE/HREE ratios [(Sm/Yb)n = 6·0–8·1 and 4·1–4·4 for high- and low-Ti picrites, respectively]. The picrites also show negative Nb anomalies, with Nb/Nb* = 0·22–0·55 and 0·43–0·99 for the high- and low-Ti picrites, respectively. The negative Nb anomalies of the high-Ti picrites are among the largest observed in the present dataset.
Sr, Nd, Hf, and Pb isotopes
Ten high-Ti dykes from the Okavango dyke swarm, eight high-Ti lava flows, seven high-Ti picrites, and eight low-Ti lava flows were analyzed for their Sr, Nd, Hf, and Pb isotope compositions (Table 2 and Fig. 2). Additionally, one low-Ti picrite also was analyzed for its Nd and Hf isotope composition. Isotopic ratios were back-calculated to their initial values at 179 Ma in agreement with the 40Ar/39Ar ages determined by Le Gall et al. (2002) and Jourdan et al. (2004, 2005); these are the values reported in the various isotope plots (Fig. 6a–f). The samples show relatively large variations in 87Sr/86Sri (0·7040–0·7064), Ndi (–9·4 to + 1·1), Hfi (–8·1 to +5·2), 206Pb/204Pbi (16·62–18·02), 207Pb/204Pbi (15·14–15·61), and 208Pb/204Pbi (36·86–38·00) (Table 2), thus readily allowing discrimination of the four groups previously defined (Table 1 and Electronic Appendix 1).
|
|
The high-Ti dykes from the Okavango dyke swarm show a large range of Sr isotopic variation (0·7040–0·7059) for a restricted range of
Ndi (+1·1 to –2·2) and Hfi (+1·3 to +5·2). On the Hfi vs Ndi plot (Fig. 6c) they lie along the mantle array (slope = 1· 4, y-axis intercept = +3, J. Blichert-Toft, published and unpublished data compilation of 2500 OIB and MORB of global distribution). In Pb–Pb isotope space (Fig. 6e and f), the dykes display relatively unradiogenic 206Pb/204Pbi (17·15–17·62), unradiogenic to radiogenic 207Pb/204Pbi (15·33–15·58), and moderately radiogenic 208Pb/204Pbi (37·40–37·95) (for a given 206Pb/204Pbi value). In the 207Pb/204Pbi vs 206Pb/204Pbi diagram (Fig. 6e), the high-Ti dykes plot to the left of the 179 Ma Geochron, extending from the Northern Hemisphere Reference Line (NHRL; Hart, 1984) toward compositions more radiogenic in 207Pb and defining a steep positive slope parallel to the Geochron. 208Pb/204Pbi is ‘decoupled’ from 206Pb/204Pbi resulting in a clustering of the samples above the NHRL, between the Depleted MORB Mantle (DMM) and Enriched Mantle I (EMI) mantle components (Fig. 6f). This indicates moderately high time-integrated Th/Pb and a difference in the Th/U of the source(s) compared with the magma plotting on the NHRL. When 206Pb/204Pbi is plotted against 87Sr/86Sri, the samples fall between Bulk Silicate Earth (BSE) and EMI (Fig. 6d).
The high-Ti lava flows (except for sample Ckp8C1-2, which displays both high- and low-Ti features) show a strong negative covariation between 87Sr/86Sri (0·7043–0·7057) and Ndi (–0·2 to –7·8) (Fig. 6a). They define a marked trend from BSE toward and beyond the EMI-like component, in the direction of the Leucite Hills lamproite field (O’Briens et al., 1995). Among these lava flow samples, the low-ITE subgroup plots close to BSE, whereas the high-ITE subgroup corresponds to the enriched end-member of the trend. Similar relationships are observed in plots of Hfi–Ndi, 87Sr/86Sri–206Pb/204Pbi, and Ndi–206Pb/204Pbi. The high-Ti lavas partially overlap with the high-Ti dykes in Pb–Pb isotope space (Fig. 6e and f), although they show slightly less radiogenic and more variable ratios (206Pb/204Pbi from 16·62 to 17· 41, 207Pb/204Pbi from 15·22 to 15·43, and 208Pb/204Pbi from 37·10 to 37·58).
