Journal of Petrology

This Article Abstract FREE Full Text (PDF) Supplementary data Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (17) Request Permissions Google Scholar Articles by EWART, A. Articles by ARMSTRONG, R. A. Search for Related Content GeoRef GeoRef Citation Social Bookmarking          What's this?

Journal of Petrology | Volume 45 | Number 1 | Pages 59-105 | 2004
© Oxford University Press 2004; all rights reserved

Petrology and Geochemistry of Early Cretaceous Bimodal Continental Flood Volcanism of the NW Etendeka, Namibia. Part 1: Introduction, Mafic Lavas and Re-evaluation of Mantle Source Components

A. EWART^1,*, J. S. MARSH², S. C. MILNER³, A. R. DUNCAN⁴, B. S. KAMBER¹ and R. A. ARMSTRONG⁵

¹ ADVANCED CENTER FOR QUEENSLAND UNIVERSITY ISOTOPE RESEARCH EXCELLENCE (ACQUIRE), THE UNIVERSITY OF QUEENSLAND, ST. LUCIA, QLD. 4072, AUSTRALIA
² DEPARTMENT OF GEOLOGY, RHODES UNIVERSITY, GRAHAMSTOWN 6140, SOUTH AFRICA
³ PANALYTICAL, LELYWEG 1, 7602EA ALMELO, THE NETHERLANDS
⁴ DEPARTMENT OF GEOLOGICAL SCIENCES, UNIVERSITY OF CAPE TOWN, RONDEBOSCH 7700, SOUTH AFRICA
⁵ RESEARCH SCHOOL OF EARTH SCIENCES, THE AUSTRALIAN NATIONAL UNIVERSITY, CANBERRA, A.C.T. 0200, AUSTRALIA

^* Corresponding author. E-mail: ewart{at}cust.caloundra.net

RECEIVED DECEMBER 3, 2001; ACCEPTED JULY 1, 2003

	ABSTRACT

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
The bimodal NW Etendeka province is located at the continental end of the Tristan plume trace in coastal Namibia. It comprises a high-Ti (Khumib type) and three low-Ti basalt (Tafelberg, Kuidas and Esmeralda types) suites, with, at stratigraphically higher level, interstratified high-Ti latites (three units) and quartz latites (five units), and one low-Ti quartz latite. Khumib basalts are enriched in high field strength elements and light rare earth elements relative to low-Ti types and exhibit trace element affinities with Tristan da Cunha lavas. The unradiogenic ²⁰⁶Pb/²⁰⁴Pb ratios of Khumib basalts are distinctive, most plotting to the left of the 132 Ma Geochron, together with elevated ²⁰⁷Pb/²⁰⁴Pb ratios, and Sr–Nd isotopic compositions plotting in the lower ¹⁴³Nd/¹⁴⁴Nd part of mantle array (EM1-like). The low-Ti basalts have less coherent trace element patterns and variable, radiogenic initial Sr (0·707–0·717) and Pb isotope compositions, implying crustal contamination. Four samples, however, have less radiogenic Pb and Sr that we suggest approximate their uncontaminated source. All basalt types, but particularly the low-Ti types, contain samples with trace element characteristics (e.g. Nb/Nb^*) suggesting metasediment input, considered source-related. Radiogenic isotope compositions of these samples require long-term isolation of the source in the mantle and depletions (relative to unmodified sediment) in certain elements (e.g. Cs, Pb, U), which are possibly subduction-related. A geodynamic model is proposed in which the emerging Tristan plume entrained subducted material in the Transition Zone region, and further entrained asthenosphere during plume head expansion. Mixing calculations suggest that the main features of the Etendeka basalt types can be explained without sub-continental lithospheric mantle input. Crustal contamination is evident in most low-Ti basalts, but is distinct from the incorporation of a metasedimentary source component at mantle depths.

KEY WORDS: Etendeka flood basalts; LOMU–EM1 Pb isotope signatures; recycled crust; three-component magma mixing; Tristan plume

	INTRODUCTION

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
Large igneous provinces, such as the Early Cretaceous Paraná–Etendeka continental flood basalt province, are characterized by the eruption of enormous magma volumes over relatively short time intervals, emplaced as regional lava sequences and dyke swarms, sills and intrusive complexes. Magma generation on this scale has long been linked to upwelling of deep, hot mantle plumes (e.g. Morgan, 1981). Petrologists have accordingly discussed the geochemical characteristics of these provinces in terms of components from the plume, sub-continental lithospheric mantle (SCLM) and the depleted asthenospheric mantle.

Understanding of mantle upwellings has improved greatly during the past two decades (e.g. Griffiths & Campbell, 1990). Stacey & Loper (1983) initially suggested that upwellings originated at the core–mantle boundary, as a result of core cooling; which has since been supported, for large plumes, by tomographic imaging. For example, Ji & Nataf (1998) documented transport of hot material in narrow cylindrical corridors from the lowermost mantle, and for the South Atlantic, a deep mantle upwelling has been linked to the so-called South African ‘superswell’ (Romanowicz & Gung, 2002). The recognition of lowermost mantle-derived plumes and possible melt layers at the bottom of the lower mantle (Williams & Garnero, 1996) is relevant to the interpretation of the isotope and geochemical systematics of plume-derived magmas (Brandon et al., 1998). Equally important petrologically is the realization that subducted oceanic plates appear to penetrate the upper–lower mantle boundary at 670 km (van der Hilst et al., 1991), or more typically, to accumulate in the mantle Transition Zone between 410 and 670 km depth (Ringwood, 1989; Simons et al., 1999; Fukao et al., 2001).

Mantle upwellings originating from the core–mantle boundary may incorporate recycled metamorphosed, subducted oceanic crust and upper mantle during ascent, especially during the initial rise of the plume head. It is therefore appropriate to consider stored eclogitic oceanic crust and associated metasediment as an additional magma source component within large basalt provinces. Because the Transition Zone is laterally heterogeneous, and may contain zones of subducted slabs of varying age, the possibility exists for the observed provinciality within large igneous provinces to be related to lateral heterogeneities within the Transition Zone rather than within the sub-continentental lithospheric mantle (SCLM) (e.g. Cox, 1988).

In this paper we argue that enriched SCLM was not a major magma source contributing to Paraná–Etendeka volcanism, but that the Tristan plume was the main trigger for the volcanism and contributed, with local asthenospheric mantle, to the petrogenesis of most of the primitive mafic magmas. We also argue that contributions from subducted slabs in the Transition Zone can be identified from their distinctive geochemical and isotopic fingerprints.

Paraná–Etendeka magmatism is generally considered closely associated, in space and time, with continental rifting above the long-lived Tristan da Cunha mantle plume (e.g. O'Connor & Duncan, 1990; Peate et al., 1990; Hawkesworth et al., 1992). Parallel studies in South America and Africa over the past two decades have provided a wealth of data used to support this interpretation (e.g. Bellieni et al., 1984, 1986; Erlank et al., 1984; Piccirillo et al., 1988a, 1988b; Peate et al., 1990, 1999; Peate & Hawkesworth, 1996; Ewart et al., 1998a; Marques et al., 1999; Schmitt et al., 2000; Marsh et al., 2001, and numerous references therein). The Paraná volcanic sequence extends over an area of 1·2 x 10⁶ km² with a preserved thickness of 1·7 km in the north (Melfi et al., 1988; Peate et al., 1990; Peate, 1997). In contrast, eroded remnants of the Etendeka sequence (Fig. 1), with maximum observed thickness of 1 km, crop out over an area of 78 000 km². In the Paraná, Bellieni et al. (1984, 1986) estimated that silicic volcanics make up 3% of the total eruptive volume. In the Etendeka, silicic volcanics commonly constitute up to 50% of the preserved volcanic thickness. This difference in the proportions of silicic rock types is a function of the asymmetry of the Paraná–Etendeka Province relative to the Atlantic rift (Turner et al., 1994), and the fact that the silicic volcanics are related spatially to the rift.

View larger version (50K): [in this window] [in a new window]   Fig. 1. Map of the Etendeka igneous province in northwestern Namibia, showing the main volcanic and intrusive outcrop areas, and the locations of Figs 2 and 3.

 

View larger version (52K): [in this window] [in a new window]   Fig. 2. Geological map (and representative section) of the northern coastal subdomain between the Hoarusib River and Angra Fria. Map compiled by A. Ewart from mapping by A. R. Duncan, A. Ewart, J. S. Marsh, R. McG. Miller (Agate Mountain carbonatite) and S. C. Milner. Numbers in legend indicate progressive upward younging in relative stratigraphy.

 

View larger version (26K): [in this window] [in a new window]   Fig. 3. Geological map of the southern coastal subdomain between the Hoanib River and Terrace Bay. Map compiled by A. Ewart from mapping by A. R. Duncan and A. Ewart.

 
Early Cretaceous Gondwana plate reconstructions (De Wit et al., 1988) place the Etendeka region adjacent to the southeastern corner of the Paraná Basin. The close similarities between the magmatism of these once contiguous areas has long been known and is reinforced by modern studies, particularly age dating (e.g. Renne et al., 1992, 1996; Stewart et al., 1996) and the trans-Atlantic correlations of specific silicic eruptive units (Milner et al., 1995b; Marsh et al., 2001). The southeastern Paraná and northwestern Etendeka regions are located where the postulated Tristan da Cunha plume trails, namely the Walvis Ridge and the Rio Grande Rise, intersect the African and South American continental margins, respectively. Principal crustal stretching along the East Brazil Rift system occurred between 130 and 120 Ma, with extension leading ultimately to the opening of the South Atlantic ocean basin (Chang et al., 1992).

A diversity of voluminous tholeiitic mafic and silicic volcanic units, together with numerous contemporaneous intrusive ring complexes (the Damaraland complexes; Fig. 1) containing a range of alkaline and tholeiitic rocks (extending to carbonatite and peralkaline granites), are characteristic of the Etendeka. This area clearly provides an unsurpassed opportunity to address questions relevant to the relationship between mantle plumes, flood volcanism, continental rifting, and sources of the mafic and silicic magmas.

Previous accounts of the Etendeka have focused on the southern Etendeka (Erlank et al., 1984; Milner et al., 1992, 1995a; Ewart et al., 1998a, 1998b), i.e. south of latitude 19°21'S that passes through Möwe Bay (Fig. 1). This location also approximately coincides with the southern limit of the high-Ti subprovinces within the Etendeka. This paper focuses on the northern Etendeka, extending northwards from, and including, the southern coastal subdomain (see below and Fig. 1).

	GEOLOGICAL SETTING

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
The northern Etendeka region comprises three structural domains, two coastal and one inland (Marsh et al., 2001). The coastal domains (Figs 1–3) lie within a 25 km zone along the coast from Terrace Bay northwards to Angra Fria. Along this zone the volcanic sequence is underlain by, and locally interbedded with aeolian Twyfelfontein (formerly Etjo) Formation sandstones (Stanistreet & Stollhoffen, 1999). The basal succession comprises 400 m of basaltic flows of the Khumib Formation (comprising both high-Ti and low-Ti basalts) overlain by a 400–500 m sequence of silicic flows with minor interbedded basalts (the Skeleton Coast Formation). The coastal domain comprises two subdomains, extending north and south of Möwe Bay (Figs 2 and 3, respectively), which expose two differing sequences of silicic volcanic units.

The entire coastal domain (preserved thickness of 1 km) has been intensively faulted by dominantly coast-subparallel normal faults. The volcanic sequences dip dominantly to the east and south and have been rotated with progressive downthrows to the west (Fig. 2). Net downthrows are estimated to be 290 m across a lateral distance of 22 km in the northernmost area, and a minimum of 200 m across 8 km in the Sarusas area (Fig. 2). Along its eastern boundary, the volcanic sequence is downfaulted against Proterozoic Damara basement granites and schists along a major detachment surface.

The inland domain comprises eroded remnants of only mafic, subhorizontal flows of the Khumib Formation, which include high-Ti (Khumib) and low-Ti (Tafelberg and Kuidas) types. No silicic volcanic rocks are exposed. Although downfaulted against basement rocks along their eastern margins, the volcanic sequence is relatively unfaulted. The overall east to west geological change across the northern Etendeka is one of increasing deformation towards the coast.

Dolerite dykes (Tafelberg, Esmeralda and Khumib types; see below) occur throughout, but are most abundant in the northern coastal domain, striking dominantly coast-parallel. Individual NW-trending dykes extend for up to 23 km, with widths from 1 to >100 m (dominantly 10–30 m). Rarer NE-trending dykes range between 1 and 20 m (dominantly 1–5 m) in width. Field relations indicate that intrusion of both dyke trends overlapped and was contemporaneous with faulting.

Inland of Cape Fria, the volcanic sequence is intruded by the Agate Mountain carbonatite complex (Fig. 2) of unknown age. The complex forms a roughly circular intrusive and breccia complex, 2·5 km in diameter, centred on a major NW-trending fault separating the basalt sequence from the stratigraphically higher Sarusas quartz latite. Fenitization extends for 2 km around much of the complex, and radially oriented dykes of carbonatite and rare phonolite extend for 7 km from the complex.

The Cape Fria ‘complex’ forms a distinctive suite of rocks cropping out between Agate Mountain and the coast. The complex comprises horizontal sheets of coarse-grained gabbroic and silicic volcanic units with chilled cappings, which are juxtaposed against the regional volcanic sequence along faulted or fault-scarp contacts. Geochemical data indicate that the gabbros and silicic volcanic rocks are chemically equivalent to some members of the regional volcanic sequence, dominantly the Esmeralda basalts and Fria quartz latite (see below), but also contain zones of hybrid composition. The Fria ‘complex’ is interpreted, from field data, as a ponded and relatively slowly cooled accumulation of basaltic and silicic lavas, locally intermixed, which have flowed into a contemporaneously active fault-controlled depression, somewhat analogous to Hawaiian lava lakes.

The Khumib, Tafelberg and Esmeralda basalt types correlate, respectively, with the Urubici, Gramado and Esmeralda types of the Paraná (Peate, 1997; Marsh et al., 2001). The Ribeira, Paranapanema and Pitanga basaltic suites, which occur in the northern Paraná, have not been recognized within the Etendeka and are therefore not considered in detail in this paper. Eight silicic eruptive units dominate the higher exposed stratigraphy of the northern subdomain. Only one (the Sarusas quartz latite) extends southwards into the southern coastal domain, where it is associated with quartz latites having affinities with the southern Etendeka types. The silicic eruptives units are described in the companion paper (Part 2, Ewart et al., 2004).

Paraná flood basalt volcanism is constrained between 129 and 134 Ma, with evidence for younger volcanism (120–128 Ma) along the coastal region (Peate, 1997). Eruptive ages of 127–131 Ma are reported for equivalent lavas from southern Uruguay (Kirstein et al., 2001). These compare with 127–132 Ma for the Etendeka (Milner et al., 1995b; Renne et al., 1996; Stewart et al., 1996; Kirstein et al., 2001). Associated Damaraland igneous complexes range in age from 123 to 137 Ma (Milner et al., 1995c; Renne et al., 1996; Schmitt et al., 2000).

	ANALYTICAL TECHNIQUES

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
Analytical techniques and sampling strategy
Approximately 550 rock analyses are available for the northern Etendeka, including those of Erlank et al. (1984) and Milner (1988), of which approximately 250 are of silicic types. As part of the mapping and documentation of the regional stratigraphy, major sampling was carried out between 1992 and 1998, the large number of samples analysed being necessitated by the difficulty of distinguishing, both in the field and petrographically, the various basaltic and the eight silicic rock types. A subset of 35 samples (Table 1) was selected for new isotopic and inductively coupled plasma mass spectrometry (ICP-MS) trace element analysis, representing unaltered samples of the Khumib and Tafelberg mafic types and the silicic types, excepting the very localized Naude quartz latite. As shown in Part 2 (Ewart et al., 2004), each of the silicic types is chemically relatively homogeneous. Additionally, three southern Etendeka Tafelberg samples (Erlank et al., 1984) were reanalysed by ICP-MS because of their unusual reported Pb isotope compositions (samples KLS048, 058, 100; Table 1). Five samples from Tristan da Cunha and Inaccessible Islands (Appendix, Table A1) were analysed by ICP-MS to provide a more complete set of trace element data than currently exists in the literature. The full set of northern Etendeka whole-rock data, including locations and isotope data, can be accessed from the Electronic Appendix, which may be downloaded from the Journal of Petrology web site at http://www.petrology.oupjournals.org.

View this table: [in this window] [in a new window]   Table 1: Major (wt %), trace element (ppm) and isotope data for selected northern Etendeka basaltic eruptives

 
X-ray fluorescence (XRF) analyses were carried out at the University of Cape Town (UCT) and Rhodes University, South Africa. After initial splitting and jaw crushing, samples were pulverized in carbon-steel mills for constant times. Splits of these were used for Sr and Nd isotope analyses at UCT. Samples for Pb isotope and ICP-MS analyses were crushed in agate.

Major elements were determined by the XRF fusion procedures of Norrish & Hutton (1969) with Na₂O analysed on pressed powder briquettes. XRF determined trace elements (Rb, Ba, Sr, Th, U, Zr, Nb, Cr, V, Sc, Ni, Co, Pb, Zn, Cu, Zn, Y, La, Ce, Nd) were also determined on pressed briquettes with corrections for background, absorption and spectral line interferences. A range of international standards were used for calibration. Duncan (1984) has provided a summary of the methods and precision of the XRF data.

ICP-MS analyses were carried out at the University of Queensland for Cs, Rb, Ba, Sr, Th, U, Pb, Zr, Hf, Nb, Ta, Ga, Ni, Cr, Cu, Zn, V, Sc, Co, Y, rare earth elements (REE), Li and Be. Samples, following crushing, were leached in 10% H₂O₅–5% HCl for 10 min at room temperature, and rinsed with distilled water. Analyses were carried out on a Fisons PQ2+ instrument, following the procedures of Eggins et al. (1997) and Murphy et al. (2002).

For consistency with the large existing set of Etendeka XRF analyses, the plots presented in this paper use the XRF values for Rb, Ba, Sr, Zr, Nb, Y, Cr, V, Co and Sc unless specified otherwise. For other elements duplicated by two techniques, ICP-MS data, or existing spark source mass spectrometry (SSMS) or instrumental neutron activation analysis (INAA) data, were preferred.

Sr–Nd–Pb isotopic analyses
Sr and Nd isotopic ratios were measured on a VG-Sector seven-collector mass spectrometer at the Radiogenic Isotope Facility, UCT, using unleached sample powders. Sr and Nd were measured in dynamic mode with mass fractionation corrected to ⁸⁶Sr/⁸⁸Sr = 0·1194, and the ⁸⁷Sr/⁸⁶Sr ratios normalized to an NBS SRM987 value of 0·71022. Nd isotope compositions were also measured in dynamic mode, corrected to ¹⁴⁶Nd/¹⁴⁴Nd = 0·7219, and normalized to the present-day ¹⁴³Nd/¹⁴⁴Nd value of 0·51264 for BCR-1, which is equivalent to ¹⁴³Nd/¹⁴⁴Nd = 0·51184 for the La Jolla standard.

Pb isotopes were analysed on purified samples in static multi-collector mode using a Finnigan MAT 261 at the Research School of Earth Sciences, Australian National University (with the exceptions noted below). Approximately 200 mg splits of rock powder were dissolved in HF–HNO₃ and the Pb was separated and purified by normal double-pass HBr–HCl anion exhange methods using Teflon microcolumns. To correct for mass fractionation, the double spike technique was used (Woodhead et al., 1995; Woodhead & Hergt, 1997) with a spiked and an unspiked split run from each sample. In two analyses where the spiked runs failed, correction for mass fractionation was made using an empirical factor based on values obtained for the standard SRM 981 (measured values, N = 17: ²⁰⁶Pb/²⁰⁴Pb = 16·9363; ²⁰⁷Pb/²⁰⁴Pb = 15·4915; ²⁰⁸Pb/²⁰⁴Pb = 36·7058). Total method Pb blanks of near 50 pg were measured during this work and no blank corrections were made to the final data. Uncertainties quoted are ±2 SE of the means for the measured unspiked runs. Five samples (KLS311, 327; SM110, 112, and 115) are previously unpublished analyses undertaken at UCT, using procedures listed by Ewart et al. (1998a). Repeat Pb isotope analyses of three southern Etendeka Tafelberg samples (see above), together with Sr–Nd–Pb analyses of the five Tristan da Cunha and Inaccessible Islands samples (see above) were analysed in the ACQUIRE laboratories at the University of Queensland, following the procedures listed by Collerson et al. (2002). New isotopic data for the Etendeka mafic samples are presented in Table 1, and the Tristan samples in the Appendix, Table A1.

	MAGMA TYPES AND THEIR DISTRIBUTION

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
The Etendeka province is bimodal with respect to the SiO₂ content of the regional magma types (including dykes; Milner et al., 1995b), a feature shared with the Paraná (Peate, 1997). Although still bimodal (Fig. 4), the northern Etendeka contains a higher proportion of intermediate compositions, mainly as latites and abundant high-Ti quartz latites (see Part 2, Ewart et al., 2004), the latter with generally lower SiO₂ concentrations than the low-Ti equivalents that predominate in the southern Etendeka (Milner, 1988; Milner et al., 1992; Ewart et al., 1998b).

View larger version (23K): [in this window] [in a new window]   Fig. 4. Histograms of silica concentrations showing the bimodal character of the northern and whole Etendeka region, excluding the Damaraland intrusive complexes. The bimodality of the northern Etendeka is reduced by the occurrence of latites that are very rare in the southern Etendeka. Data from Erlank et al. (1984); Ewart et al. (1998a, 1998b); Milner (1988), and references therein; Marsh et al. (2001); and this study (see Electronic Appendix for the full dataset).

 
In the total alkalis–silica (TAS) classification, the low-Ti Tafelberg and high-Ti Khumib samples range from basalt, through basaltic andesite, basaltic trachyandesite to andesite and trachyandesite [as discussed by Marsh et al. (2001)]. For clarity, we simply refer to these as the Tafelberg and Khumib basaltic series, unless otherwise specified. Within the Etendeka, the distinction between the high-Ti and low-Ti basaltic series is based on TiO₂ and Sr, the boundaries taken at 2·2 wt % and 450 ppm, respectively (Marsh et al., 2001). Peate et al. (1992) used the Ti/Y ratio for the Paraná lavas with the boundary taken at 310. This criterion, however, does not adequately subdivide the chemistries of the Etendeka mafic lavas, although it is a useful discriminant ratio.

Eight geochemically distinct mafic magma types are recognized in the entire Etendeka Province (Marsh et al., 2001), of which four—the high-Ti Khumib and the low-Ti Tafelberg, Kuidas and Esmeralda types—occur in the northern Etendeka. The Tafelberg and Khumib types dominate the lower part of the volcanic sequence (together forming the Khumib Formation), and the Esmeralda types are interbedded with silicic types in the overlying part of the sequence (comprising the Skeleton Coast Formation). The proportion of Khumib type flows increases inland, but decreases southwards, terminating north of Terrace Bay (Fig. 1). Only five Khumib dykes are known, all from north of latitude 18°27'S. Esmeralda type lavas and dykes, although more common in the northern Etendeka, extend into the southern Etendeka. The Tafelberg type flows and dykes are widespread along the coastal domains and throughout the southern Etendeka.

The Khumib basaltic series are compositionally variable (MgO 2·4–11·4 wt %; SiO₂ 49·1–58·5 wt %), being more diverse than the equivalent Paraná Urubici basalts (MgO 3·7–5·4 wt %). More Mg-rich flows occur at or close to the base of the sequence. Despite extensive flow-by-flow sampling, we cannot recognize the compositional stratigraphy described by Peate et al. (1999) in the Urubici sequence. Northern Etendeka Tafelberg basaltic series [including the regional dolerites of Erlank et al. (1984)] are also compositionally variable (MgO 2·9–8·0 wt %; SiO₂ 50·0–57·8 wt %), but without the high MgO compositions (up to 23·5 wt %) that sporadically occur in the southern Etendeka (see Fig. 6, below).

View larger version (36K): [in this window] [in a new window]   Fig. 6. Selected variation diagrams, vs SiO2 and MgO, for the Khumib type basalts showing the existence of high-MgO and low-MgO subsets. The higher-MgO subset is considered part cumulative, and the sample numbers shown in (d) are those used in specific modelling (Table 4) of trends 1 and 2 shown in inset in (d). The inset in (a) shows schematic fractionation vectors at low fO2 (trend 1) and high fO2 (trend 2), as discussed in text.

 
The low-Ti Esmeralda basalts are also compositionally variable (MgO 6·1–3·2 wt %), but with characteristic high FeO_tot. This, and their low Zr and Zr/Y ratios [for a given Mg number; mol. MgO x 100/(MgO + FeO_tot)], distinguishes them from the Tafelberg series. The Esmeralda lavas and dykes occur interbedded with, and intruding the silicic volcanic sequences. The low-Ti Kuidas type is volumetrically insignificant, although widespread as flows throughout the Etendeka. The lavas are relatively magnesian (MgO 7·8–8·9 wt %), showing little variation, and isotopically distinct with higher _Nd compared with the Tafelberg series. The distributions of these magma types throughout the Etendeka region have been summarized by Marsh et al. (2001).

Although not exposed within the northern Etendeka, three additional southern Etendeka magma-types are important petrogenetically and are included in the petrogenetic interpretations (see below): (1) the Horingbaai dolerites, a volumetrically minor dyke phase with no known volcanic equivalents, were emplaced in the later stages of Etendeka volcanism; they are chemically similar to E-MORB (Erlank et al., 1984; Duncan et al., 1990; Thompson et al., 2001); (2) the relatively small-volume Tafelkop series (ferropicrites of Gibson et al., 2000; Thompson et al., 2001), limited to the Goboboseb Mountains and Doros regions (Fig. 1; Jerram et al., 1999), and identified as plume related (Milner & le Roex, 1996; Ewart et al., 1998a; Thompson et al., 2001); (3) the Nil Desperandum dolerites identified by Thompson et al., (2001), which may be related to the Huab sill complex (Duncan et al., 1989). We note that the use of Nil Desperandum duplicates a name previously used for a southern Etendeka latite (Marsh et al., 2001). We therefore use the alternative name ‘Florida type’ for these dolerites, from one of the relevant localities.

	MINERALOGY

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
The low-Ti Tafelberg, Esmeralda and Kuidas lavas are aphyric to sparsely phyric, the rare phenocryst or microphenocryst phases being plagioclase, augite, less common pigeonite and rare titaniferous magnetite (Table 2). Groundmass phases are the same, with the calcic pyroxenes extending into the subcalcic range, and coexisting with pigeonite. Olivine has not been identified in the northern Etendeka low-Ti basalts, although olivine (Fo_85–87) has been reported in magnesian dykes from the southern Etendeka (Erlank et al., 1984). Application of the Lindsley (1983) two-pyroxene geothermometer to phenocryst, microphenocryst and groundmass phases in lavas indicates temperatures of 1150–1200°C. Dolerite dyke pyroxenes are compositionally more variable than those in the lavas, with extreme development of subcalcic compositions, although whether these pyroxenes are homogeneous is uncertain (Mellini et al., 1988). Two pyroxene temperatures indicate a range between 1100 and 1250°C.

View this table: [in this window] [in a new window]   Table 2: Summary of mineralogy of northern Etendeka mafic rock series

 
The Khumib lavas are aphyric to sparsely phyric in the lower MgO types, but phyric in the lavas with >5 wt % MgO (Table 2). Phenocrysts in the latter comprise augite, orthopyroxene, plagioclase, and altered olivine and pigeonite. Orthopyroxene has not been found in the less magnesian lavas. Groundmass phases are typically plagioclase, augite–subcalcic augite, pigeonite, titaniferous magnetite and ilmenite. Two-pyroxene thermometry (Lindsley, 1983) suggests temperatures in the range 1050–1150°C (<5 wt % MgO types) and 1100–1200°C (>5 wt % MgO types). The occurrence of orthopyroxene in the more magnesian basalts is significant as it appears to be rare in tholeiites of large igneous provinces (Demarchi et al., 2001). In some Khumib samples, the orthopyroxene is cumulate (see below) which may explain the existence of bimodal orthopyroxene compositions in some samples (Table 2).

The Cape Fria gabbros comprise coarse phenocrystal plagioclase, augite, magnetite and microphenocrystal apatite set in very fine-grained interstitial matrix (20–35 vol. %, Table 2). Crystallization occurred in situ within surface ponded lava lakes, not all necessarily interconnected, explaining the subtly differing mineralogical and chemical variations within different outcrops. Major element chemistry (Fig. 5) suggests that most gabbros correlate with the low-MgO end of the Esmeralda basalt arrays. Localized, roughly linear hybrid zones of intermediate composition are recognized, especially near the western margin of the complex, in which up to three populations of pyroxene are observed, as well as bimodal plagioclase compositions (Table 2). These features indicate magma mixing, probably between ponded quartz latite and low-Ti (Esmeralda) basaltic flows (see above).

View larger version (38K): [in this window] [in a new window]   Fig. 5. Selected variation diagrams, relative to MgO, of the northern Etendeka basalts and dolerites. The enclosed fields are the fields of the southern Etendeka Tafelberg series. The Khumib series are classed as a high-Ti suite, the remaining series being low-Ti suites. The calculated fractionation curves [inset in (d)] are derived from a Khumib basalt with plag + aug + pig + mt (curve 1), plag + oliv + aug + mt + ilm (curve 2), with basaltic partition coefficients (Appendix, Table A2). Curves 3 and 4 are calculated using dacitic partition coefficients with the same mineral assemblages as for curves 1 and 2, respectively.

 

	CHEMISTRY

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
Introduction
All basalt types, irrespective of MgO concentrations, are predominantly silica-saturated tholeiites (Fig. 5a), with only a small number marginally silica-undersaturated. Compared with the Tafelberg type, Khumib basalts are enriched in TiO₂, P₂O₅, Zr, Sr, Nb, Ba, light REE (LREE) and middle REE (MREE) for a given MgO concentration (Fig. 5c, e and f) or Mg number. MgO, Ni and Cr also tend to be higher in the Khumib basalts (Fig. 5d), especially in the inland domain. Other incompatible elements, such as K₂O, Rb, U, Th, Pb, Y and heavy REE (HREE), show substantial to complete overlap between the two types. Ratios of incompatible elements between these two groups are useful discriminators, e.g. Ti/Y, Zr/Y, Th/Yb, Nb/Yb, Zr/Nb, Ti/Nb, Tb/Yb, Ce/Pb and Sr/Pb (Table 3).

View this table: [in this window] [in a new window]   Table 3: Selected trace element ratios for Paraná–Etendeka basaltic magma types and comparative data

 
The low-Ti Esmeralda and Kuidas basalts share many geochemical features with the Tafelberg basalts (Table 3), but are distinguished by higher Ti/Zr and lower Zr/Y (Fig. 5e), and for the Esmeralda basalts, high FeO_tot (for given MgO; Fig. 5b). Esmeralda basalts possess relatively low Mg number (50–28) whereas Kuidas basalts have consistently high Mg number (67–60). In the elemental plots (Fig. 5; and Figs 10–15, 17 and 18, below), the Paraná (Peate & Hawkesworth, 1996) and Etendeka Esmeralda chemical data are combined.

View larger version (27K): [in this window] [in a new window]   Fig. 10. Trace element ratio plots for the mafic magma types of the Etendeka; the Cretaceous ultra-alkaline rocks (lamproites, kimberlites, lamprophyres and related types) of Namibia the Paraná Gramado (low-Ti), Esmeralda (low-Ti; Paraná and Etendeka data combined) and Urubici (high-Ti) series; and Walvis Ridge and Tristan da Cunha lavas. Data sources listed in the text. Compositional arrays of low-Ti and high-Ti silicic volcanics are shown as dashed closed fields; the low-Ti quartz latites are from the southern Etendeka (Milner, 1988; Ewart et al., 1998b). Upper-crustal field based on Rudnick & Fountain (1995) and Taylor & McLennan (1995); Namibian and Paraná crustal basement compositions have been listed by Ewart et al. (1998b). Comparative mean MORB and OIB (Sun & McDonough, 1989), GLOSS (Plank & Langmuir, 1998), Gaussberg [Murphy et al., 2002; not shown in (b) as out of range of plot] and Pitcairn (Eisele et al., 2002) compositions are shown. Schematic mantle array is shown in (a). The Walvis Ridge Site 525A data are highlighted in (b) because of their geochemical similarity to the Khumib lavas. The insets show fractional crystallization vectors calculated for plag + aug + pig + mt + apat phase assemblage, to F = 0·5 (at termination of each vector), using basaltic (1) and dacitic (2) partition coefficients (Appendix, Table A2).

 

View larger version (30K): [in this window] [in a new window]   Fig. 11. (a) Sr/Pb–Nb/Nb* plot illustrating the correlation between two potential sediment (crust) trace element indices within the Etendeka and equivalent Paraná basalts, the Tristan lavas and the Namibian ultra-alkaline suites. The Paraná ultra-alkaline suites are shown by the enclosed field, as are the low-Ti and high-Ti silicic volcanic series, and model upper-crustal compositions. Nb/Nb* = NbN/(ThN · LaN), normalized to primitive mantle, following Eisele et al. (2002). (b) Eu/Eu*–Sr plot for the Etendeka and equivalent Paraná volcanic series (as above) showing the absence of systematic negative Eu/Eu* anomalies in the mafic series excepting the low-Ti basalts. The Namibian and Paraná (field only) ultra-alkaline suites exhibit negative Eu/Eu* anomalies only in a very small number of samples. The high-Ti silicic series exhibit progressive Eu depletion, extending from the Khumib array, and into the more strongly Eu-depleted low-Ti quartz latites. The insets represent fractional crystallization vectors, as described in Fig. 10 caption. Symbols, model compositions, and data sources as in Fig. 10 key and caption. SCLM composition from McDonough (1990).

 

View larger version (36K): [in this window] [in a new window]   Fig. 12. Initial (132 Ma) Nd–Sr isotopic compositions of the Etendeka and equivalent Paraná basalts, Walvis Ridge and Tristan, the fields of the Namibian and Paraná ultra-alkaline suites, and the fields of the high-Ti (extending from the Khumib compositions) and low-Ti (in part) silicic series. The postulated EM1 composition is after Zindler & Hart (1986). Symbols, model compositions and data sources as in Fig. 10 key and caption.

 

View larger version (48K): [in this window] [in a new window]   Fig. 13. Initial Pb isotopic compositions of the Etendeka and equivalent Paraná basalts, Walvis Ridge and Tristan (a) and (c), and (b) the Namibian and Paraná ultra-alkaline suites (see text for data sources). The EM (Erosion Mix) and DM (Depleted, MORB Mantle) growth curves are from Kramers & Tolstikhin (1997). The LM (Lower Mantle) growth curve and OIB array are from Kamber & Collerson (1999). Ticks on the growth curves represent 0·1 Gyr intervals, except where indicated otherwise. Symbols in (a) and (d), and data sources as in Fig. 10 key and caption. Data sources for (d) are listed in the text. (c) illustrates an interpretation of the chemistry of the Tristan da Cunha compositions in terms of mixing between plume and depleted asthenospheric magmas (Kamber & Collerson, 1999), and their extrapolation to the inferred original Tristan plume compositions at 70 Ma (Walvis Ridge, DSDP 525A site age) and 132 Ma (Etendeka age). The extension of the Tristan array is calculated as a linear extrapolation between the DM curve and the HIMU extension of the LM evolution curve, first by a least-squares fit to the Tristan array, and second by constraining the DM composition at 132 Ma. The HIMU extension is calculated from two-stage LM Pb isotope evolution following an abrupt increase in µ2 accompanying the fractionation event terminating stage 1. This is back-calculated 150 Myr to allow for evolution and rise of the Tristan plume, based on Discovery Tablemount data, following Kamber & Collerson (1999). The extension represents points of progressively increasing µ2, trending towards HIMU compositions. The two calculated intersections of the Tristan data lie at approximately µ2 = 47, reducing to 35 when the data are forced through the 132 Ma point on the DM curve. These µ2 values are used to recalculate two sets of initial 206Pb/204Pb and 207Pb204Pb ratios at both 70 Ma and 132 Ma.

 

View larger version (38K): [in this window] [in a new window]   Fig. 14. Initial Sr (a) and Nd (b) isotopic compositions vs initial 206Pb/204Pb ratios of the Etendeka and equivalent Paraná mafic basalts, Tristan and Walvis Ridge compositions, the subset of unradiogenic Jacupiranga carbonatite (Brazil) compositions, and Namibian and Paraná ultra-alkaline suites. The low-Ti and high-Ti silicic compositions are represented as fields, the latter extending from the Khumib array. It should be noted that in (a), the Jacupiranga, Khumib–Urubici–Tafelkop, Walvis Ridge, Tristan, and ultra-alkaline data define a linear array with increasingly radiogenic Sr correlating with increasingly unradiogenic Pb isotope compositions. In (b), a similar array is apparent, but excludes the Khumib–Urubici compositions, which project below the array. Symbols, data sources and mean compositions of GLOSS and Gaussberg as in Fig. 10 key and caption; ultra-alkaline data sources listed in text.

 

View larger version (30K): [in this window] [in a new window]   Fig. 15. Nb/Nb* vs initial Sr and Nd isotopic compositions, providing one example of the correlations between a possible ‘sediment-index’ trace element parameter and Sr–Nd isotopic compositions. Symbols, model compositions, and data sources as in Fig. 10 key and caption and text (ultra-alkaline suites).

 

View larger version (33K): [in this window] [in a new window]   Fig. 17. Plots of (a) La/Sr–La/Yb, and (b) Sr/Cs–Pb/Cs, for the Etendeka and equivalent Paraná basalts, Tristan and Walvis Ridge basalts, and Namibian ultra-alkaline suite. The Paraná ultra-alkaline compositions (comprising three separate arrays), and the compositions of the low-Ti and high-Ti silicic series are shown in the enclosed fields. The paucity of Cs data as illustrated in (b) should be noted. In both plots, two broad trends are apparent. Symbols, data sources and model compositions as in Fig. 10 key and caption, and in text.

 

View larger version (60K): [in this window] [in a new window]   Fig. 18. Trace element and isotopic plots to illustrate the proposed multiple mixing model (see text) of mafic magma genesis in the Etendeka and the Paraná, involving sub-lithospheric mixing between plume-derived magma (Tristan), depleted asthenospheric mantle magma (N-MORB), and the Gaussberg magma composition (an inferred product of ancient subducted sediment remelting; Murphy et al., 2002). GLOSS represents a pre-subduction sediment and crustal contaminant composition. The mixing curves between each end-member are marked at 20% mixing intervals, except where otherwise noted. The shaded fields represent the main Paraná–Etendeka magma types: K-U, Khumib–Urubici; TK, Tafelkop; F, Florida; H, Horingbaai; E, Esmeralda; T-G, Tafelberg–Gramado. Also shown in (d) are the high-Ti/Y and low-Ti/Y boundary, and the ranges of Ti/Y ratios of the high Ti/Y Ribeira, Paranapanema and Pitanga basalt types of the northern and central Paraná (Peate, 1997).

 
The Tafelberg, Khumib and Esmeralda basalt types exhibit variable and overlapping ranges of MgO concentrations. Incompatible element abundances (e.g. Zr, Ti, P, Nb, K, Ba) increase and compatible element (e.g. Ca, Cr, Ni) abundances decrease systematically with declining Mg number within each of the basalt types (e.g. Fig. 5c and f). This is consistent with fractional crystallization control on the compositional spectrum observed within the basalts, an aspect documented in detail for the southern Etendeka Tafelberg series by Erlank et al. (1984) and Ewart et al. (1998a). It has also been documented by Peate & Hawkesworth (1996) and Peate et al. (1999) for the Gramado and Esmeralda equivalents in the Paraná, and is therefore not treated further in this paper. Major and trace element compositions (and isotopic compositions; see below) of the southern and northern Etendeka Tafelberg types exhibit almost complete overlap (Fig. 5), attesting to the widespread distribution of this magma type. It forms a regionally distributed and strongly heterogeneous magma type within the Etendeka, mirrored by the equivalent Paraná Gramado rocks. In the descriptions and discussions that follow we use all the available Tafelberg data.

Despite the northern Etendeka region lying at the intersection of the Tristan plume trace (i.e. the Walvis Ridge) with the continental margin (Fig. 1), none of the northern Etendeka mafic magma types have compositions similar to the plume-related Tafelkop picritic basalts identified in the southern Etendeka.

Khumib basalt type—major element trends
Two significant features of the Khumib basalts are: (1) their compositional range with element variations broadly consistent with crystal fractionation; (2) the presence of phyric, relatively MgO-rich lavas at the base of the sequence. Figure 6 highlights some distinctive trends identified within the high-MgO subset (>5 wt % MgO) of Khumib samples and demonstrates that they do not group at the primitive end of a simple liquid line of descent.

View this table: [in this window] [in a new window]   Table 4: Least-squares phenocryst accumulation models for the Khumib type high-MgO variant lavas

 
Amongst the MgO-rich basalt subset (Fig. 6), the SiO₂-poor samples contain phenocrystal olivine and augite, whereas olivine is absent in the more SiO₂-rich rocks, which have abundant augite + orthopyroxene ± pigeonite + plagioclase. The displaced, but parallel trend of the high-MgO samples in the MgO–SiO₂ plot (Fig. 6d) suggests that they may have developed from low-MgO basalts by accumulation of mafic phenocrysts. This has been modelled by least-squares mixing calculations (Table 4), using samples identified in Fig. 6, which support an accumulation trend for Si-poor samples (trend 2 in Fig. 6d inset), involving olivine and orthopyroxene only. In contrast, the more evolved compositions (trend 1) are dominated by accumulation of augite and lesser amounts of plagioclase with minor involvement of pigeonite (or olivine). Although the calculated r² values for these models are slightly high (range: 0·17–0·34, the largest residuals for Na and K), the results support the feasibility of the proposed accumulation processes.