The low-Ti lava flows show more radiogenic 87Sr/86Sri (0·7053–0·7063) at similar Hfi (–1·7 to +1· 8) and Ndi (–3·6 to –0·7) compared with the high-Ti dykes. In the Ndi vs 87Sr/86Sri diagram (Fig. 6a), they extend the trend of the high-Ti dykes towards more radiogenic Sr isotope ratios, at similar Ndi. A noteworthy feature is the presence of one outlier (Ckp8A1), which has more radiogenic 87Sr/86Sri (0·7073) for a similar Ndi, apparently related to its evolved character (Mg-number = 41). Compared with the high-Ti group, the low-Ti lavas (1) have higher 206Pb/204Pbi (17·60–18·02) for a similar range in 207Pb/204Pbi (15·34–15·56) and 208Pb/204Pbi (36·86–37·68), thus shifting them to the right of the Geochron (Fig. 6e and f), and (2) define opposite ‘trends’ in the Ndi and 87Sr/86Sri vs 206Pb/204Pbi diagrams (Fig. 6b and d).
The high-Ti picrites share some features with the high-ITE subgroup of the high-Ti lava flows, but extend to more enriched isotopic compositions with even lower Hfi (–5·2 to –8·1) and Ndi (–5·0 to –9·4) (Fig. 6a–c), locating them close to the field defined by the Leucite Hills and Gaussberg lamproites (O’Brien et al., 1995; Murphy et al., 2002, J. Blichert-Toft & B. Hanan, unpublished data). In the Hfi vs Ndi plot (Fig. 6c), the picrites (and possibly the high-ITE, high-Ti lava flows) define a shallower trend (slope = 0·6) than observed for the other high-Ti rocks, thus deviating slightly from the mantle (MORB–OIB) array. In Pb–Pb isotope space, the picrites overlap the fields of both high-Ti lava flows and dykes, defining an array parallel to (and to the left of) the Geochron (Fig. 6e). In the Hfi vs Ndi plot, the low-Ti picrite plots slightly below the mantle array, but has a composition similar to the high-Ti picrites.
Pb and Sr isotopic data for the high-Ti samples Bot0098 (Okavango dyke) and P11-2 (lava flow) were obtained for both the whole-rock and plagioclase separates (Table 2). For sample Bot0098, plagioclase, compared with the whole-rock, has slightly more radiogenic 206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb and noticeably less radiogenic 87Sr/86Sr (Fig. 6d–f). For sample P11-2, plagioclase has overall less radiogenic Pb isotope compositions and less radiogenic 87Sr/86Sr. The deviation between the Pb isotope compositions of the whole-rocks and the plagioclase separates is within (or slightly larger than) the error bars (Table 2), whereas the discrepancies observed for 87Sr/86Sr are significantly larger than what can be accounted for by the expected analytical uncertainties and suggest a possible contribution from alteration processes, as will be discussed further below.