Compositional variation in the Paraná low-MgO Urubici lavas was shown by Peate et al. (1999) to be consistent with fractional crystallization of olivine + augite + plagioclase for generating evolved TiO₂-enriched derivatives from lower TiO₂ parental magma compositions. As this fractionation assemblage produces little SiO₂ enrichment, Peate et al. (1999) postulated models incorporating magnetite to generate more evolved SiO₂-rich compositions. In the low-MgO Khumib basalts, although trends of decreasing CaO, Cr and Ni, and increasing SiO₂, TiO₂, P₂O₅, FeO and K₂O (and incompatible trace elements) correlate with decreasing MgO (e.g. Fig. 6d–f), comparable plots vs SiO₂ do not exhibit such clear correlations (Fig. 6a–c). The higher TiO₂ concentrations occur within the lower SiO₂ samples, with FeO, TiO₂ (Fig. 6a and b) and V concentrations decreasing with increasing SiO₂. It appears that although the bulk of the low-MgO Khumib suite has evolved by fractional crystallization, discrete oxide-poor and oxide-rich trends are not evident within the data. It is suggested that either the proportion of oxide in the fractionating assemblage varied widely and unsystematically between magma batches, or there has been mixing of magmas evolving along the oxide-present and oxide-absent trends.

The MELTS software (Ghiorso & Sack, 1995) was used to test the possible effects of f_O2 on magnetite crystallization (at 2 kbar) of sample KLS574 (51·57 and 2·96 wt % SiO₂ and TiO₂, respectively; see Electronic Appendix). At f_O2 = NNO (nickel–nickel oxide), the phase assemblage of augite, low-Ca pyroxene and plagioclase is joined at F = 0·53 by Ti-magnetite, which forms 40 wt % of the precipitating solid fraction, producing a liquid with 55·43 and 3·85 wt % SiO₂ and TiO₂. At f_O2 = QFM + 3 log units (where QFM is quartz–fayalite–magnetite), the assemblage augite and plagioclase is joined by Ti-magnetite at F = 0·68, which forms 56 wt % of the precipitating solid assemblage, producing a liquid with 58·79 and 3·21 wt % SiO₂ and TiO₂, respectively. We suggest that these calculations support the existence of contrasting fractionation paths, at differing f_O2, leading to differing levels of Ti and Si enrichment in derivative magmas (Fig. 6a inset).

Comparable calculations to model the crystallization of orthopyroxene, and its observed bimodal composition, within Khumib sample KLS554 (Table 1), predict the following (anhydrous; f_O2= QFM). At 1 bar; liquidus olivine (Mg₇₈) olivine (Mg₇₃) + plagioclase (An₆₀) pigeonite (Mg₆₀Ca₁₂) + augite (Mg₄₄Ca₃₉) + plagioclase (An₄₉) + spinel. At 2–10 kbar, the sequence is: liquidus orthopyroxene (Mg_75–77Ca₄) pigeonite (Mg₆₈Ca₁₀) pigeonite (Mg_58–51Ca_13–23) + augite (Mg_49–37Ca_28–39) + plagioclase (An_49–52, reduced stability at 10 kbar). Olivine occurs only at the lowest pressures, under both hydrous (4 kbar) and anhydrous (<2 kbar) conditions. The predicted phase mineralogy and compositions agree with observed phenocryst phases at pressures 2 kbar. Mg-rich orthopyroxene compositions observed within this sample are not consistent with the rock bulk composition, but may represent a relict composition from an earlier phase of cumulate crystallization.

	TRACE ELEMENTS

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
Introduction
The trace element characteristics of the Etendeka Tafelberg and Khumib type lavas and dykes have been extensively described by Erlank et al. (1984), Ewart et al. (1998a) and Marsh et al. (2001), and are only briefly noted here. In the following section we focus on samples whose XRF determined trace element compositions have been supplemented by additional trace elements (REE, Li, Be, Cs, Pb, Th, U, Ta, Hf, Nb) determined by SSMS, INAA or ICP-MS. Approximately 300 samples have been selected on the basis of these criteria. In the plots and discussions that follow, we concentrate on data from the northern Etendeka, but include analyses of the wider Paraná–Etendeka Province as follows:

1. Southern Etendeka (Erlank et al., 1984; Ewart et al., 1998a), including the southern Etendeka Tafelkop, Horingbaai and Florida magma types (Jerram et al., 1999; Gibson et al., 2000; Thompson et al., 2001) excluding samples identified by the last authors as crustally contaminated, plus the Tafelberg series;
   
2. The Gramado, Esmeralda and Urubici series from the Paraná (Peate & Hawkesworth, 1996; Peate et al., 1999) and other high-Ti and low-Ti basalts from southern and northern Paraná (Marques et al., 1999);
   
3. comparative data from Tristan da Cunha and Inaccessible Islands, and Walvis Ridge (O'Nions & Pankhurst, 1974; Richardson et al., 1982; Humphris & Thompson, 1983; Weaver et al., 1987; le Roex et al., 1990; Cliff et al., 1991; Appendix, Table A1);
   
4. Namibian Cretaceous lamproitic, lamprophyric, carbonatitic and kimberlitic intrusives (Milner & le Roex, 1996; le Roex & Lanyon, 1998; Davies et al., 2001);
   
5. the complex series of alkaline and ultra-alkaline igneous suites (lamprophyres, lamproites, kimberlites, etc.) associated with the Paraná region (Comin-Chiaramonti et al., 1992; Gibson et al., 1995a, 1995b; Carlson et al., 1996; Thompson et al., 1998).
   

Relations between magma types
Northern Etendeka REE data (Table 1; Fig. 7a) show overlapping HREE concentrations, but LREE enrichment of the Khumib compared with Tafelberg samples. Comparison with the new Tristan da Cunha data (Appendix, Table A1) and the southern Etendeka Tafelkop, and Horingbaai basalts (Fig. 7b) indicate: (1) similar MREE and HREE abundance patterns between the Tristan and Khumib lavas, with higher LREE–MREE fractionation in the Tristan lavas (mean La/Yb = 27 compared with 14·6 for Khumib; Table 3); (2) overlap of LREE abundances between the Tafelkop and Tafelberg types, but La/Yb ratios differ (mean 13·5 compared with 7·2, respectively); (3) markedly lower LREE abundances within the Horingbaai dykes, the mean La/Yb ratio being 2·8.

View larger version (29K): [in this window] [in a new window]   Fig. 7. N-MORB-normalized REE plots comparing: (a) new ICP-MS analysed NW Etendeka Khumib and Tafelberg basalts (Table 1), and (b) ICP-MS analysed Horingbaai and Tafelkop basalts (Gibson et al., 2000; Thompson et al., 2001; using only samples showing least crustal contamination) and the two most magnesian Tristan da Cunha basalts (Appendix, Table A1). Normalizing values after Sun & McDonough (1989).

 
Normal mid-ocean ridge basalt (N-MORB)-normalized multi-element variation plots (Fig. 8a) for the northern Etendeka Tafelberg and Khumib samples are consistent with the well-established patterns of enhanced LILE and HFSE abundances in the Khumib magmas, except for Cs, Rb, Th, U and K. The most conspicuous features of both magma types, relative to N-MORB, are the enhanced concentrations of Ba, Pb (and Cs, Rb and ±U within the Tafelberg samples), with relative depletions of Nb–Ta, Sr and Ti and possible weak depletions of Zr and Hf. These features are most strongly developed within the Tafelberg magmas and comparable trace element signatures, of varying intensities (Fig. 8b), are also present in the Tafelkop and Horingbaai samples (and Esmeralda and Florida magma types, not shown for clarity), and weakly developed in the new Tristan da Cunha data. Similar characteristics are also observed in lamproites (e.g. Fraser et al., 1985; Ellam & Cox, 1991; Murphy et al., 2002).

View larger version (37K): [in this window] [in a new window]   Fig. 8. N-MORB-normalized multi-element plots of the same sample sets as shown in Fig. 7, comparing (a) ICP-MS analysed northern Etendeka Khumib and Tafelberg basalts (Table 1), and (b) ICP-MS analysed most magnesian Tristan da Cunha (Appendix, Table A1) and Horingbaai and Tafelkop basalts (Gibson et al., 2000; Thompson et al., 2001; using only samples showing least crustal contamination).

 
The possible effects of crustal contamination on the Tafelberg magma compositions require consideration in view of their variable and generally radiogenic Sr isotope compositions. As shown by Erlank et al. (1984), Tafelberg samples with high and low Sr isotopic compositions have very similar major and trace element compositions, and this observation is not changed by the extended data presented here, as clearly illustrated in Fig. 9. The Tafelberg trace element patterns have a strongly developed ‘crustal’ character that is little modified by addition of, for example, 20% of model bulk continental crust composition (Taylor & McLennan, 1995) irrespective of their initial Sr isotopic ratios. This is consistent with the conclusions of Erlank et al. (1984) that simple crustal contamination or assimilation–fractional crystallization (AFC) processes alone cannot explain the combined isotopic and compositional characters.

View larger version (39K): [in this window] [in a new window]   Fig. 9. N-MORB-normalized multi-element plots of Etendeka (whole region) Tafelberg basalts with initial 87Sr/86Sr ratios <0·710 (a) and >0·710 (b). The overall similarity of element abundance patterns, and overlapping SiO2 concentrations, should be noted. The <0·710 subset includes the unradiogenic Pb isotope samples. Data from Erlank et al. (1984), Milner (1988), Ewart et al. (1998a) and Table 1.

 
In the Th/Yb vs Nb/Yb plot (Fig. 10a), the mantle array is defined by MORB, Tristan and the Namibian ultra-alkaline suites. Within this array project the Horingbaai, Florida and Tafelkop series. The Tafelberg rocks, however, are clearly displaced towards crustal compositions, extending into the fields of low-Ti quartz latites. The Khumib basalts and the high-Ti lavas from northern Paraná (Marques et al., 1999) are slightly displaced from the mantle array towards the crustal fields, as is the high-Ti latite–quartz latite array, which is also shifted towards higher Th/Yb and Nb/Yb ratios. The overlap of Tafelkop compositions with the Khumib and mantle arrays is noted. Zr/Yb vs Nb/Yb relations (Fig. 10b, illustrated because of the absence of Th and Pb data for Walvis Ridge rocks) show that the Etendeka, Tristan, Walvis Ridge and the Namibian ultra-alkaline suites each define separate arrays, suggestive of mixing, with the Etendeka arrays also overlapping the crustal fields (not shown for clarity). The Walvis Ridge samples from Deep Sea Drilling Project (DSDP) Site 525A overlap the Khumib compositions, the overlap extending to other trace element and isotopic compositions, as recognized by Milner & le Roex (1996). The main Walvis Ridge array (DSDP Sites 527, 528) effectively connects the Tristan and enriched MORB (E-MORB) compositions, consistent with observations of Humphris & Thompson (1983).

Crustal–sediment signatures
The Tafelberg–Gramado and Esmeralda basalts exhibit pronounced Nb–Ta depletions, also developed more weakly in the Khumib–Urubici series. Nb–Ta depletion is a characteristic of arc-related magmas, continental crust and crustally derived sediments (e.g. Taylor & McLennan, 1995). Crustal contamination will therefore result in negative Nb–Ta anomalies. This was quantified by Eisele et al. (2002), who defined the parameter Nb/Nb^* [= Nb_N/(Th_N · La_N), normalized to primitive mantle], which provides a potential index of sediment input.

Nb/Nb^* correlates with a range of trace element (and isotopic) ratios (see below), for example, Ce/Pb, Nb/U, Eu/Eu^*, Ti/Y, La/Yb, La/Nb and Sr/Pb (e.g. Fig. 11a, Table 3). The Tafelberg and Esmeralda rocks have the lowest Nb/Nb^* of the mafic magma types, i.e. a signature of highest crustal input. Nb/Nb^* in the Khumib lavas extends to higher values, similar to the Tafelkop lavas, which in turn overlap the Tristan da Cunha and Horingbaai compositions. Thus there is a continuous trend of apparently increasing crust–sediment input extending from the Namibian ultra-alkaline intrusions through the high-Ti Khumib and Tafelkop types to the low-Ti Tafelberg magmas and the high-Ti and low-Ti silicic suites (Fig. 11a). The Tristan samples also appear slightly offset towards lower Nb/Nb^* compared with the compiled ocean island basalt (OIB) composition of Sun & McDonough (1989) and the ultra-alkaline compositions. The Horingbaai dykes exhibit some variation of Nb/Nb^*, but project close to the model N-MORB compositions (also seen in Fig. 10).

Eu/Eu^* vs Sr systematics (Fig. 11b) illustrate the development in the Tafelberg (and some Esmeralda) magmas of negative Eu/Eu^* anomalies, which are not, however, correlated with decreasing Sr. Similar negative Eu anomalies are not present in other Etendeka mafic types, nor the Tristan lavas (except some very fractionated trachytes). Although feldspar-dominated fractionation will produce negative Eu anomalies, reference to the upper-crustal and Global Subducting Sediment (GLOSS; Plank & Langmuir, 1998) compositions indicates that crust–sediment mixing or assimilation can also contribute to the development of such anomalies.

In summary, most aspects of the trace element geochemistry suggest possible, but variable extent of continental crust–sediment contributions to the Etendeka mafic magmas, the strongest input occurring in the compositionally very variable Tafelberg and Esmeralda basalts. This is certainly not a new observation, as previous workers identified crustal signatures in the Tafelberg magmas (e.g. Erlank et al., 1984; Peate & Hawkesworth, 1996; Ewart et al., 1998a). Trace element signatures of sediment input are also apparent in the compositions of the Khumib, primitive Tafelkop, Horingbaai and Tristan magmas. In most previous work, these signatures have been attributed to crustal assimilation. Importantly, however, trace element data do not provide implicit information as to where and how these sediment signatures were acquired, whether it was by direct assimilation by the magmas, or inherited from their mantle source(s).

	ISOTOPIC CHEMISTRY

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
Sr and Nd isotope compositions
Initial Nd–Sr isotopic compositions (corrected to 132 Ma) of the northern Etendeka rocks (Table 1), together with relevant previously published data, are plotted in Fig. 12. The Khumib and Tafelberg basalts define separate arrays with minimal overlap. Khumib compositions lie on the extension of the mantle array in the enriched quadrant, towards the field of the Paraná low-Ti potassic mafic rocks (Gibson et al., 1995a). Only one sample has a significantly higher ⁸⁷Sr/⁸⁶Sr ratio. Milner & le Roex (1996) recognized the overlap of Sr–Nd isotopic ratios between the Khumib compositions and Walvis Ridge DSDP Site 525A, as noted above.

The Tafelberg and Gramado magmas are very heterogeneous with regard to their Nd and Sr isotopic compositions, consistent with their variable trace element characteristics (Figs 10 and 11). The Tafelberg compositions scatter in a broad flat array extending from the field of Esmeralda samples towards crustal–sediment compositions. Three samples, however, are notably displaced to lower ¹⁴³Nd/¹⁴⁴Nd ratios (Fig. 12; crosses enclosed by circles). Very few mantle-derived magmas contain such unradiogenic Nd, but we note that our samples project towards the Gaussberg lamproite composition (calculated at 132 Ma from the isotopic composition of its inferred 2·5 Ga source; Murphy et al., 2002), and the Paraná low-Ti potassic mafic rocks. We attach particular significance to the fact that these three Tafelberg samples have distinctive unradiogenic Pb isotopic compositions (see below).

As noted by Milner & le Roex (1996), the Nd–Sr isotope fields of the Tafelkop and Tristan basalts overlap with Walvis Ridge basalt compositions from Sites 527 and 528. One Tafelkop composition extends to lower ¹⁴³Nd/¹⁴⁴Nd ratios. This sample again is also characterized by a relatively unradiogenic Pb isotope composition (see below). The Nd–Sr isotope compositions of the Horingbaai dykes are consistent with their E-MORB-like trace element chemistry. The isotopic compositions of the Namibian Cretaceous ultra-alkaline rocks overlap with those of the Tafelkop and Tristan samples, but not the Khumib–Urubici array. In contrast, the compositions of the Paraná ultra-alkaline igneous suites completely overlap all the Etendeka data that project within the mantle array.

Pb isotopes
New and previously published Pb isotope data for the Etendeka and Paraná are shown in Fig. 13 in relation to relevant reference points and evolution curves: the 132 Ma Geochron; the Erosion Mix (EM) growth curves of Kramers & Tolstikhin (1997), serving as a proxy for upper-crustal erosion material; the Depleted Mantle (DM) growth curve of Kramers & Tolstikhin (1997); and the Lower Mantle growth curve of Kamber & Collerson (1999). It should be noted that the present-day lower mantle (LM) curve of Kamber & Collerson (1999) is very similar to that of Galer & Goldstein (1996). Pb isotope compositions, except for the Gramado samples (Paraná), have been recalculated to 132 Ma, using, where necessary for some previously published data, extrapolated U, Th and/or Pb concentrations from similar rock compositions.

Excluding data for the carbonatites and ultra-alkaline rocks, the Paraná–Etendeka data define a broad array extending from left of the 132 Ma Geochron to near the t = 0 termination of the EM growth curve (Fig. 13a). Apart from the Tafelberg–Gramado rocks and the low-Ti Paraná basalts described by Marques et al. (1999), other samples plot almost entirely between the LM and DM growth curves. The Tafelberg, Gramado and the other Paraná low-Ti basalt compositions, with a few important exceptions (see below), lie between the LM and EM growth curves (also shown by ²⁰⁸Pb/²⁰⁴Pb data, Fig. 13d), consistent with their crust–sediment signatures indicated by the trace element and Nd–Sr isotope data. Added support for this is provided by the crustally contaminated Horingbaai and Tafelkop samples documented by Thompson et al. (2001), which lie within the main Tafelberg field (but are omitted from Fig. 13 for clarity). The ²⁰⁶Pb/²⁰⁴Pb ratios of low-Ti basalts from the southern Paraná (Marques et al., 1999) overlap the main Tafelberg array, whereas the compositions of their northern Paraná high-Ti and low-Ti basalts are less radiogenic and lie closer to the Khumib array.

Walvis Ridge samples plot between the Tristan and the Khumib–Urubici fields. Samples from DSDP Site 525A plot close to the Khumib–Urubici basalt field, but are separated by a clear compositional break from the samples of Sites 527 and 528 (Milner & le Roex, 1996).

The most intriguing aspect of the Pb isotope data is the relatively unradiogenic ²⁰⁶Pb/²⁰⁴Pb ratios of the Khumib–Urubici rocks, and small subsets of the Tafelkop, Horingbaai, Florida and northern Paraná high-Ti basaltic samples, which lie close to and to the left of the 132 Ma Geochron, yet clearly above the DM growth curve in terms of ²⁰⁷Pb/²⁰⁴Pb. The unradiogenic ²⁰⁶Pb/²⁰⁴Pb compositions projecting to the left of the Geochron cannot be approximated by known 132 Ma crustal, upper-mantle, or plume mantle reservoirs (Kramers & Tolstikhin, 1997), nor are they typical of OIB (Fig. 13a and d). The least radiogenic Khumib Pb isotope compositions overlap those of the 130 Ma Jacupiranga carbonatites from the eastern Paraná Basin (Toyoda et al., 1994; Huang et al., 1995), also seen in the ⁸⁷Sr/⁸⁶Sr vs ²⁰⁶Pb/²⁰⁴Pb isotope plot (Fig. 14a), but not in the corresponding ¹⁴³Nd/¹⁴⁴Nd vs ²⁰⁶Pb/²⁰⁴Pb plot (Fig. 14b). Significance is also attached to the small number of Tafelberg samples exhibiting unradiogenic Pb isotope compositions that lie close to, and left of the 132 Ma Geochron. Their possible interpretation is considered below.

The Namibian and Paraná ultra-alkaline suites and carbonatites (Fig. 13b) extensively overlap the compositional fields of the Etendeka volcanics, being predominantly constrained between the LM and DM growth curves. The trend defined by the field of the modern Tristan da Cunha compositions (Fig. 13c) suggests that it may contain a MORB-like admixed Pb isotope component. If correct, the original Pb isotope composition of the Tristan magmas should lie along the two-stage HIMU extension of the LM growth curve, following the arguments of Kamber & Collerson (1999). The calculated Tristan compositions at 132 Ma and 70 Ma (Walvis Ridge DSDP site age) are shown in Fig. 13c (calculation details in Fig. 13 caption; this assumes no change in the Tristan plume composition over 132 Ma). Relative to the ²⁰⁶Pb/²⁰⁴Pb ratios of the lavas of the Walvis Ridge DSDP sites 527 and 528, the inferred 70 Ma Tristan Pb isotope compositions are consistent with the MORB–Tristan mixing trend recognized by Humphris & Thompson (1983).

In ²⁰⁸Pb/²⁰⁴Pb vs ²⁰⁶Pb/²⁰⁴Pb space, spatial relations between the various Paraná–Etendeka rock suites are further complicated by the complexity of both Th/U ratios and time (Fig. 13d). A scattered Khumib (Urubici)–Tafelberg array extends from the LM to the EM and Old Upper Crust (OUC) growth curves (from Kramers & Tolstikhin, 1997). The Tafelkop and Horingbaai data are displaced to lower ²⁰⁸Pb/²⁰⁴Pb, defining two separate arrays extending from the DM and OUC growth curves. The Tafelkop array parallels that of the Khumib lavas but does not intersect it. If these represent mixing curves, the arrays trend towards the LM growth curves and their extensions would intersect this growth curve at Pb isotope compositions more unradiogenic than the Khumib–Urubici or Jacupiranga compositions.

The OIB field (Kamber & Collerson, 1999; outlined in Fig. 13a and d) shows that the unradiogenic Pb isotope compositions present in certain of the Etendeka mafic suites cannot be attributed to an unmodified OIB source. These unradiogenic Pb isotope compositions also plot away from compositions of crust-generated melts, as shown by the OUC growth curve and the compositions of the low-Ti silicic volcanic suites (inferred to be dominantly crustal in origin; see Part 2, Ewart et al., 2004).

The Pb isotope data of many suites of oceanic and continental basalts are more complex than the more tightly correlated Nd–Sr isotope and trace element ratios. Provided that the complexity can be interpreted, Pb isotopes afford higher temporal resolution for the reconstruction of magma source histories. An important aspect of Pb isotope data shown by many provinces is an apparent decoupling from other isotope systems. Two reasons are suggested. First, although individual OIBs define strongly correlated Pb isotope mixing arrays, the correlations with Sr and Nd isotope compositions break down on a global scale. This may be because relatively limited Rb/Sr and Sm/Nd fractionation in ascending mantle plumes cannot greatly influence the Sr and Nd isotope compositions during the plume's lifetime. However, over the >100 Myr lifetime of large-scale plumes, U/Pb fractionation, feasibly caused by CaSi-perovskite, may be capable of producing a spread of ²⁰⁶Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb ratios, leaving ²⁰⁷Pb/²⁰⁴Pb ratios unchanged (Kamber & Collerson, 1999). Second, the extent of depletion of the asthenosphere is much greater for Pb than for Sr, Nd, Hf or Os because the degree of enrichment in continental crust is overproportional for Pb (compared with Nd and Sr). This concentration contrast implies that the isotopic signature of depleted basalt sources is easily modified by very small contributions from enriched sources, making Pb isotopes ideal for the detection of crustal components (e.g. Hergt et al., 1989).

Sr–Nd–Pb isotope–trace element relationships
⁸⁷Sr/⁸⁶Sr vs ²⁰⁶Pb/²⁰⁴Pb (Fig. 14a) relations suggest a broad mantle array linking the Tristan–Walvis Ridge–Tafelkop–Khumib–Urubici and Jacupiranga carbonatite suites in which a systematic increase in radiogenic Sr correlates with decreasing ²⁰⁶Pb/²⁰⁴Pb, overlapping the EM1 field (Zindler & Hart, 1986). This trend is not consistent with crustal contamination. Tafelberg lavas possessing unradiogenic ²⁰⁶Pb/²⁰⁴Pb compositions have initial ⁸⁷Sr/⁸⁶Sr ratios that define a narrow range between 0·7076 and 0·7087, again atypical of crustal contamination. Crustal contamination is implied, however, in those Tafelberg rocks possessing more strongly radiogenic Sr and Pb isotope compositions.

¹⁴³Nd/¹⁴⁴Nd vs ²⁰⁶Pb/²⁰⁴Pb (Fig. 14b) indicates differing relations between suites: (1) the Jacupiranga carbonatites have initial ¹⁴³Nd/¹⁴⁴Nd ratios significantly higher than the Khumib–Urubici suite, and form a subhorizontal array containing the Walvis Ridge Sites 527 and 528, Tafelkop, Namibian ultra-alkaline and the Tristan samples, all of which lie above EM1; (2) the Khumib–Urubici and Walvis Ridge DSDP Site 525A suites define a separate field, in part including the EM1 domain. The Horingbaai compositions define a separate field, plotting at much higher ¹⁴³Nd/¹⁴⁴Nd values.

Sr and Nd isotopic compositions correlate with Ti/Zr, Th/Nb, Nd/Pb, Nb/U, La/Nb, Nb/Pb, Ce/Pb, Ti/Y, Th/Yb, Nb/Yb, Sr/Pb and Nb/Nb^* as shown, for example, in Fig. 15a and b. This plot illustrates a trend of decreasing Nb/Nb^* (i.e. higher sediment signatures) with increasingly radiogenic Sr, and decreasing ¹⁴³Nd/¹⁴⁴Nd ratios; a similar correlation occurs between decreasing Nb/Nb^* and increasing ²⁰⁶Pb/²⁰⁴Pb (not shown). The overlap between the Tristan and Tafelkop data is considered significant. The isotope–trace element correlations are consistent with the interpretation by Peate & Hawkesworth (1996) that the Esmeralda magmas developed by mixing between an asthenospheric melt (MORB-like) and a Gramado (Tafelberg)-type magma, followed by fractional crystallization.

	PETROGENESIS

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
Introduction
The major controversial question regarding Paraná–Etendeka petrogenesis concerns the relative contributions of plume, asthenosphere, SCLM and the continental crust to the origin of the mafic magmas (see references listed below). Nevertheless, a consensus has developed with regard to the role of the SCLM as the major source for a diverse range of high-Ti and low-Ti magmas, which may or may not include asthenospheric or plume contributions. Notable exceptions are the Horingbaai dykes with E-MORB-like character (considered asthenospheric by Erlank et al., 1984; Duncan et al., 1990; Thompson et al., 2001); the Tafelkop ferropicrites (considered as Tristan plume-related; Milner & le Roex, 1996; Ewart et al., 1998a; Gibson et al., 2000); and the low-Ti suites for which models involving AFC processes in the crust have been proposed (e.g. Erlank et al., 1984; Bellieni et al., 1986; Peate & Hawkesworth, 1996; Ewart et al., 1998a).

The attraction of the SCLM as a primary magma source in the Paraná–Etendeka Province and other continental flood basalt provinces, such as the Karoo, is its potential to act as an ancient, stagnant and major element depleted mantle reservoir, able to accumulate ‘evolved’ isotopic signatures over time, periodically enhanced by repeated episodes of metasomatic enrichment, carbonation, halogenation and hydration (e.g. Hawkesworth et al., 1983, 1984, 1992, 1999; Erlank et al., 1984; Piccirillo et al., 1989; Ellam & Cox, 1991; Hergt et al., 1991; Gallagher & Hawkesworth, 1992; Turner & Hawkesworth, 1995; Peate, 1997; Marques et al., 1999). The SCLM is accordingly frequently referred to as heterogeneous (particularly in terms of trace element and isotope chemistry) and variably metasomatized. As the diverse Paraná–Etendeka mafic magmas, and those of many other continental flood basalt provinces, differ chemically from OIB and MORB, it is inferred that they originated from a distinct magma source not present within the oceanic mantle (e.g. Turner & Hawkesworth, 1995; Peate, 1997). This view, however, may require modification in the light of results from more recent studies on oceanic plateaux, especially Ontong Java (Neal et al., 1997; Collerson et al., 2000).

Evidence for the existence of metasomatically enriched SCLM was initially recognized in kimberlite xenoliths. Hawkesworth et al. (1983, 1984) and Erlank et al. (1984) provided detailed discussions of the inferred relevance of such xenoliths to the petrogenesis of the Etendeka and Karoo flood basalts, including the various styles of enrichment common to both xenoliths and volcanic rocks [see review by Menzies (1992)]. Erlank et al. (1984) recognized the broadly subduction-related geochemical signatures of the low-Ti mafic rocks and discussed in detail the different possible metasomatic enrichment processes responsible. These included metasomatism by fluids emanating continuously from the deep mantle, veining by small-volume partial melts, and metasomatism by fluids generated in ancient subduction environments. They concluded that there was no compelling evidence for the last alternative. In contrast, Hawkesworth et al. (1984), dealing with a wider dataset, including representative Karoo basalts, advocated an introduced Proterozoic subduction component into the southern African SCLM. Some researchers, however, have questioned the reality of large volumes of mafic magma being derived from cool, refractory SCLM (e.g. McKenzie & Bickle, 1988; Campbell & Griffiths, 1990; Arndt & Christensen, 1992; Menzies, 1992; Arndt et al., 1993). Some of these objections were addressed by Gallagher & Hawkesworth (1992) in terms of melting of amphibole-bearing SCLM in the presence of H₂O + CO₂.

In the Cretaceous–Jurassic igneous suites of southern Africa and South America, the SCLM has been credited as the dominant, or even only source for a remarkable range of magma compositions. An abbreviated listing includes:

1. the regional low-Ti/Y Paraná–Etendeka tholeiitic series (Erlank et al., 1984; Piccirillo et al., 1988b, 1989; Peate, 1997; Hawkesworth et al., 1999), including the Paraná Gramado series (modified by crustal contamination; Peate & Hawkesworth, 1996), the northern Paraná tholeiites (Marques et al., 1999), the Jurassic regional Karoo tholeiites (Hawkesworth et al., 1984), and the regional Jurassic and Cretaceous low-Ti flood basalts of the wider Gondwanaland provinces (Hergt et al., 1991; compare Hergt et al., 1989);
   
2. the high-Ti/Y Paraná Urubici magmas (Peate et al., 1999);
   
3. alkaline and ultramafic lamprophyres and carbonatites in certain alkaline Namibian Damaraland central complexes (le Roex & Lanyon, 1998);
   
4. Brazilian Jacupiranga carbonatites, which were interpreted as being derived from ‘similar mantle source regions’ to the high-Ti Paraná and Walvis Ridge basalts (Huang et al., 1995), although we note that Toyoda et al. (1994) preferred an ‘asthenospheric mantle plume’ source for carbonatites of southeastern Brazil;
   
5. the potassic silica-undersaturated and -saturated mafic–intermediate dykes of eastern Paraguay (Comin-Chiaramonti et al., 1992);
   
6. magnesian potassic–ultrapotassic and associated carbonatitic extrusives and intrusives emplaced extensively along the Paraná Basin margins during late and early Cretaceous (Gibson et al., 1995a, 1995b; Carlson et al., 1996). In the case of the Serra do Mar complexes, Thompson et al. (1998) proposed a sub-lithospheric convecting mantle source with ‘sporadic input from lithospheric metasomites’.
   

The above listing highlights the extreme range of magma compositions, and their wide range of isotopic compositions, advocated as originating in the SCLM. Also of note is the wide range in scale of magma generation that is believed to have taken place in the SCLM, from smaller volume carbonatites to flood basalt sequences exceeding 10⁶ km³ in volume, all within the Jurassic–Late Cretaceous time period. We seriously question whether constraints exist to justify the ability of SCLM to be compositionally diverse enough, in sufficient volumes, to act as the source of all the magmas in question.

The data presented in this paper add nothing new to arguments supporting the major role of the SCLM in Paraná–Etendeka magmatism. Instead, we offer an alternative model for debate, one that explores a sub-lithospheric origin for the mafic magmas. This specifically involves interaction between the Tristan plume, convecting asthenosphere, and a component derived from subducted material stored in (or above) the Transition Zone. Interpretations are, however, potentially complicated by variable degrees of SCLM and crustal interaction that are likely to have occurred.

Khumib basalts
General considerations
There are five relevant aspects to the petrogenesis of the Khumib lavas:

1. they are interbedded with low-Ti Tafelberg lavas throughout the northern Etendeka.
   
2. Most samples have been modified by fractional crystallization (e.g. wt % MgO <5·5%, excepting the basal MgO-rich flows). Isotopic evidence for crustal assimilation is minimal, with a single exception (Figs 12–15).
   
3. Khumib trace element and Sr–Nd isotope compositions consistently lie within, or extend from the mantle arrays defined by MORB–OIB.
   
4. The Khumib compositions show some trace element and Sr–Nd isotopic similarities with the Tafelkop and Tristan da Cunha basalts (e.g. Figs 10, 12 and 16b).
   
5. They have distinctive Pb isotope compositions with ²⁰⁶Pb/²⁰⁴Pb ratios that plot mainly to the left of the 132 Ma Geochron, and relatively high ²⁰⁷Pb/²⁰⁴Pb ratios. In ²⁰⁶Pb/²⁰⁴Pb–²⁰⁷Pb/²⁰⁴Pb–²⁰⁸Pb/²⁰⁴Pb plots, the Khumib compositions project above the DM and close to the LM growth curves (Fig. 13a and d). Khumib basalts clearly have Pb isotopic compositions that are atypical of magmas derived from asthenosphere, plume or OIB sources. Comparable compositions have been reported in lamproites (Murphy et al., 2002), Pitcairn lavas (Eisele et al., 2002), some carbonatites (e.g. Jacupiranga, Toyoda et al., 1994) and related ultra-alkaline suites from southern Brazil (Carlson et al., 1996), minettes from Schirmacher Oasis in East Antarctica (Hoch et al., 2001), in certain Archaean lower-crustal granulites (e.g. Rudnick & Goldstein, 1990), and in the oceanic environments of the Walvis Ridge and the southern Mid-Atlantic Ridge (see below).
   

View larger version (38K): [in this window] [in a new window]   Fig. 16. Multi-element plots of the northern Etendeka Khumib and Tafelberg samples (as in Figs 7 and 8) normalized to GLOSS and the Tristan da Cunha sample 11898 (Appendix, Table A1). In (a) the Pb, U, K, Th and Cs (Khumib only) depletions should be noted, whereas in (b) Pb, Ba, K and Cs (Tafelberg only) spikes are apparent, with marked Nb ± Ta depletions.

 
U/Pb and Th/Pb ratios in Khumib rocks (Table 3) are lower than observed in Tristan da Cunha and the Cretaceous alkaline–ultra-alkaline igneous phases of the Paraná and Etendeka, but overlap those of the tholeiitic Horingbaai, Tafelkop and Tafelberg suites. These ratios do not significantly correlate with Pb isotope composition, i.e. decoupling is apparent between Th, U, Pb and Pb isotopes.

The enhanced ²⁰⁷Pb/²⁰⁴Pb but low ²⁰⁶Pb/²⁰⁴Pb ratios imply a two-stage evolution (over extended periods of time), first in a high U/Pb environment followed by a low U/Pb stage. This is traditionally thought to occur in the SCLM [see review by Murphy et al. (2003)], where high ²⁰⁷Pb/²⁰⁴Pb is interpreted to be derived by fluids from subducted sediments followed by a long, low U/Pb evolution in the stable SCLM (see, e.g. Hergt et al., 1991; Hoch et al., 2001). Another possible source for unradiogenic Pb is ancient lower crust (see Zartman & Haines, 1988; Kramers & Tolstikhin, 1997; Thompson et al., 2001). Compilations of the Pb isotope compositions of lower-crustal xenoliths show that samples with Pb isotope compositions less radiogenic than the Geochron characterize certain Archaean xenoliths that have suffered metamorphic U loss. The majority of xenoliths, however, have Pb isotope compositions more radiogenic than the Geochron (Murphy et al., 2003).

With regard to the South Atlantic, the existence of relatively unradiogenic Pb isotope compositions in the Walvis Ridge Site 525A oceanic basalts is very important. These compositions overlap those of the Khumib rocks, but the lack of Archaean lower crust or an ancient SCLM keel beneath the Walvis Ridge clearly argues for a sub-lithospheric origin of the unradiogenic Pb. Those advocating an SCLM source to account for these distinctive Pb isotope compositions have proposed that delamination of the continental lithosphere during continental rifting has allowed lithospheric contamination of the source of some Walvis Ridge basalts, specifically those from Site 525A (Milner & le Roex, 1996). We suggest this to be unlikely. Further insight on their origin can be gained from the occurrence of the unradiogenic ²⁰⁶Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb compositions of the very localized LOMU basalts associated with the southern Mid-Atlantic Ridge (Douglass et al., 1999; Kamenetsky et al., 2001). These are characterized by unusually low ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁸Pb/²⁰⁴Pb and ¹⁴³Nd/¹⁴⁴Nd ratios, variable ²⁰⁷Pb/²⁰⁴Pb, relatively high ⁸⁷Sr/⁸⁶Sr ratios, and are associated with mantle plumes, specifically the off-ridge Discovery and the ridge-centred Shona mantle plumes (Douglass et al., 1999), and also near the Bouvet Triple Junction (Kamenetsky et al., 2001). In keeping with the conventional models, their occurrences have been attributed to SCLM and lower-crustal delamination during Gondwana breakup by these researchers. We suggest that their widespread and localized occurrences, especially near the Mid-Atlantic Ridge and an associated triple junction, almost certainly require a sub-lithospheric reservoir, sampled occasionally by deep mantle upwellings. Unradiogenic oceanic Pb is not restricted to the Atlantic but has also been reported, for example, from the Indian Ocean (e.g. Mahoney et al., 1996).

When continental Pb isotope signatures were initially recognized in apparently uncontaminated basalts (e.g. Hergt et al., 1989, 1991), higher resolution tomographic images of the mantle were not available and the presence of subducted material in the deep mantle was not widely accepted (but see Ringwood, 1991). The SCLM was therefore identified as the logical source of such Pb, and transport of continental Pb into the lithospheric mantle was inferred to occur via fluid migration from subducted slabs with a sedimentary veneer (e.g. Hoch et al., 2001). With modern seismic tomography, however, it is now recognized that subduction and storage of slabs in the deep mantle, especially the Transition Zone, does occur (e.g. van der Hilst et al., 1991; van der Hilst & Seno, 1993; Christensen, 1996; Simons et al., 1999; Fukao et al., 2001).

In response to this changing chemical geodynamic picture of the mantle, Murphy et al. (2002) proposed a model in which comparable Pb isotopic compositions, sampled by the Gaussberg lamproites, were stored in the deep mantle rather than the lithosphere. Specifically for the Gaussberg lavas, Murphy et al. (2002) modelled the isotope data in terms of subducted sediment stored in the Transition Zone for 2–3 Gyr. Their proposed two-stage evolution model therefore requires: (1) that enhanced ²⁰⁷Pb/²⁰⁴Pb ratios be inherited from an ancient crustal source; (2) subsequent evolution of this crustal source, with reduced µ, slowing isotope evolution. The reduced µ is attributed to metamorphism accompanying subduction (U–Pb partitioning by, for example, K-hollandite; Irifune et al., 1994). In an extension of this model, Murphy et al. (2003) considered the wider case of melting of a subducted slab with both sediment and MORB (eclogite) to explain the range of Pb isotopic compositions of a wider spectrum of alkaline magmas. Younger sediment components result in more radiogenic ²⁰⁶Pb/²⁰⁴Pb compositions in their two-stage model, and variable U/Pb fractionation during subduction is an additional parameter. Eisele et al. (2002; compare Gasperini et al., 2000) have also proposed a minor recycled 0·7–1·9 Ga continental crustal component in the source of the oceanic Pitcairn lavas, mixed with a depleted mantle component, to explain the observed unradiogenic ²⁰⁶Pb/²⁰⁴Pb and high ²⁰⁸Pb/²⁰⁴Pb ratios. The existence of Walvis Ridge DSDP Site 525A oceanic basalts (70 Ma) confirms that Walvis Ridge–Etendeka magmas had been able to access an unradiogenic Pb isotope source for at least 60 Myr. This period is significantly extended to >100 Myr by the above-noted LOMU occurrences near the southern Mid-Atlantic Ridge.

Trace elements
N-MORB-normalized multi-element plots (Fig. 8) exhibit enhanced concentrations of Pb and Ba, but depleted abundances of Sr, Nb and Ta. When normalized to GLOSS (Fig. 16a), enhanced abundances of Nb, Ta, ±Sr, ±Zr and Ti are evident, with depleted abundances of Cs, U, Pb ± Th. The same compositions normalized to the Tristan da Cunha sample 11 898 (Appendix, Table A1; Fig. 16b) indicate Ba and Pb peaks, marked Ta and Ti depletions, and small depletions of Nb and Cs. We regard these enrichment or depletion patterns as supportive of an actual metasedimentary source (albeit modified by subduction metamorphism) rather than selective fertilization of a lithospheric source in certain elements (e.g. Pb).