Chemical stratigraphy
The chemical stratigraphy of the lava pile in Botswana was assessed from the 370 m thick volcanic succession of the P11 borehole (Fig. 2a), from which 21 basalts were sampled at regular depth intervals. Based on the variations of Mg-number, TiO2, Nb, and La/Yb as a function of depth through the P11 section, three sequences can be identified (Fig. 7). The lower part of the section (from 500 m to 280 m depth in the borehole) consists of low-Ti basalts displaying homogeneous compositions with subtle but regular variations of TiO2 and Nb contents (0·85–1· 02 wt % and 3·54–5·09 ppm, respectively) and trace element ratios [(La/Yb)n = 2·00–2·61]. These compositions are similar to those of other Karoo low-Ti stratigraphic sequences (Cox, 1988; Sweeney et al., 1994; Marsh et al., 1997) and suggest the steady-state outpouring of magmas from a regularly replenished open magma chamber. An abrupt change occurs at the depth of 280 m, which is marked by the emplacement of a primary lava flow (Mg-number = 70) that shifts compositions towards higher Nb concentrations and La/Yb ratios and is in turn followed by a progressive decrease of these variables up to a depth of 160 m, whereas TiO2 contents gradually increase. These small changes probably reflect the re-injection of a fresh pulse of undifferentiated magma that subsequently evolved by fractional crystallization processes. The most striking feature of the section is a dramatic shift within its uppermost 30 m (i.e. between depths of 160 and 130 m) of TiO2 and Nb contents and La/Yb towards the significantly higher values diagnostic of the high-Ti magma group. These changes are broadly similar to those observed in some Lebombo (Sweeney et al., 1994) and Springbok Flats (Marsh et al., 1997) stratigraphic sequences. The switch from low- to high-Ti basalts is also marked by a change in isotopic composition [e.g. Ndi decreases from –2·4 (P11-19) to –7·0 (P11-2)]. This abrupt stratigraphic shift from low- to high-Ti compositions likewise is observed in the P8 (close to the P11 borehole) and the Maun (northwestern Botswana) boreholes (Fig. 2a).
|
The sharp changes observed in major and trace element and isotopic compositions towards the top of the P11 borehole are inconsistent with progressive partial melting and differentiation processes being responsible for the low- to high-Ti transition, and rather suggest the tapping of two distinct magma sources as previously suggested by Cox (1988
), Sweeney & Watkeys (1990), Sweeney et al. (1994) and Elburg & Goldberg (2000). |
DISCUSSION |
|---|
Alteration
In thin section, the effects of alteration on the rocks appear to be minor (especially for the samples selected for isotopic analysis) and, when optically discernible, mostly affect olivine (transformed into bowlingite). This observation is corroborated by the major and trace element compositions of the samples, which do not correlate with loss on ignition (LOI, not shown), suggesting that the rocks are unaltered. However, as mentioned above, the plagioclase and whole-rock isotopic compositions of samples Bot0098 and P11-2 show substantial differences for 87Sr/86Sr (Fig. 8), which suggest, on the contrary, a possible contribution from alteration. Effects of alteration are more likely to affect the Sr isotopic composition of the whole-rock than that of the separated plagioclase because (1) plagioclase was hand-picked under a binocular microscope, carefully selecting only the cleanest minerals, (2) the Bot0098 whole-rock is characterized by a large amount of a glassy oxidized groundmass, which is particularly sensitive to alteration, and (3) the Bot0098 whole-rock sample is shifted towards more radiogenic Sr (at a given
Nd; Fig. 6a) compared with other high-Ti dykes.
|
The isotope ratio vs LOI plots show no visible covariation for Nd, Hf (not shown), or Pb (Fig. 8) isotopic compositions regardless of the group of rocks considered. In contrast, a positive correlation between 87Sr/86Sr and LOI is observed for the high-Ti dykes. This trend shows that (1) the less radiogenic samples have the same isotopic composition as the plagioclase separates from sample Bot0098 and (2) the trend is roughly parallel to the tie-line linking the Bot0098 whole-rock and plagioclase. This suggests that the high-Ti dykes are variably affected by alteration processes, with the samples with more radiogenic Sr being the most strongly affected (Fig. 8). Therefore, the variable 87Sr/86Sr compositions of the high-Ti dykes (Fig. 6) apparently do not reflect mantle source processes, but rather subsequent alteration (along with possible assimilation; see below). We propose that the pristine magmatic composition of the high-Ti dykes is more likely to be located near BSE (as suggested by their
Nd values). Similarly, the Sr isotopic composition also differs between the P11-2 whole-rock and fresh pl