Relative to GLOSS, observed depletion of especially Cs, U and Pb is suggested to be consistent with a subduction signature, as these elements are known to be lost during slab dehydration (e.g. Brenan et al., 1994, 1995a, 1995b; Becker et al., 2000). They are not consistent with simple continental crust contamination. The Nb–Ta depletion relative to N-MORB contrasts with Nb–Ta enrichment relative to GLOSS. The Nb/Ta ratios (18·2–23·4) of the two Khumib samples analysed by ICP-MS (Tables 1 and 3) are relatively high (consistent with the high Nb/Ta ratios of the associated high-Ti silicic volcanic units; see Part 2, Ewart et al., 2004). The reported Nb/Ta ratios of the Urubici samples (Peate et al., 1999) are significantly lower than the Khumib data, but were analysed by a different technique. High Nb/Ta ratios may develop in the slab during slab dehydration (e.g. Kamber & Collerson, 2000a; Tang & Liu, 2002). During slab subduction and accompanying dehydration and metamorphism of sediment + MORB oceanic crust, concentrations of LILE, HFSE and LREE will be dominated by the sediment component relative to MORB. This will be most critical for Pb, but less so for Sr (important in resulting isotopic compositions), as verified by simple mixing calculations using, for example, the altered MORB data of Becker et al. (2000).

The presence of a sedimentary component in the Khumib basalts is consistent with various trace element parameters (e.g. Nb/Nb^*, Ce/Pb, Sr/Pb, Th/Nb) and correlated with Sr and Nd isotope compositions (e.g. Figs 10, 11 and 15; Table 3). As previously noted, however, the Khumib lavas also exhibit certain geochemical similarities to the Tristan da Cunha lavas, whose possible significance is discussed below.

The above interpretations imply sub-lithospheric, high-pressure phase equilibria controls on magma compositions. One indication of this is the behaviour of La, Sr and Yb (Fig. 17a). Sr abundances are sensitive to lower-pressure fractionation of plagioclase-bearing (gabbroic) mineral assemblages, whereas Yb should be buffered by eclogite and garnetite (majorite) assemblages. Complications can arise, however, from potential Sr partitioning into K-hollandite and Ca-perovskite (Ringwood et al., 1992; Irifune et al., 1994; Stachel et al., 2000; Taura et al., 2001). Considering the whole range of magma compositions represented in the Paraná–Etendeka Province, including the ultra-alkaline types, Fig. 17a does suggest the existence of two trends. One is the steeper trend towards higher La/Sr ratios defined by the low-Ti Tafelberg–Esmeralda basalts, the high-Ti and low-Ti silicic rocks (see Ewart et al., 2004), and a small number of more strongly fractionated compositions (i.e. lowest wt % MgO concentrations) within the Khumib and Urubici lavas. This is suggested to reflect lower-pressure plagioclase-controlled fractionation, with or without crustal assimilation. The second trend is a shallow trend extending to high La/Yb ratios, defined by the broad array extending through the Horingbaai, Tafelkop, Khumib and Tristan basalts, continuing to the Namibian and Paraná lamproitic-kimberlite and the related alkaline rocks. We interpret this shallow trend as reflecting sub-lithospheric or deep lithospheric, higher-pressure phase equilibria in the presence of garnet (majorite). A lower-mantle and sub-lithospheric origin has been proposed for OIB and lamproitic-kimberlite magmas, respectively (e.g. Ringwood et al., 1992; Kamber & Collerson, 1999, 2000a, 2000b, and references therein; Collerson et al., 2000; Stachel et al., 2000; Murphy et al., 2002). If these views are correct, then the observed shallow La/Yb trend, leading to high La/Yb ratios, is consistent with sub-lithospheric phase control.

Tafelberg basalts
In this section we consider the combined data for the Tafelberg basalts of both the southern and northern Etendeka. The following aspects are considered petrogenetically significant.

1. The Tafelberg suite is compositionally very variable as illustrated by its range in trace element abundances and isotopic compositions (e.g. Figs 10–13). There is general agreement that low-pressure fractional crystallization, with or without crustal contamination, in which plagioclase is a dominant fractionating phase (e.g. Fig. 11b), is an important process in the origin of the Tafelberg basalts (Erlank et al., 1984; Peate & Hawkesworth, 1996; Ewart et al., 1998a).
   
2. Erlank et al. (1984) showed that the full range of initial ⁸⁷Sr/⁸⁶Sr ratios occurs in both the most primitive and evolved Tafelberg basalts (see also Fig. 9), i.e. the parental magmas had very variable Sr isotopic compositions. Peate & Hawkesworth (1996) reported comparable features for the Gramado series, and identified groups of samples showing little or no evidence for crustal contamination. These have initial ⁸⁷Sr/⁸⁶Sr ratios of 0·710 (Caxias do Sul) and 0·708 (São Joaquim), interpreted in terms of two-stage development of the Gramado magmas, the initial stage attributed to a lithospheric, early crustal contamination phase.
   
3. In trace element and Sr–Nd isotope plots, the Tafelberg data lie away from the main mantle arrays (e.g. Figs 10 and 12), extending from near the mantle array into crustal compositional fields.
   
4. Although the majority of Tafelberg rocks have relatively radiogenic Pb and Sr and unradiogenic Nd isotope compositions, four samples exhibit unradiogenic ²⁰⁶Pb/²⁰⁴Pb ratios (Table 1) and lie close to, or to the left of the 132 Ma Geochron in Pb isotope plots (Figs 13 and 14).
   

Marques et al. (1999) have reported similar, but not as strongly unradiogenic ²⁰⁶Pb/²⁰⁴Pb compositions for low-Ti basalts from northern Paraná (Fig. 13a). The unradiogenic ²⁰⁶Pb/²⁰⁴Pb Tafelberg samples, nevertheless, have Sr and Nd isotopic compositions that overlap those of the main Tafelberg array. We suggest that these samples may preserve geochemical signatures predating any Tafelberg magma interaction with continental crust (±SCLM). This possibility receives support from the trace element and isotope compositions of the unradiogenic Pb isotopic samples, which, except for the evolved sample (KLS662), lie displaced from the crustal compositions (e.g. Figs 10, 12 and 14). Their initial ⁸⁷Sr/⁸⁶Sr ratios lie between 0·7076 and 0·7078 (0·7087 for the evolved sample), similar to Sr isotope compositions of the ‘primitive’ Gramado rocks of the Paraná [see (2) above].

Tafelberg multi-element plots (Figs 8 and 16) show similar, but more strongly developed trace element depletion–enrichment patterns to the Khumib rocks. Relative to GLOSS (Fig. 16a), Nb, Ta, ±Sr, ±Zr, Ti are enriched and U and Pb depleted, with Cs behaviour variable between samples. Tafelberg basalts differ considerably from Tristan lavas in their incompatible element abundances, more so than the Khumib samples (Figs. 16b), and this is consistent with other trace element and isotope data (Figs 10, 11, 12 and 15).

The majority of the Tafelberg suite samples possess geochemical signatures indicative of a strong crustal–sediment component in their petrogenesis as illustrated by Nb/Nb^* ratios (Table 3; Figs 11 and 15). Differences in trace element ratios are apparent between the three most primitive (i.e. unradiogenic ²⁰⁶Pb/²⁰⁴Pb ratios) Tafelberg samples and the main Tafelberg dataset, specifically Ti/Y, U/Pb, Th/Pb, Nb/U and Zr/Nb (Table 3), reflecting the distinctive Pb, U and Nb abundances within these rocks. To explain the origin of the unradiogenic Pb isotope compositions in samples that have initial ⁸⁷Sr/⁸⁶Sr ratios of 0·708, we suggest that these magmas have inherited their crustal signatures, not through high-level crustal contamination, but by incorporating an ancient recycled, subducted sediment–MORB component from the subducted oceanic lithosphere (compare Hergt et al., 1989). If correct, we suggest that this occurred during upwelling of the Tristan plume. The Tafelberg magmas represent the most voluminous volcanic phase within the southern Etendeka region, presumably maximizing the potential for mixing between sub-lithospheric mantle domains. It is nevertheless clear that extensive overprinting by crustal contamination has occurred.

Distinguishing crustal contamination from deeper mantle mixing or assimilation of recycled subducted sediment is not straightforward. Pb isotope data are critical in this regard but the GLOSS-normalized compositions (Fig. 16) suggest that Cs, U, Sr and Pb may offer the potential to distinguish between the two processes. Fractionation between Sr, Cs, U and Pb differs between mantle, crust and subduction melting–crystallization processes (e.g. Brenan et al., 1994, 1995a, 1995b; Irifune et al., 1994; Stachel et al., 2000; Taura et al., 2001). Sr/Cs vs Pb/Cs ratios (Fig. 17b, noting the paucity of Cs data and the possible post-eruptive mobility of Cs) suggest two trends defining mantle and crustal arrays. The former is defined by the Paraná ultra-alkaline rocks, Tristan, OIB and MORB compositions, the high-Ti lavas from northern Paraná (Marques et al., 1999) and the Khumib–Urubici suite. This is designated as a sub-lithospheric source trend. The second trend, inferred to define a crustal fractionation and crustal contamination trend, contains most of the Tafelberg, Esmeralda, southern Paraná low-Ti samples and the silicic eruptive units (see Ewart et al., 2004). The Tafelberg samples with unradiogenic Pb isotope compositions scatter across the two trends. We suggest that the Cs, Pb and Sr geochemistry of the Tafelberg, and the Gramado and low-Ti basalts of the Paraná (Marques et al., 1999) reflects their complex history, including inherited sub-lithospheric source characteristics and the superimposed signatures of crustal contamination.

A model for Paraná–Etendeka magmatism
Plume ascent driven by thermal buoyancy will entrain surrounding mantle with relatively uncontaminated material preserved in the plume axis (e.g. Campbell & Griffiths, 1990, 1992; Griffiths & Campbell, 1990, 1991). Following Campbell & Griffiths, we envisage that: (1) the magnitude of the Paraná–Etendeka volcanism, together with tomographic evidence (Romanowicz & Gung, 2002), indicates that the Tristan plume probably emanates from the lower mantle; (2) entrainment of ambient mantle material during plume ascent and cooling results in compositional zoning within the plume head, which increases in volume with progressive entrainment, accompanied by stirring (but not mixing) of plume head components; (3) continued ascent of the plume initiates progressive adiabatic melting as the various entrained materials pass through their respective solidi, allowing varying degrees of mixing of the partial melts; (4) the plume head mushrooms as it impacts the base of the lithosphere, where it may induce thermal erosion.

If the Paraná–Etendeka mafic magmas are fundamentally plume-derived, then their different compositions could reflect mixing, in varying proportions, of partial melts of the different mantle components within the plume head. Plausible mantle components include the Tristan plume (lower mantle), convecting and depleted upper mantle (MORB-source), stored subducted slab components in the Transition Zone, and the SCLM. Thus, not only is the initial Tristan plume the thermal source, but we believe it contributes materially as well (White et al., 1987; White & McKenzie, 1989). This is contrary to the views of Turner et al. (1996) and Peate (1997), who regard the Tristan plume as having played a largely passive role, providing only heat for lithospheric mantle melting.

Subducted slabs stored within the Transition Zone are expected to exist as discontinuous lensoid masses within the Transition Zone space, representing a compositionally variable mantle reservoir (Ringwood, 1991; Collerson et al., 2000). Geological observations suggest phases of subduction beneath the Namibian and southeastern Brazilian crust during the Early and Late Proterozoic and Archaean (e.g. Miller, 1983; Becker et al., 1996; Machado et al., 1996; Teixeira et al., 1996; Seth et al., 1998). Majorite, K-hollandite, CaSiO₃ perovskite, a CAS phase, and stishovite (Irifune et al., 1994), plus MORB eclogitic assemblages, are expected to be present in subducted sediment and mixed sediment–MORB phase assemblages. We suggest that entrainment of pre-existing slab components occurred in the Transition Zone, during plume ascent, with additional incorporation of shallow depleted upper mantle occurring during continued ascent. The plume head may therefore have evolved into an incompletely mixed, three-component system prior to impacting the base of the SCLM. Additional compositional heterogeneities may arise when melts derived from this three-component system pass through the SCLM and the continental crust. The Pb isotope compositions of the modern Tristan da Cunha lavas are consistent with the presence of a depleted upper-mantle component in the plume (see above), whereas plume–MORB source mixing has been proposed to explain the compositions of the Walvis Ridge Sites 527 and 528 basalts (Humphris & Thompson, 1983).

The possible importance of the convecting asthenosphere as a major source of continental flood basalts has been emphasized by previous workers, for example, Ellam & Cox (1991) and Gibson et al. (1995a). A characteristic aspect of the Paraná–Etendeka Province, and indeed of the wider Gondwanaland Jurassic–Cretaceous provinces, is the existence, distribution and significance of coexisting ‘high-Ti’ and ‘low-Ti’ magma types, which occur over wide areas. As noted by Turner et al. (1994), chemically defined magma types seem to have been erupted at different times in different places. Their regional occurrences have been referred to as ‘geochemical provinces’, and attributed to large-scale SCLM heterogeneities (e.g. Duncan, 1987; Cox, 1988; Piccirillo et al., 1989; Gibson et al., 1995a; Peate & Hawkesworth, 1996).

To gain insight into the petrogenetic processes operating within the plume model outlined above, we present a series of mixing scenarios involving melts representing the following assumed end-members: (1) N-MORB (data after Sun et al., 1979; Sun & McDonough, 1989); (2) modern Tristan da Cunha (plume) compositions, using only lavas with >6·0 wt % MgO (data sources as listed above); (3) melts derived from recycled subducted sediment–MORB component, specifically based on the Gaussberg lamproites (Murphy et al., 2002); (4) pre-subduction sediments based on GLOSS (Plank & Langmuir, 1998), also representing a composite upper-crustal composition. Although mixing of the different source materials within the expanding plume may be possible, we argue for the presence and mixing of the derived, but compositionally different melts. We suggest this receives support from melt inclusion studies (e.g. Saal et al., 1998). Therefore, the mixing calculations use observed magma compositions for the proposed mantle end-members and an average sediment proxy for the crustal contaminant.

With regard to the plume component, we note the arguments of Gibson et al. (1995a) for using a magnesian ‘oceanic plume magma’, which they considered to be more representative of plume head magmas (e.g. the relatively localized southern Etendeka Tafelkop type). In view of the scale of magmatism, and the sub-lithospheric depths at which we argue magma mixing had occurred, we prefer to use the modern Tristan compositions.

The melt derived from the proposed recycled subducted sediment–MORB component is modelled on the relatively homogeneous Gaussberg lamproite lavas. These have been interpreted as melts derived from an ancient (2–3 Ga) sediment-dominated subducted slab component, with partial re-equilibration with enclosed mantle, stored in the Transition Zone (Murphy et al., 2002, 2003; see also Eisele et al., 2002). This represents a very distinctive composition not observed in the spectrum of continental sediments. The reasons for its choice are: (1) lamproites represent magmas derived from sources which, with one exception, exhibit unradiogenic ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁸Pb/²⁰⁴Pb, and usually enhanced ²⁰⁷Pb/²⁰⁴Pb ratios, which plot to the left of the Geochron (Murphy et al., 2002, 2003); (2) the Gaussberg lamproites are fresh and relatively young lavas that are chemically relatively homogeneous, exhibit minimal evidence for crustal contamination, and for which new and comprehensive major and trace element and isotopic data are available; (3) the Gaussberg lavas exhibit very similar Nb, Ta, Cs, Ba, U and Pb abundance patterns, relative to N-MORB and GLOSS, as seen (to variable degrees) in the Etendeka high-Ti and most low-Ti basaltic types, which are attributed to subduction metamorphism; (4) plausible histories have been modelled for the isotopic evolution of the Gaussberg source (Murphy et al., 2002).

We use GLOSS (Plank & Langmuir, 1998) as a proxy for evaluating the effects of crustal contamination. A global sediment average was preferred over a modelled local crustal average because GLOSS is an actual observed composition based on a large number of comprehensive datasets. The Etendeka crustal column is expected to exhibit significant regional, lithological and temporal variation, for which few geochemical data exist, and some of which may extend outside the modelled chemical space.

Mixing calculations, based on the four end-member compositions defined above, are illustrated in Fig. 18a–f, based on selected element ratio and element-isotope ratio combinations. General observations arising from the mixing calculations (which are not confined to the combinations of parameters plotted in Fig. 18), include the following.

1. The disparity in LREE, LILE and HFSE abundances between end-members is such that the compositions of N-MORB melts are very sensitive to even very small (1–5 wt %) inputs of the other end-member components. This is illustrated by ‘sediment signature’ parameters such as Nb/Nb^*, Nd/Pb and La/Nb (Fig. 18c–e).
   
2. Trace element and Nd–Sr isotope compositions suggest that the various Paraná–Etendeka magma types considered in this paper (Tafelberg–Gramado, Khumib–Urubici, Tafelkop, Horingbaai, Florida) can be plausibly modelled, to a first approximation, in terms of the four components shown (Fig. 18). In fact, the mixing models provide an explanation for their trace element and isotopic characteristics without the necessity to invoke SCLM.
   
3. The mixing calculations suggest a possible ‘magma-genic’ explanation for the existence of high-Ti and low-Ti magma types. Magmas with relatively high proportions of the Tristan plume component, and consequently smaller input of N-MORB, have inherently higher Ti/Y ratios. The reverse is true in low-Ti magma types. Such an explanation seems to receive support from the wide range of Ti/Y ratios within the Paraná–Etendeka Province when the Pitanga, Paranapanema and Ribeira magma types of the northern Paraná are taken into account (Peate, 1997; Fig. 18d).
   
4. The Gaussberg rocks have Nd–Sr isotope compositions at 132 Ma that are appropriate as a potential sediment end-member for the Khumib–Urubici magmas (Fig. 18a). This does not strictly apply to the Pb isotope data, where the Khumib–Urubici samples extend to less radiogenic ²⁰⁶Pb/²⁰⁴Pb and ²⁰⁷Pb/²⁰⁴Pb compositions than those observed at Gaussberg (e.g. Fig. 18f). Applying the same calculation procedures as outlined by Murphy et al. (2002; fig. 5) to those Khumib lavas with least radiogenic ²⁰⁶Pb/²⁰⁴Pb compositions, an ‘age’ of 3·7–3·8 Ga is obtained for the inferred subducted ‘Erosion Mix’ source. This is calculated to satisfy both ²⁰⁶Pb/²⁰⁴Pb and ²⁰⁷Pb/²⁰⁴Pb compositional constraints. A similar calculation applied to sample KLS048 (Table 1), the Tafelberg sample exhibiting the least radiogenic ²⁰⁶Pb/²⁰⁴Pb ratio, gives an ‘age’ of 2·8 Ga. Although suggesting an Archaean source component for the Pb in these Khumib and Tafelberg magmas, the inferred presence of other magma components in the magmas (see below) makes the significance of the calculated ‘ages’ uncertain, although these are consistent with an ancient recycled sediment source component in these magmas.
   

	SUMMARY AND CONCLUSIONS

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
We present a new interpretation for the diversity of magma chemistries and isotope compositions of mafic lavas in the Etendeka. The new model operates with three sub-lithospheric magma types (N-MORB, Tristan OIB and lamproite) and a crustal contaminant. Given the complexity of magma chemistries, it is not our intention to attempt to provide a unique model. Rather, we illustrate, to a first approximation, the feasibility of matching observed trace element and isotope characteristics of the various basaltic magma types without a lithospheric magma source. The marked disparity in element concentration ranges in the assumed end-members produces differing geometries of the fields defined both by the mixing curves and by the projected compositional fields of each basalt type, often with only small changes in degrees of mixing. Summarizing the above data and mixing models, we consider more specifically each magma type below and suggest very approximate proportions at which the end-member components may be present. This is based only on the trace element and isotope data, and does not consider bulk compositional balances. The variations in the estimated proportions of the various end-member components reflect the inherent variation within each of the basalt types in respect to the element parameters plotted.

Khumib–Urubici type. We interpret the data to indicate that this magma type comprises a predominant plume component (45–75%) with subordinate N-MORB (5–10%) and recycled subducted sediment (+MORB) components (20–40%), the sediment component having compositional attributes similar to the Gaussberg lamproites (e.g. Fig. 18a, b and d). The dominance of the inferred plume component appears to be consistent with the distribution of the outcrop of the Urubici–Khumib lavas, which are effectively bisected by the zone of incipient splitting along the proto-Atlantic rift, according to the pre-drift reconstruction of Peate et al. (1999, fig. 2). This suggests that these magmas developed above the axis of the plume head where the Tristan magma composition was best represented. The inferred presence of a major plume component provides one possible answer to the enigma (see Introduction) that there had been no plume-related magma previously recognized within the northern Etendeka, in spite of the intersection of the Walvis Ridge plume trace with the northern Etendeka coast.

Tafelberg–Gramado type. This strongly heterogeneous, low-Ti magma type is less enriched in a range of incompatible elements compared with the Khumib magmas, and also possesses a strong crustal–sediment trace element signature. This suggests significant N-MORB input (5–55%), with subordinate subducted sediment (2–20%) and plume (2–25%) components in the magmas. In the Etendeka, the Tafelberg basalts dominate away from the Walvis Ridge, and therefore feasibly emanated from the more peripheral parts of the plume head where the Tristan component was lessened by the increased entrainment of asthenospheric mantle and sediment components. The inferred sediment component is clearly biased towards the GLOSS-type composition (20–80%) and plausibly reflects overprinting by crustal contamination of many of these magma compositions. As discussed above, the presence of a subordinate Gaussberg-type component is suggested by unradiogenic Pb isotope compositions preserved within a small number of samples.

Esmeralda type. Peate & Hawkesworth (1996) concluded that these magmas represent mixing between asthenospheric (MORB) melt and Gramado (Tafelberg) type magma at relatively shallow (crustal) depths, with superimposed fractional crystallization. Our data are consistent in showing the overlap between the Tafelberg and Esmeralda magma types and appear generally consistent with the above model. The mixing plots suggest N-MORB input of 60–90%, GLOSS-type sediment input of 10–30%, and 10% input of Gaussberg or Tristan components.

Horingbaai type. Data discussed here are consistent with the strong E-MORB-like characteristics of these magmas (Duncan et al., 1990; Thompson et al., 2001), with a source within the convecting asthenosphere. The N-MORB component is suggested to be in the range 85 to >95%, with <5–15% of the Tristan component. As discussed by Thompson et al. (2001), some samples exhibit evidence for crustal contamination, which could explain some of the spread of composition (e.g. Figs 10 and 18), even though samples with the most pronounced crustal contamination signatures were excluded from the plots. We note unradiogenic Pb isotopic compositions within three samples (Fig. 13; Erlank et al., 1984), indicating, from our previous interpretations, the presence of a very minor subducted ancient sediment component, possibly 2% (e.g. Fig. 18d). Duncan et al. (1990) and Thompson et al. (2001) have proposed a model for these magmas of decompression melting during advanced stages of lithospheric thinning.

Tafelkop type. Ewart et al. (1998a) identified these as predominantly mantle plume related magmas, specifically associated with the Tristan plume. Gibson et al. (2000), emphasizing their Fe-rich and picritic chemistry, considered them to represent decompression melting of Fe-rich ‘streaks’ within the mantle plume starting-head. Although the samples exhibiting the strongest indications of crustal contamination (Thompson et al., 2001) have been excluded from the various plots used in this work, the Tafelkop series show significant isotopic variation. The trace element and isotopic data suggest affinities with both Tristan da Cunha and Khumib–Urubici compositions. We infer a major Tristan plume (25–75%) component with N-MORB (10–50%) and a minor subducted sediment component (<3–15%) in this magma type. We suggest that the latter is supported by the unradiogenic ²⁰⁶Pb/²⁰⁴Pb composition in one of the samples (97SB73; Gibson et al., 2000), in line with arguments given above.

We argue that this new interpretation of the Etendeka magma types, and, by analogy, the equivalent Paraná magma types, provides a plausible alternative to the conventional SCLM model. We have not attempted to include an SCLM component as an end-member, to demonstrate that sub-lithospheric sources alone can account for the diversity of observed trace element chemistries and isotopic compositions in the lavas. In reality, it is likely that upon thermal erosion, some degree of SCLM melting would occur. The refractory SCLM material is thought, however, to contribute minimally to the lithophile element abundances and isotope compositions modelled here, but is expected to affect compatible element abundances and Os isotope compositions. Os isotope data have been interpreted to support a lithospheric mantle source ‘variably depleted in Re’ for ultra-alkaline magmas in the Paraná region (Carlson et al., 1996), whereas Ellam et al. (1992) interpreted Os isotopic data for the picrites within the Mwenezi region of the Jurassic Karoo province of southern Africa as supporting mixing of an enriched SCLM component with a highly depleted, plume-derived, ‘ultra-MORB magma’. Although unradiogenic Os isotope ratios may provide an effective way of recognizing SCLM components in magmas, it is not clear, given the frequently proposed heterogeneity of the SLCM, to what extent the Os isotope data can quantitatively constrain the SCLM contributions.

Proposed enriched SCLM components have feasibly contributed to the chemical and isotopic diversity of Etendeka magmas. It is, however, the relative importance of metasomatized, enriched and isotopically extreme SCLM components that we question. We suggest that the role of such components has been greatly over-emphasized in the past, one reason being that, until recently, it was not considered geodynamically feasible for the source of chemical and isotopic provinciality to exist in the deeper mantle. The previously noted relationship between styles of enrichment common to both mantle xenoliths and host volcanic rocks (Erlank et al., 1984) may also be interpreted in terms of the interaction of sub-lithosphere-derived (i.e. plume and plume modified) magmas with the SCLM. In fact, our model predicts that geochemical signatures of the sub-lithospheric magmas are imparted to the SCLM, possibly to be resampled during later episodes or phases of magma ascent and intrusion. We emphasize four aspects of our model that we suggest may contribute to critical reassessment of the prevailing and complicated SCLM model for continental basalt provinces, as follows.

1. Large, deep plumes emanating from the core–mantle boundary naturally entrain undegassed lower-mantle material that can incorporate accumulated recycled crustal eclogitic and related materials en route to the depleted asthenosphere. Large igneous provinces, such as the Paraná–Etendeka, are characterized by exceedingly wide compositional ranges and scales of magmatism, which we suggest to be a natural consequence of mantle upwellings of this scale. The longevity of such upwellings over the lifetime of plume magmatism will lead, at times, to the possibility that ²⁰⁶Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb ratios evolve significantly over the lifetime of plume magmatism, and result in apparent decoupling of Pb from other radiogenic isotope systems.
   
2. The relatively unradiogenic ²⁰⁶Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb, and high ²⁰⁷Pb/²⁰⁴Pb compositions exhibited by most Khumib magmas and small subsets of samples of the Tafelberg, and some other basalt types, are difficult to reconcile with an origin in the SCLM (or lower crust given the scales of magmatism). We note the lack of lithospheric mantle xenoliths that contain Pb plotting to the left of the Geochron. We therefore concur with Eisele et al. (2002) and Murphy et al. (2002) that U-depleted subducted sediment (±MORB) stored in the deep mantle is a more likely source for Pb isotope compositions plotting to the left of the Geochron, yet significantly above the DM evolution curve. We further argue that the presence of similar Pb isotopic signatures in localized, plume-related oceanic basalts from the Walvis Ridge Site 525A, the Discovery and central Shona Ridges, and near the Bouvet Triple Junction, each near the southern Mid-Atlantic Ridge, are entirely inconsistent with an SCLM source for these Pb isotope signatures.
   
3. The contrasting Sr and Pb abundance patterns and isotopic compositions observed in the Khumib lavas (and small subsets of the Tafelberg and other basalt types) are difficult to explain with an SCLM origin, which would require special enrichment events whereby Pb is preferentially enriched and Sr depleted relative to elements of similar compatibility. Furthermore, because lithospheric mantle xenoliths have chondritic and sub-chondritic Nb/Ta ratios, and Pb isotope ratios plotting to the right of the Geochron, SCLM models need to resort to melting domains unsampled by xenoliths.
   
4. Sub-lithospheric mixing of melts derived from depleted asthenospheric upper mantle, Tristan plume, and a subducted sediment (±MORB) component provides an alternative, simpler explanation for the discontinuities and provinciality of the different magma types present (e.g. high-Ti and low-Ti types), rather than appealing to somewhat arbitrary, large-scale chemical discontinuities within the SCLM.
   

The model presented here for the Paraná–Etendeka Province provides an alternative to SCLM melting and needs to be tested in other plume-dominated large igneous provinces. Unusual isotopic and geochemical characteristics, exhibited by even only a small subset of studied lavas, deserve attention for the possible identification of long-lived subduction-modified crustal reservoirs in the mantle. Isotopic fingerprints alone may not yield an unequivocal solution, as exemplified by the Pb-isotope compositions of the studied rock suites. Distinctive low ²⁰⁶Pb/²⁰⁴Pb but high ²⁰⁷Pb/²⁰⁴Pb isotope compositions, plotting close to the Geochron, were first identified in oceanic basalts from the Walvis Ridge. They served to define the enriched-mantle I (EM-1) reservoir of Zindler & Hart (1986). If our model is correct these Pb isotope compositions characterize the recycled crustal end-member identified here (and in Pitcairn and Gaussberg) and the term ‘enriched’ is justified, despite the fact that the enriched character (i.e. the high ²⁰⁷Pb/²⁰⁴Pb) was created in continental crust, and not in the mantle.

It is now well established that the bulk silicate Earth plots onto the core-corrected Geochron at higher ²⁰⁷Pb/²⁰⁴Pb than the depleted MORB source (e.g. Galer & Goldstein, 1996; Kamber & Collerson, 1999; Murphy et al., 2003), in an area in common Pb space very similar to the EM-1 source defined by Zindler & Hart (1986). Undepleted material could exist in portions of the lowermost mantle and in deep plumes originating at the core–mantle boundary that have the potential to erupt melts with Pb isotope compositions that developed for >4·5 Gyr in the lowermost mantle. This creates a situation where Pb isotope ratios of very similar composition can have very different geodynamic implications.

The combination, however, of trace element fingerprints, together with Pb and other radiogenic isotope ratios, should allow distinction between melts from recycled material [the EM-1 mantle end-member component of Milner & le Roex (1996) and Zindler & Hart (1986)] from undepleted mantle, which does not possess a super-chondritic Nb/Ta, low Nb/Nb^*, low Cs/Rb or super-chondritic Sr-isotope ratios. Large igneous provinces offer the opportunity to gain further insight into the regional and temporal variation of deep mantle sources depending on the detailed chemistry of the plume magmas, and the age and compositional attributes of the subducted slab material (see Bell & Tilton, 2001).

	APPENDIX

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 

Table A1: Major and trace element analyses (unnormalized) of lavas from the Tristan da Cunha islands

Sample: 11897 11898 11900 11902 11903 TAS: TePh Bas Tep KTrBas TePh SiO2 52·89 43·11 45·78 47·28 55·04 TiO2 1·73 3·95 3·33 3·02 1·58 Al2O3 18·79 14·89 16·30 16·27 19·01 Fe2O3 6·51 13·67 12·16 10·76 5·60 MnO 0·18 0·17 0·19 0·14 0·17 MgO 1·62 6·71 4·36 4·40 1·29 CaO 5·90 10·74 9·26 9·20 5·16 Na2O 5·29 2·85 3·93 3·22 5·80 K2O 4·46 2·22 3·18 2·00 4·99 P2O5 0·44 0·67 0·95 0·42 0·31 LOI 0·17 0·1 0·1 1·25 0·37 Total 97·98 98·98 99·44 97·96 99·32 Rb 114 67·2 73·4 42·0 126 Ba 1254 732 857 654 1407 Sr 1479 1189 1268 920 1454 Th 14·9 7·16 9·94 7·14 16·2 U 3·50 1·73 2·32 0·955 3·80 Zr 506 271 365 293 519 Hf 10·3 6·22 8·30 6·61 10·7 Nb 124 67·2 90·4 62·9 141 Ta 7·66 4·05 5·47 3·74 9·36 Cr 0·02 82·2 4·84 35·0 V 84·4 367 228 262 75·4 Sc 3·52 21·5 13·3 18·3 2·14 Ni 2·76 42·3 9·19 42·3 2·73 Co 7·46 47·3 29·4 33·8 4·99 Pb 9·37 4·48 6·39 5·80 10·1 Zn 112 120 129 107 112 Cu 7·90 35·6 22·2 93·4 6·68 Y 32·3 25·2 30·0 25·1 35·2 La 111 58·7 79·1 54·8 119 Ce 217 123 162 110 236 Pr 24·0 14·9 19·1 13·0 26·2 Nd 84·3 58·9 72·7 49·7 92·3 Sm 13·1 10·6 12·6 9·29 14·1 Eu 3·73 3·19 3·69 2·86 4·00 Gd 9·79 8·66 10·0 7·85 10·6 Tb 1·34 1·14 1·34 1·07 1·45 Dy 6·86 5·77 6·74 5·51 7·44 Ho 1·26 1·02 1·22 1·00 1·38 Er 3·14 2·43 2·95 2·43 3·49 Tm 0·436 0·317 0·385 0·320 0·474 Yb 2·66 1·83 2·30 1·93 2·89 Lu 0·379 0·250 0·321 0·265 0·409 Li 9·38 4·80 7·04 5·52 10·8 Be 2·96 1·52 2·09 1·78 3·06 Cs 1·41 0·745 0·909 0·454 1·57 (206Pb/204Pb)m ± 2 SE 18·545 ± 2 18·481 ± 3 18·513 ± 2 18·045 ± 2 18·481 ± 5 (207Pb/204Pb)m ± 2 SE 15·533 ± 1 15·533 ± 1 15·533 ± 0 15·506 ± 1 15·519 ± 2 (208Pb/204Pb)m ± 2 SE 39·020 ± 2 38·961 ± 3 38·964 ± 2 38·278 ± 2 38·938 ± 4

Trace elements in ppm. Total Fe expressed as Fe₂O₃. Samples from British Museum of National History, collected during Royal Society Expedition 1962. Sample numbers refer to Department of Earth Sciences collections, the University of Queensland. Major element analyses by inductively coupled plasma optical emission spectrometry. Trace element analyses by ICP-MS, the University of Queensland (see text). TAS nomenclature: Bas, basanite; Tep, tephrite; KTrBas, potassic trachybasalt; TePh, tephriphonolite. 11897, Tristan da Cunha ‘C’; flow from Stony Beach, Cinder Cove, at sea level, latest flow prior to 1961 eruption. 11898, Tristan da Cunha ‘D’; flow-sill near Big Point, 10 feet above sea level. 11900, Tristan da Cunha ‘H’; seashore below settlement and east of Herald Point. 11902, Inaccessible Island; from behind huts at landing location, near the waterfall, 75 feet above sea level. 11903, Tristan da Cunha ‘G’; new lava, 1961; seaward margin of flow, 100 yards from extreme western limit of flow, 10 feet above sea level.

Table A2: Basaltic, andesitic and dacitic partition coefficients used in modelling calculations

Plagioclase Olivine Augite Pigeonite Orthopyroxene Magnetite Ilmenite Apatite Basaltic Rb 0·027 0·0002 0·011 0·0003 0·0006 0·004 0·001 0·001 Ba 0·48 0·017 0·077 0·0004 0·014 0·10 0·01 0·014 Sr 1·9 0·012 0·11 0·003 0·07 0·03 0·01 0·88 Th 0·13 0·025 0·052 0·14 0·0001 0·17 0·10 5·2 Zr 0·020 0·015 0·26 0·05 0·03 0·15 0·31 0·034 Nb 0·025 0·025 0·022 0·01 0·01 0·22 3·5 0·011 Cr 0·027 18 20·0 8·6 20·0 30·0 0·1 0 V 0·017 0·09 2·2 1·5 0·9 20·0 0·1 0·06 Ni 0·05 13·0 2·0 5·6 7·0 5·0 1·46 0 Sc 0·014 0·22 1·8 1·5 0·33 3·0 1·8 0·22 Pb 0·36 0·01 0·07 0·1 0·001 0·8 0·1 0 Y 0·026 0·036 0·57 0·35 0·3 0·1 0·005 2·47 La 0·128 0·0044 0·057 0·0125 0·002 0·446 0·098 9·3 Ce 0·104 0·0056 0·109 0·020 0·0039 0·419 0·11 9·6 Nd 0·0707 0·0096 0·222 0·049 0·016 0·435 0·14 10·0 Sm 0·056 0·0125 0·353 0·10 0·066 0·455 0·15 10·16 Eu 0·61 0·0167 0·425 0·125 0·084 0·46 0·145 10·56 Gd 0·035 0·0191 0·46 0·155 0·110 0·47 0·14 10·95 Tb 0·028 0·0228 0·50 0·185 0·160 0·484 0·14 10·93 Yb 0·0077 0·0394 0·458 0·40 0·57 0·473 0·17 6·45 Dacitic Rb 0·08 — 0·09 0·005 0·02 0·004 0·001 0 Ba 0·72 — 0·082 0·083 0·04 0·11 0·01 0·014 Sr 4·4 — 0·52 0·01 0·03 0·08 0·05 10·0 Th 0·05 — 0·06 0·14 0·10 0·50 0·43 1·1 Zr 0·1 — 1·0 0·068 0·064 0·26 0·13 0·034 Nb 0·05 — 0·04 0·08 0·027 1·6 4·5 0·011 Cr 0·05 — 7·5 20·0 15·0 10·0 0·1 0 V 0·02 — 5·2 4·0 2·1 20·0 0·1 0·06 Ni 0·21 — 6·0 5·6 10·0 5·0 1·0 0 Sc 0·02 — 10·0 14·0 4·7 3·4 2·0 0·22 Pb 0·58 — 0·27 0·05 0·1 0·3 0·05 0 Y 0·08 — 2·6 1·2 0·61 1·1 2·3 70·0 La 0·25 — 0·35 0·625 0·9 0·5 3·6 25·6 Ce 0·20 — 0·52 0·775 0·5 0·6 3·5 33·4 Nd 0·20 — 1·2 1·0 0·22 0·8 3·2 55·8 Sm 0·20 — 1·7 1·28 0·27 0·80 2·8 65·5 Eu 1·6 — 1·75 0·68 0·31 0·80 2·7 71·0 Gd 0·10 — 1·8 1·36 0·34 0·80 2·6 76·0 Tb 0·10 — 2·0 1·4 0·50 0·80 2·5 82·0 Yb 0·04 — 1·8 1·56 0·85 0·6 1·6 64·7 Andesitic Rb 0·017 — 0·013 0·003 0·015 0·004 0 0·001 Ba 0·48 — 0·024 0·0004 0·014 0·4 0 0·014 Sr 3·1 — 0·11 0·003 0·032 0·10 0 0·88 Th 0·19 — 0·035 0·14 0·054 0·17 0 5·2 Zr 0·10 — 0·3 0·10 0·10 0·4 0 0·034 Nb 0·022 — 0·043 0·02 0·02 2·0 0 0·011 Cr 0·027 — 35 8·6 15·0 58 0 0 V 0·017 — 1·7 1·5 0·60 30 0 0·061 Ni 0·12 — 3·7 5·6 9·5 10 1·5 0 Sc 0·014 — 6·0 1·5 2·0 5·9 1·8 0·22 Pb 0·58 — 0·10 0·10 0·10 0·3 0 0 Y 0·04 — 2·1 0·35 0·46 0·5 0 2·47 La 0·29 — 0·158 0·0125 0·238 0·446 0·098 9·3 Ce 0·169 — 0·223 0·02 0·301 0·419 0·11 9·6 Nd 0·114 — 0·36 0·049 0·144 0·435 0·14 10·0 Sm 0·091 — 0·424 0·10 0·141 0·455 0·15 10·16 Eu 0·33 — 0·487 0·125 0·133 0·413 0·10 6·93 Gd 0·073 — 0·569 0·155 0·131 0·470 0·14 10·95 Tb 0·063 — 0·556 0·185 0·130 0·484 0·14 10·93 Yb 0·021 — 0·563 1·56 0·257 0·473 0·17 6·45

Partition coefficients after Ewart et al. (1998a), with additional and modified data from unpublished database (A. Ewart, 2002); Hart & Dunn (1993), Dunn & Sen (1994), Ewart & Griffin (1994) and Jeffries et al. (1995).

	SUPPLEMENTARY DATA

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
Supplementary data for this paper are available on Journal of Petrology online.

	ACKNOWLEDGEMENTS

 
We wish to acknowledge the logistic support, over some 6 years, of the Namibian Geological Survey that allowed the mapping programme to be undertaken. A.E. wishes to thank the FRD, South Africa for some financial support, and R. Rasch of the University of Queensland Centre for Microscopy and Microanalysis for technical support whilst undertaking microprobe analyses. We are also indebted to Drs A. le Roex, D. Peate, E. M. Piccirillo and M. Wilson for detailed and constructive comments on versions of the manuscript.

	REFERENCES

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING ANALYTICAL TECHNIQUES MAGMA TYPES AND THEIR... MINERALOGY CHEMISTRY TRACE ELEMENTS ISOTOPIC CHEMISTRY PETROGENESIS SUMMARY AND CONCLUSIONS APPENDIX SUPPLEMENTARY DATA REFERENCES

 
Arndt, N. T. & Christensen, U. (1992). The role of lithospheric mantle in continental flood volcanism: thermal and geochemical constraints. Journal of Geophysical Research 97(B7), 10967–10981.[CrossRef]

Arndt, N. T., Czamanske, G. K., Wooden, J. L. & Fedorenko, V. A. (1993). Mantle and crustal contributions to continental flood volcanism. Tectonophysics 223, 39–52.[CrossRef][Web of Science]

Becker, H., Jochum, K. P. & Carlson, R. W. (2000). Trace element fractionation during dehydration of eclogites from high-pressure terranes and the implications for element fluxes in subduction zones. Chemical Geology 163, 65–99.[CrossRef][Web of Science]

Becker, T., Hansen, B. T., Weber, K. & Wiegand, B. (1996). U–Pb and Rb–Sr isotopic data for the Mooirivier Complex, Weener Igneous Suite and Gaub Valley Formation (Rehoboth sequence) in the Nauchas area and their significance for Paleoproterozoic crustal evolution in Namibia. Communications of the Geological Survey of Namibia 11, 31–46.

Bell, K. & Tilton, G. R. (2001). Nd, Pb and Sr isotopic compositions of East African carbonatites: evidence for mantle mixing and plume inhomogeneity. Journal of Petrology 42, 1927–1945.[Abstract/Free Full Text]

Bellieni, G., Brotzu, P., Comin-Chiaramonti, P., Ernesto, M., Melfi, A., Pacca, I. G. & Picirillo, E. M. (1984). Flood basalt to rhyolite suites in the southern Paraná Plateau (Brazil): palaeomagnetism, petrogenesis and geodynamic implications. Journal of Petrology 25, 579–618.[Abstract/Free Full Text]

Bellieni, G., Comin-Chiaramonti, P., Marques, L. S., Melfi, A. J., Nardy, A. J. R., Papatrechas, C., Piccirillo, E. M., Roisenberg, A. & Stolfa, D. (1986). Petrogenetic aspects of acid and basaltic lavas from the Paraná Plateau (Brazil): geological, mineralogical and petrochemical relationships. Journal of Petrology 27, 915–944.[Abstract/Free Full Text]

Brandon, A. D., Walker, R. J., Morgan, J. W., Norman, M. D. & Prichard, H. M. (1998). Coupled Os-186 and Os-187 evidence for core–mantle interaction. Science 280, 1570–1573.[CrossRef][Web of Science][Medline]

Brenan, J. M., Shaw, H. F. & Ryerson, F. J. (1994). Rutile–aqueous fluid partitioning of Nb, Ta, Hf, U and Th: implications for high field strength element depletions in island-arc basalts. Earth and Planetary Science Letters 128, 327–339.[CrossRef][Web of Science]

Brenan, J. M., Shaw, H. F. & Ryerson, F. J. (1995a). Experimental evidence for the origin of lead enrichment in convergent-margin magmas. Nature 378, 54–56.[CrossRef]

Brenan, J. M., Shaw, H. F., Ryerson, F. J. & Phinney, D. L. (1995b). Mineral–aqueous fluid partitioning of trace elements at 900°C and 2·0 GPa: constraints on the trace element chemistry of mantle and deep crustal fluids. Geochimica et Cosmochimica Acta 59, 3331–3350.[CrossRef][Web of Science]

Campbell, I. H. & Griffiths, R. W. (1990). Implications of mantle plume structure for the evolution of flood basalts. Earth and Planetary Science Letters 99, 79–93.[CrossRef][Web of Science]

Campbell, I. H. & Griffiths, R. W. (1992). The changing nature of mantle hotspots through time: implications for the chemical evolution of the mantle. Journal of Geology 92, 497–523.

Carlson, R. W., Esperança, S. & Svisero, D. P. (1996). Chemical and Os isotopic study of Cretaceous potassic rocks from Southern Brazil. Contributions to Mineralogy and Petrology 125, 393–405.[CrossRef][Web of Science]

Chang, H. K., Kowsmann, R. O., Figueiredo, A. M. F. & Bender, A. A. (1992). Tectonics and stratigraphy of the East Brazil Rift system: an overview. Tectonophysics 213, 97–138.[CrossRef][Web of Science]

Christensen, U. R. (1996). The influence of trench migration on slab penetration into the lower mantle. Earth and Planetary Science Letters 140, 27–39.[CrossRef][Web of Science]

Cliff, R. A., Baker, P. E. & Mateer, N. J. (1991). Geochemistry of Inaccessible Island volcanics. Chemical Geology 92, 251–260.[CrossRef][Web of Science]

Collerson, K. D., Hapugoda, S., Kamber, B. S. & Williams, Q. (2000). Rocks from the mantle transition zone: majorite-bearing xenoliths from Malaita, Southwest Pacific. Science 288, 1215–1223.[CrossRef][Web of Science][Medline]

Collerson, K. D., Kamber, B. S. & Schoenberg, R. (2002). Applications of accurate high-precision Pb isotope ratio measurements by multi-collector ICP-MS. Chemical Geology 188, 65–83.[CrossRef][Web of Science]

Comin-Chiaramonti, P., Cundari, A., Gomes, C. B., Piccirillo, E. M., Censi, P., DeMin, A., Bellieni, G., Velazquez, V. F. & Orué, D. (1992). Potassic dyke swarm in the Sapucai Graben, eastern Paraguay: petrographical, mineralogical and geochemical outlines. Lithos 28, 283–301.[CrossRef][Web of Science]

Cox, K. G. (1988). The Karoo Province. In: Macdougall, J. D. (ed.) Continental Flood Basalts. Dordrecht: Kluwer Academic, pp. 239–271.

Davies, G. R., Spriggs, A. J. & Nixon, P. H. (2001). A non-cognate origin for the Gibeon kimberlite megacryst suite, Namibia: implications for the origin of Namibian kimberlites. Journal of Petrology 42, 159–172.[Abstract/Free Full Text]

Demarchi, G., Antonini, P., Piccirillo, E. M., Orsi, G., Civetta, L. & D'Antonio, M. (2001). Significance of orthopyroxene and major element constraints on the petrogenesis of Ferrar tholeiites from southern Prince Albert Mountains, Victoria Land, Antarctica. Contributions to Mineralogy and Petrology 142, 127–146.[Web of Science]

De Wit, M., Jeffrey, M., Berg, H. & Nicolaysen, L. (1988). Geological map of sectors of Gondwana reconstructed to their disposition 150 Ma. Tulsa, OK: American Association of Petroleum Geologists; Johannesburg: University of the Witswatersrand, scale 1:10 000 000.

Douglass, J., Schilling, J.-G. & Fontignie, D. (1999). Plume–ridge interactions of the Discovery and Shona mantle plumes with the southern Mid-Atlantic Ridge (40°–55°S). Journal of Geophysical Research 104(B2), 2941–2962.[CrossRef]

Duncan, A. R. (1984). Appendix 1. Analytical techniques and database descriptions. In: Erlank, A. J. (ed.) Petrogenesis of the Volcanic Rocks of the Karoo Province. Geological Society of South Africa, Special Publication 13, 389–395.

Duncan, A. R. (1987). The Karoo Igneous Province—a problem area for inferring tectonic setting from basalt geochemistry. Journal of Volcanology and Geothermal Research 32, 13–34.[CrossRef][Web of Science]

Duncan, A. R., Newton, S. R., van den Berg, C. & Reid, D. L. (1989). Geochemistry and petrology of dolerite sills in the Huab River valley, Damaraland, north western Namibia. Communications of the Geological Survey of Namibia 5, 5–18.

Duncan, A. R., Armstrong, R. A., Erlank, A. J., Marsh, J. S. & Watkins, R. T. (1990). MORB-related dolerites associated with the final phase of Karoo flood basalt volcanism of southern Africa. In: Parker, J. A. J., Rickwood, P. C. & Tucker, D. H. (eds) Mafic Dykes and Emplacement Mechanisms. Proceedings of 2nd International Dyke Conference. Rotterdam: A. A. Balkema, pp. 119–130.

Dunn, T. & Sen, C. (1994). Mineral/matrix partition coefficients for orthopyroxene, plagioclase, and olivine in basaltic to andesitic systems: a combined analytical and experimental study. Geochimica et Cosmochimica Acta 58, 717–733.[CrossRef][Web of Science]

Eggins, S. M., Woodhead, J. D., Kinsley, L. P. J., Mortimer, G. E., Sylvester, P., McCulloch, M. T., Hergt, J. M. & Handler, M. R. (1997). A simple method for the precise determination of 40 trace elements in geological samples by ICPMS using enriched isotope internal standardisation. Chemical Geology 134, 311–326.[CrossRef][Web of Science]

Eisele, J., Sharma, M., Galer, S. J. G., Blichert-Toft, J., Devey, C. W. & Hofmann, A. W. (2002). The role of sediment recycling in EM-1 inferred from Os, Pb, Hf, Nd, Sr isotope and trace element systematics of the Pitcairn hotspot. Earth and Planetary Science Letters 196, 197–212.[CrossRef][Web of Science]

Ellam, R. M. & Cox, K. G. (1991). An interpretation of Karoo picrite basalts in terms of interaction between asthenospheric magmas and the mantle lithosphere. Earth and Planetary Science Letters 105, 330–342.[CrossRef][Web of Science]

Ellam, R. M., Carlson, R. W. & Shirley, S. B. (1992). Evidence from Re–Os isotopes for plume–lithosphere mixing in Karoo flood basalt genesis. Nature 359, 718–721.[CrossRef]

Erlank, A. J., Marsh, J. S., Duncan, A. R., Miller, R. McG., Hawkesworth, C. J., Betton, P. J. & Rex, D. C. (1984). Geochemistry and petrogenesis of the Etendeka volcanic rocks from SWA/Namibia. In: Erlank, A. J. (ed.) Petrogenesis of the Volcanic Rocks of the Karoo Province. Geological Society of South Africa, Special Publication 13, 195–245.

Ewart, A. & Griffin, W. L. (1994). Application of proton-microprobe data to trace element partitioning in volcanic rocks. Chemical Geology 117, 251–284.[CrossRef][Web of Science]

Ewart, A., Milner, S. C., Armstrong, R. A. & Duncan, A. R. (1998a). Etendeka volcanism of the Goboboseb Mountains and Messum Igneous complex, Namibia. Part I: Geochemical evidence of Early Cretaceous Tristan plume melts and the role of crustal contamination in the Paraná–Etendeka CFB. Journal of Petrology 39, 191–225.

Ewart, A., Milner, S. C., Armstrong, R. A. & Duncan, A. R. (1998b). Etendeka volcanism of the Goboboseb Mountains and Messum Igneous Complex, Namibia. Part II: Voluminous quartz latite volcanism of the Awahab magma system. Journal of Petrology 39, 227–253.

Ewart, A., Marsh, J. S., Milner, S. C., Duncan, A. R., Kamber, B. S. & Armstrong, R. A. (2004). Petrology and geochemistry of early Cretaceous bimodal continental flood volcanism of the NW Etendeka, Namibia. Part 2: Characteristics and petrogenesis of the high-Ti latite and high-Ti and low-Ti voluminous quartz latite eruptives. Journal of Petrology 44, 107–138.

Fraser, K. J., Hawkesworth, C. J., Erlank, A. J., Mitchell, R. H. & Scott-Smith, B. H. (1985). Sr, Nd and Pb isotope and minor element geochemistry of lamproites and kimberlites. Earth and Planetary Science Letters 76, 57–70.[CrossRef][Web of Science]

Fukao, Y., Widiyantoro, S. & Obayashi, M. (2001). Stagnant slabs in the upper and lower mantle transition region. Reviews of Geophysics 39, 291–323.[CrossRef][Web of Science]

Galer, S. J. G. & Goldstein, S. L. (1996). Influence of accretion on lead in the Earth. In: Basu, A. & Hart, S. (eds) Earth Processes: Reading the Isotopic Code. Washington, DC: American Geophysical Union, pp. 75–98.

Gallagher, K. & Hawkesworth, C. (1992). Dehydration melting and the generation of continental flood basalts. Nature 358, 57–59.[CrossRef]

Gasperini, D., Blichert-Toft, J., Bosch, D., Del Moro, A., Macera, P., Télouk, P. & Albarède, F. (2000). Evidence from Sardinian basalt geochemistry for recycling of plume heads into the Earth's mantle. Nature 408, 701–704.[CrossRef]

Ghiorso, M. S. & Sack, R. O. (1995). Chemical mass transfer in magmatic processes. IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid–solid equilibria in magmatic systems at elevated temperatures and pressures. Contributions to Mineralogy and Petrology 119, 197–212.[Web of Science]

Gibson, S. A., Thompson, R. N., Dickin, A. P. & Leonardos, O. H. (1995a). High-Ti and low-Ti mafic potassic magmas: key to plume–lithosphere interactions and continental flood-basalt genesis. Earth and Planetary Science Letters 136, 149–165.[CrossRef][Web of Science]

Gibson, S. A., Thompson, R. N., Leonardos, O. H., Dickin, A. P. & Mitchell, J. G. (1995b). The late Cretaceous impact of the Trindade mantle plume: evidence from large-volume, mafic, potassic magmatism in S. E. Brazil. Journal of Petrology 36, 189–229.[Abstract/Free Full Text]

Gibson, S. A., Thompson, R. N. & Dickin, A. P. (2000). Ferropicrites: geochemical evidence for Fe-rich streaks in upswelling mantle plumes. Earth and Planetary Science Letters 174, 355–374.[CrossRef][Web of Science]

Griffiths, R. W. & Campbell, I. H. (1990). Stirring and structure in mantle starting plumes. Earth and Planetary Science Letters 99, 66–78.[CrossRef][Web of Science]

Griffiths, R. W. & Campbell, I. H. (1991). On the dynamics of long-lived plume conduits in the convecting mantle. Earth and Planetary Science Letters 103, 214–227.[CrossRef][Web of Science]

Hart, S. R. & Dunn, T. (1993). Experimental cpx/melt partitioning of 24 trace elements. Contributions to Mineralogy and Petrology 113, 1–8.[CrossRef][Web of Science]

Hawkesworth, C. J., Erlank, A. F., Marsh, J. S., Menzies, M. A. & van Calsteren, P. W. C. (1983). Evolution of continental lithosphere: evidence from volcanics and xenoliths in Southern Africa. In: Hawkesworth, C. J. & Norry, M. J. (eds) Continental Flood Basalts and Mantle Xenoliths. Nantwich: Shiva, pp. 111–138.

Hawkesworth, C. J., Marsh, J. S., Duncan, A. R., Erlank, A. J. & Norry, M. J. (1984). The role of continental lithosphere in the generation of the Karoo volcanic rocks: evidence from combined Nd- and Sr-isotope studies. In: Erlank, A. J. (ed.) Petrogenesis of the Volcanic Rocks of the Karoo Province. Geological Society of South Africa, Special Publication 13, 341–354.

Hawkesworth, C. J., Gallagher, K., Kelley, S., Mantovani, M. S. M., Peate, D. W., Regelous, M. & Rogers, N. W. (1992). Paraná magmatism and the opening of the South Atlantic. In: Storey, B. C., Alabaster, A. & Pankhurst, R. J. (eds) Magmatism and the Causes of Continental Break-up. Geological Society, London, Special Publications 68, 221–240.

Hawkesworth, C., Kelley, S., Turner, S., Le Roex, A. & Storey, B. (1999). Mantle processes during Gondwana break-up and dispersal. Journal of African Earth Sciences B28, 239–261.

Hergt, J. M., Chappell, B. W., Faure, G. & Mensing, T. M. (1989). The geochemistry of Jurassic dolerites from Portal Peak, Antarctica. Contributions to Mineralogy and Petrology 102, 298–305.[CrossRef][Web of Science]

Hergt, J. M., Peate, D. W. & Hawkesworth, C. J. (1991). The petrogenesis of Mesozoic Gondwana low-Ti flood basalts. Earth and Planetary Science Letters 105, 134–148.[CrossRef][Web of Science]

Hoch, M., Rehkamper, M. & Tobschall, H. J. (2001). Sr, Nd, Pb and O isotopes of minettes from Schirmacher Oasis, East Antarctica: a case of mantle metasomatism involving subducted continental material. Journal of Petrology 42, 1387–1400.[Abstract/Free Full Text]

Huang, Y.-M., Hawkesworth, C. J., van Calsteren, P. & McDermott, F. (1995). Geochemical characteristics and origin of the Jacupiranga carbonatites, Brazil. Chemical Geology 119, 79–99.[CrossRef][Web of Science]

Humphris, S. E. & Thompson, G. (1983). Geochemistry of rare earth elements in basalts from the Walvis Ridge: implications for its origin and evolution. Earth and Planetary Science Letters 66, 223–242.[CrossRef][Web of Science]

Irifune, T., Ringwood, A. E. & Hibberson, W. O. (1994). Subduction of continental crust and terrigenous and pelagic sediments: an experimental study. Earth and Planetary Science Letters 126, 351–368.[CrossRef][Web of Science]

Jeffries, T. E., Perkins, W. T. & Pearce, N. J. G. (1995). Measurements of trace elements in basalts and their phenocrysts by laser probe microanalysis inductively coupled plasma mass spectrometry (LPMA–ICP-MS). Chemical Geology 121, 131–144.[CrossRef][Web of Science]

Jerram, D., Mountney, N., Holzförster, F. & Stollhofen, H. (1999). Internal stratigraphic relationships in the Etendeka Group in the Huab Basin, NW Namibia: understanding the onset of flood volcanism. Journal of Geodynamics 28, 393–418.[CrossRef][Web of Science]

Ji, Y. & Nataf, H.-C. (1998). Detection of mantle plumes in the lower mantle by diffraction tomography. Earth and Planetary Science Letters 159, 99–115.[CrossRef][Web of Science]

Kamber, B. S. & Collerson, K. D. (1999). Origin of ocean island basalts: a new model based on lead and helium isotope systematics. Journal of Geophysical Research 104, 25479–25491.[CrossRef]

Kamber, B. S. & Collerson, K. D. (2000a). Role of ‘hidden’ deeply subducted slabs in mantle depletion. Chemical Geology 166, 241–254.[CrossRef][Web of Science]

Kamber, B. S. & Collerson, K. D. (2000b). Zr/Nb systematics of ocean island basalts reassessed—the case for binary mixing. Journal of Petrology 41, 1007–1021.[Abstract/Free Full Text]

Kamenetsky, V. S., Maas, R., Sushchevskaya, N. M., Norman, M. D., Cartwright, I. & Peyve, A. (2001). Remnants of Gondwanan continental lithosphere in oceanic upper mantle: evidence from the South Atlantic Ridge. Geology 29, 243–246.[Abstract/Free Full Text]

Kirstein, L. A., Kelley, S., Hawkesworth, C., Turner, S., Mantovani, M. & Wijbrans, J. (2001). Protracted felsic magmatic activity associated with opening of the South Atlantic. Journal of the Geological Society, London 158, 583–592.[Abstract/Free Full Text]

Kramers, J. D. & Tolstikhin I. N. (1997). Two terrestrial lead isotope paradoxes, forward transport modelling, core formation and the history of the continental crust. Chemical Geology 139, 75–110.[CrossRef][Web of Science]

le Roex, A. P. & Lanyon, R. (1998). Isotope and trace element geochemistry of Cretaceous Damaraland lamprophyres and carbonatites, northwestern Namibia: evidence for plume–lithosphere interactions. Journal of Petrology 39, 1117–1146.

le Roex, A. P., Cliff, R. A. & Adair, B. J. I. (1990). Tristan da Cunha, South Atlantic: geochemistry and petrogenesis of a basanite–phonolite lava series. Journal of Petrology 31, 779–812.[Free Full Text]

Lindsley, D. H. (1983). Pyroxene thermometry. American Mineralogist 68, 477–493.[Abstract]

Machado, N., Valladares, C., Heilbron, M. & Valeriano, C. (1996). U–Pb geochronology of the Ribeira belt (Brazil) and implications for the evolution of the Brazilian Orogeny. Precambrian Research 79, 347–361.[CrossRef][Web of Science]

Mahoney, J. J., White, W. M., Upton, B. G. J., Neal, C. R. & Scrutton, R. A. (1996). Beyond EM-1: lavas from Afanasy-Nikitin Rise and the Crozet Archipelago, Indian Ocean. Geology 24, 615–618.[Abstract/Free Full Text]

Marques, L. S., Dupré, B. & Piccirillo, E. M. (1999). Mantle source compositions of the Paraná Magmatic Province (southern Brazil): evidence from trace element and Sr–Nd–Pb isotope geochemistry. Journal of Geodynamics 28, 429–459.

Marsh, J. S., Ewart, A., Milner, S. C., Duncan, A. R. & Miller, R. McG. (2001). The Etendeka Igneous Province: magma types and their stratigraphic distribution with implications for the evolution of the Paraná–Etendeka flood basalt province. Bulletin of Volcanology 62, 464–486.[CrossRef][Web of Science]

McDonough, W. F. (1990). Constraints on the composition of the continental lithospheric mantle. Earth and Planetary Science Letters 101, 1–18.[CrossRef][Web of Science]

McKenzie, D. & Bickle, M. J. (1988). The volume and composition of melt generated by extension of the lithosphere. Journal of Petrology 29, 625–679.[Abstract/Free Full Text]

Melfi, A. J., Piccirillo, E. M. & Nardy, A. J. R. (1988). Geological and magmatic aspects of the Paraná Basin—an introduction. In: Piccirillo, E. M. & Melfi, A. J. (eds) The Mesozoic Flood Volcanism of the Paraná Basin: Petrogenetic and Geophysical Aspects. São Paulo: Instituto Astronomico e Geofisico Publishers, IAG–USP Press, pp. 1–13.

Mellini, M., Carbonin, S., Dal Negro, A. & Piccirillo, E. M. (1988). Tholeiitic hypabyssal dykes: how many clinopyroxenes? Lithos 22, 127–134.[CrossRef][Web of Science]

Menzies, M. A. (1992). The lower lithosphere as a major source for continental flood basalts: a re-appraisal. In: Storey, B. C., Alabaster, A. & Pankhurst, R. J. (eds) Magmatism and the Causes of Continental Break-up. Geological Society, London, Special Publications 68, 31–39.

Miller, R. McG. (1983). The Pan-African Damara Orogen of South West Africa/Namibia. In: Miller, R. McG. (ed.) Evolution of the Damara Orogen of South West Africa/Namibia. Geological Society of South Africa, Special Publication 11, 431–515.

Milner, S. C. (1988). The geology and geochemistry of the Etendeka Formation quartz latites, Namibia. Ph.D. thesis, University of Cape Town, 263 pp.

Milner, S. C. & le Roex, A. P. (1996). Isotope characteristics of the Okenyenya igneous complex, northwestern Namibia: constraints on the composition of the early Tristan plume and the origin of the EM1 mantle component. Earth and Planetary Science Letters 141, 277–291.[CrossRef][Web of Science]

Milner, S. C., Duncan, A. R. & Ewart, A. (1992). Quartz latite rheoignimbrite flows of the Etendeka formation, north-western Namibia. Bulletin of Volcanology 54, 200–219.[Web of Science]

Milner, S. C., Duncan, A. R., Ewart, A. & Marsh, J. S. (1995a). Promotion of the Etendeka Formation to Group status: a new integrated stratigraphy. Communications of the Geological Survey of Namibia 9, 5–11.

Milner, S. C., Duncan, A. R., Whittingham, A. M. & Ewart, A. (1995b). Trans-Atlantic correlation of eruptive sequences and individual silicic volcanic units within the Paraná–Etendeka igneous province. Journal of Volcanology and Geothermal Research 69, 137–157.[CrossRef][Web of Science]

Milner, S. C., le Roex, A. P. & O'Connor, J. M. (1995c). Age of Mesozoic igneous rocks in northwestern Namibia and their relationship to continental breakup. Journal of the Geological Society, London 152, 97–104.[Abstract/Free Full Text]

Morgan, W. J. (1981). Hotspot tracks and the opening of the Atlantic and Indian Oceans. In: Emiliani, C. (ed.) The Sea. Volume 7: The Oceanic Lithosphere. New York: Wiley, pp. 443–487.

Murphy, D. T., Collerson, K. D. & Kamber, B. S. (2002). Lamproites from Gaussberg, Antarctica: possible Transition Zone melts of Archaean subducted sediments. Journal of Petrology 43, 981–1001.[Abstract/Free Full Text]

Murphy, D. T., Kamber, B. S. & Collerson, K. D. (2003). A refined solution to the first terrestrial Pb-isotope paradox. Journal of Petrology 44, 39–53.[Abstract/Free Full Text]

Neal, C. R., Mahoney, J. J., Kroenke, L. W., Duncan, R. A. & Petterson, M. G. (1997). The Ontong Java plateau. In: Mahoney, J. J. & Coffin, M. E. (eds) Large Igneous Provinces: Continental, Oceanic, and Planetary Flood Volcanism. Geophysical Monograph, American Geophysical Union 100, 183–216.

Norrish, K. & Hutton, J. T. (1969). An accurate X-ray spectrographic method for the analysis of a wide range of geological samples. Geochimica et Cosmochimica Acta 32, 431–453.

O'Connor, J. M. & Duncan, R. A. (1990). Evolution of the Walvis Ridge and Rio Grande Rise hotspot system—implications for African and South American plate motions over plumes. Journal of Geophysical Research 95, 17475–17502.[CrossRef]

O'Nions, R. K. & Pankhurst, R. J. (1974). Petrogenetic significance of isotope and trace element variations in volcanic rocks from the Mid-Atlantic. Journal of Petrology 15, 603–634.[Abstract/Free Full Text]

Peate, D. W. (1997). The Paraná–Etendeka Province. In: Mahoney, J. J. & Coffin, M. F. (eds) Large Igneous Provinces: Continental Oceanic and Planetary Flood Volcanism. Geophysical Monograph, American Geophysical Union 100, 217–245.

Peate, D. W. & Hawkesworth, C. J. (1996). Lithospheric to asthenospheric transition in low-Ti flood basalts from southern Paraná, Brazil. Chemical Geology 127, 1–24.[CrossRef][Web of Science]

Peate, D. W., Hawkesworth, C. J., Mantovani, M. S. M. & Shukovsky, W. (1990). Mantle plumes and flood-basalt stratigraphy in the Paraná, South America. Geology, 18, 1223–1226.[Abstract/Free Full Text]

Peate, D. W., Hawkesworth, C. J. & Mantovani, M. S. M. (1992). Chemical stratigraphy of the Paraná lavas (South America): classification of magma types and their spatial distribution. Bulletin of Volcanology 55, 119–139.[CrossRef][Web of Science]

Peate, D. W., Hawkesworth, C. J., Mantovani, M. S. M., Rogers, N. W. & Turner, S. P. (1999). Petrogenesis and stratigraphy of the high Ti/Y Urubici magma type in the Paraná flood basalt province and implications for the nature of ‘Dupal’-type mantle in the South Atlantic region. Journal of Petrology 40, 451–473.[CrossRef][Web of Science]

Piccirillo, E. M., Comin-Chiaramonti, P., Melfi, A. J., Stolfa, D., Bellieni, G., Marques, L. S., Giaretta, A., Nardy, A. J. R., Pinese, J. P. P., Raposo, M. I. B. & Roisenberg, A. (1988a). V. Petrochemistry of continental flood basalt–rhyolite suites and related intrusives from the Paraná Basin (Brazil). In: Piccirillo, E. M. & Melfi, A. J. (eds) The Mesozoic Flood Volcanism of the Paraná Basin: Petrogenetic and Geophysical Aspects. São Paulo: Instituto Astronomica e Geofisico Publishers, IAG–USP Press, pp. 107–156.

Piccirillo, E. M., Comin-Chiarmonti, P., Bellieni, G., Civetta, L., Marques, L. S., Melfi, A. J., Petrini, R., Raposo, M. I. B. & Stolfa, D. (1988b). VII. Petrogenetic aspects of continental flood basalt–rhyolite suites from the Paraná Basin (Brazil). In: Piccirillo, E. M. & Melfi, A. J. (eds) The Mesozoic Flood Volcanism of the Paraná Basin: Petrogenetic and Geophysical Aspects. São Paulo: Instituto Astronomica e Geofisico Publishers, IAG–USP Press, pp. 179–205.

Piccirillo, E. M., Civetta, L., Petrini, R., Longinelli, A., Bellieni, G., Comin-Chiaramonti, P., Marques, L. S. & Melfi, A. J. (1989). Regional variations within the Paraná flood basalts (Southern Brazil): evidence for subcontinental mantle heterogeneity and crustal contamination. Chemical Geology 75, 103–122.[CrossRef][Web of Science]

Plank, T. & Langmuir, C. H. (1998). The chemical composition of subducting sediment and its consequences for the crust and mantle. Chemical Geology 145, 325–394.[CrossRef][Web of Science]

Renne, P. R., Ernesto, M., Pacca, I. G., Coe, R. S., Glen, J. M., Prévot, M. & Perrin, M. (1992). The age of the Paraná flood volcanism, rifting of Gondwanaland, and the Jurassic–Cretaceous boundary. Science 258, 975–979.[Abstract/Free Full Text]

Renne, P. R., Glen, J. M., Milner, S. C. & Duncan, A. R. (1996). Age of Etendeka flood volcanism and associated intrusions in southwestern Africa. Geology 24, 659–662.[Abstract/Free Full Text]

Richardson, S. H., Erlank, A. J., Duncan, A. R. & Reid, D. L. (1982). Correlated Nd, Sr and Pb isotope variation in the Walvis Ridge basalts and implications for the evolution of their mantle source. Earth and Planetary Science Letters 59, 327–342.[CrossRef][Web of Science]

Ringwood, A. E. (1989). Constitution and evolution of the mantle. In: Ross, A. L. J. J., Ferguson, J. et al. (eds) Kimberlites and Related Rocks: their Composition, Origin and Emplacement. Geological Society of Australia, Special Publication 457–485.

Ringwood, A. E. (1991). Phase transformations and their bearing on the constitution and dynamics of the mantle. Geochimica et Cosmochimica Acta 55, 2083–2110.[CrossRef][Web of Science]

Ringwood, A. E., Kesson, S. E., Hibberson, W. & Ware, N. (1992). Origin of kimberlites and related magmas. Earth and Planetary Science Letters 113, 521–538.[CrossRef][Web of Science]

Romanowicz, B. & Gung, Y. (2002). Superplumes from the core–mantle boundary to the lithosphere: implications for heat flux. Science 296, 513–516.[CrossRef][Web of Science][Medline]

Rudnick, R. L. & Fountain, D. M. (1995). Nature and composition of the continental crust: a lower crustal perspective. Reviews of Geophysics 33, 267–309.[CrossRef][Web of Science]

Rudnick, R. L. & Goldstein, S. L. (1990). The Pb isotopic compositions of lower crustal xenoliths and the evolution of lower crustal Pb. Earth and Planetary Science Letters 98, 192–207.[CrossRef][Web of Science]

Saal, A. E., Hart, S. R., Shimizu, N., Hauri, E. H. & Layne, G. D. (1998). Pb isotopic variability in melt inclusions from oceanic island basalts, Polynesia. Science 282, 1481–1484.[CrossRef][Web of Science][Medline]

Schmitt, A. K., Emermann, R., Trumbull, R. B., Bühn, B. & Henjes-Kunst, F. (2000) Petrogenesis and ⁴⁰Ar/³⁹Ar geochronology of the Brandberg Complex, Namibia: evidence for a major mantle contribution in metaluminous and peralkaline granites. Journal of Petrology 41, 1207–1239.[Abstract/Free Full Text]

Seth, B., Kröner, A., Mezger, K., Nemchin, A. A., Pidgeon, R. T. & Okrusch, M. (1998). Archaean to Neoproterozoic magmatic events in the Kaoko belt of NW Namibia and their geodynamic significance. Precambrian Research 92, 341–363.[CrossRef][Web of Science]

Simons, F. J., Zielhuis, A. & van der Hilst, R. D. (1999). The deep structure of the Australian continent from surface wave tomography. Lithos 48, 17–43.[CrossRef][Web of Science]

Stacey, F. D. & Loper, D. E. (1983). The thermal boundary-layer interpretation of D'' and its role as a plume source. Physics of the Earth and Planetary Interiors 33, 45–55.[CrossRef][Web of Science]

Stachel, T., Harris, J. W., Brey, G. P. & Joswig, W. (2000). Kankan diamonds (Guinea) II: lower mantle inclusion parageneses. Contributions to Mineralogy and Petrology 140, 16–27.[CrossRef][Web of Science]

Stanistreet, I. G. & Stollhoffen, H. (1999). Onshore equivalents of the Kudu gas reservoir in Namibia. In: Cameron, N. R., Bate, R. H. & Clure, V. S. (eds) The Oil and Gas Habitats of the South Atlantic. Geological Society, London, Special Publications 153, 345–365.

Stewart, K., Turner, S., Kelley, S., Hawkesworth, C., Kirsten, L. & Mantovani, M. (1996). ⁴⁰Ar–³⁹Ar geochronology in the Paraná continental flood basalt province. Earth and Planetary Science Letters 143, 95–109.[CrossRef][Web of Science]

Sun, S.-S. & McDonough, W. F. (1989). Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. In: Saunders, A. D. & Norry, M. J. (eds) Magmatism in the Ocean Basins. Geological Society, London, Special Publications 42, 313–345.

Sun, S.-S., Nesbitt, R. W. & Sharaskin, A. Ya. (1979). Geochemical characteristics of Mid-Ocean Ridge basalts. Earth and Planetary Science Letters 44, 119–138.[CrossRef][Web of Science]

Tang, H.-F. & Liu, C.-Q. (2002). Trace element geochemistry during metamorphic dehydration: a case study from the Xingzi Group of Lushan, southeast China. Geochemical Journal 36, 545–561.[Web of Science]

Taura, H., Yurimoto, H., Kato, T. & Sueno, S. (2001). Trace element partitioning between silicate perovskites and ultracalcic melt. Physics of the Earth and Planetary Interiors 124, 25–32.[CrossRef][Web of Science]

Taylor, S. R. & McLennan, S. M. (1995). The geochemical evolution of the continental crust. Reviews of Geophysics 33, 241–265.[CrossRef][Web of Science]

Teixeira, W., Carneiro, M. A., Noce, C. M., Machado, N., Sato, K. & Taylor, P. N. (1996). Pb, Sr and Nd isotope constraints on the Archaean evolution of gneissic–granitoid complexes in the southern São Francisco craton, Brazil. Precambrian Research 78, 151–164.[CrossRef][Web of Science]

Thompson, R. N., Gibson, S. A., Mitchell, J. G., Dickin, A. P., Leonardos, O. I. H., Brod, J. A. & Greenwood, J. C. (1998). Migrating Cretaceous–Eocene magmatism in the Serra do Mar alkaline Province, SE Brazil: melts from the deflected Trindade mantle plume? Journal of Petrology 39, 1493–1526.[CrossRef][Web of Science]

Thompson, R. N., Gibson, S. A., Dickin, A. P. & Smith, P. M. (2001). Early Cretaceous basalt and picrite dykes of the Southern Etendeka region, NW Namibia: windows into the role of the Tristan mantle plume in Paraná–Etendeka magmatism. Journal of Petrology 42, 2049–2081.[Abstract/Free Full Text]

Toyoda, K., Horiuchi, H. & Tokonami, M. (1994). Dupal anomaly of Brazilian carbonatites: geochemical correlations with hotspots in the South Atlantic and implications for the mantle source. Earth and Planetary Science Letters 126, 315–331.[CrossRef][Web of Science]

Turner, S. & Hawkesworth, C. J. (1995) The nature of the continental mantle lithosphere: constraints from the major element composition of continental flood basalts. Chemical Geology 120, 295–314.[CrossRef][Web of Science]

Turner, S. P., Regelous, M., Kelley, S., Hawkesworth, C. J. & Mantovani, M. (1994). Magmatism and continental break-up in the South Atlantic: high precision ⁴⁰Ar–³⁹Ar geochronology. Earth and Planetary Science Letters 121, 333–348.[CrossRef][Web of Science]

Turner, S. P., Hawkesworth, C. J., Gallagher, K., Stewart, K., Peate, D. W. & Mantovani, M. S. M. (1996). Mantle plumes, flood basalts and thermal models for melt generation beneath continents: assessment of a conductive heating model and application to the Paraná. Journal of Geophysical Research 101, 11503–11518.[CrossRef]

van der Hilst, R. & Seno, T. (1993). Effects of relative plate motion on the deep structure and penetration depth of slabs below the Izu–Bonin and Mariana island arcs. Earth and Planetary Science Letters 120, 395–407.[CrossRef][Web of Science]

van der Hilst, R. D, Engdahl, E. R., Spakman, W. & Nolet, G. (1991). Tomographic imaging of subducted lithosphere below northwest Pacific island arcs. Nature 353, 37–43.[CrossRef]

Weaver, B. L., Wood, D. A., Tarney, J. & Joron, J. L. (1987). Geochemistry of ocean island basalts from the South Atlantic: Ascension, Bouvet, St. Helena and Tristan da Cunha. In: Fitton, J. G. & Upton, B. G. J. (eds) Alkaline Igneous Rocks. Geological Society, London, Special Publications 30, 253–267.

White, R. & McKenzie, D. (1989). Magmatism at rift zones: the generation of volcanic continental margins and flood basalts. Journal of Geophysical Research 94, 7685–7729.

White, R. S., Spence, G. D., Fowler, S. R., McKenzie, D. P., Westbrook, G. K. & Bowen, A. N. (1987). Magmatism at rifted continental margins. Nature 330, 439–444.[CrossRef]

Williams, Q. & Garnero, E. J. (1996). Seismic evidence for partial melt at the base of Earth's mantle. Science 273, 1528–1530.[Abstract]

Woodhead, J. D. & Hergt, J. M. (1997). Application of the double spike technique to Pb-isotope geochronology. Chemical Geology 138, 311–321.[CrossRef][Web of Science]

Woodhead, J. D., Volker, P. & McCulloch, M. T. (1995). Routine lead isotope determination using a ²⁰⁷Pb–²⁰⁴Pb double spike: a long-term assessment of analytical precision and accuracy. Analyst 120, 35–39.[CrossRef]

Zartman, R. E. & Haines, S. M. (1988). The plumbotectonic model for Pb isotopic systematics among major terrestrial reservoirs—a case for bi-directional transport. Geochimica et Cosmochimica Acta 52, 1327–1339.[CrossRef][Web of Science]

Zindler, A. & Hart, S. (1986). Chemical geodynamics. Annual Review of Earth and Planetary Sciences 14, 493–571.[CrossRef][Web of Science]

CiteULike    Connotea    Del.icio.us    What's this?

This article has been cited by other articles:

				R. B. Trumbull, D. L. Reid, C. de Beer, D. van Acken, and R. L. Romer Magmatism and continental breakup at the west margin of southern Africa: A geochemical comparison of dolerite dikes from northwestern Namibia and the Western Cape South African Journal of Geology, September 1, 2007; 110(2-3): 477 - 502. [Abstract] [Full Text] [PDF]

				F. Jourdan, H. Bertrand, U. Scharer, J. Blichert-Toft, G. Feraud, and A. B. Kampunzu 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 J. Petrology, June 1, 2007; 48(6): 1043 - 1077. [Abstract] [Full Text] [PDF]

				R. N. Thompson, A. J. V. Riches, P. M. Antoshechkina, D. G. Pearson, G. M. Nowell, C. J. Ottley, A. P. Dickin, V. L. Hards, A.-K. Nguno, and V. Niku-Paavola Origin of CFB Magmatism: Multi-tiered Intracrustal Picrite-Rhyolite Magmatic Plumbing at Spitzkoppe, Western Namibia, during Early Cretaceous Etendeka Magmatism J. Petrology, June 1, 2007; 48(6): 1119 - 1154. [Abstract] [Full Text] [PDF]

				J. B. Murphy and J. Dostal Continental mafic magmatism of different ages in the same terrane: Constraints on the evolution of an enriched mantle source Geology, April 1, 2007; 35(4): 335 - 338. [Abstract] [Full Text] [PDF]

				P. Comin-Chiaramonti, A. Marzoli, C. de Barros Gomes, A. Milan, C. Riccomini, V. F. Velazquez, M. M.S. Mantovani, P. Renne, C. C. G. Tassinari, and P. M. Vasconcelos The origin of post-Paleozoic magmatism in eastern Paraguay Geological Society of America Special Papers, January 1, 2007; 430(0): 603 - 633. [Abstract] [Full Text] [PDF]

				C. Class and A. P. le Roex Continental material in the shallow oceanic mantle--How does it get there? Geology, March 1, 2006; 34(3): 129 - 132. [Abstract] [Full Text] [PDF]

				T. Furman, J. Bryce, T. Rooney, B. Hanan, G. Yirgu, and D. Ayalew Heads and tails: 30 million years of the Afar plume Geological Society, London, Special Publications, January 1, 2006; 259(1): 95 - 119. [Abstract] [PDF]

				T. R. RILEY, P. T. LEAT, M. L. CURTIS, IAN. L. MILLAR, R. A. DUNCAN, and A. FAZEL Early-Middle Jurassic Dolerite Dykes from Western Dronning Maud Land (Antarctica): Identifying Mantle Sources in the Karoo Large Igneous Province J. Petrology, July 1, 2005; 46(7): 1489 - 1524. [Abstract] [Full Text] [PDF]

This Article Abstract FREE Full Text (PDF) Supplementary data Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (17) Request Permissions Google Scholar Articles by EWART, A. Articles by ARMSTRONG, R. A. Search for Related Content GeoRef GeoRef Citation Social Bookmarking          What's this?

Online ISSN 1460-2415 - Print ISSN 0022-3530

Copyright © 2009 Oxford University Press

Oxford Journals Oxford University Press

• Site Map
• Privacy Policy
• Frequently Asked Questions

Other Oxford University Press sites: Oxford University Press Oxford Journals China Oxford Journals Japan American National Biography Booksellers' Information Service Children's Fiction and Poetry Children's Reference Corporate & Special Sales Dictionaries Dictionary of National Biography Digital Reference English Language Teaching Higher Education Textbooks Humanities International Education Unit Law Medicine Music Online Products Oxford English Dictionary Reference Rights and Permissions Science School Books Social Sciences Very Short Introductions World's Classics