Journal of Petrology Advance Access originally published online on December 8, 2005
Journal of Petrology 2006 47(3):541-566; doi:10.1093/petrology/egi085
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Overlap of Karoo and Ferrar Magma Types in KwaZulu-Natal, South Africa
1 BRITISH ANTARCTIC SURVEY, HIGH CROSS, MADINGLEY ROAD, CAMBRIDGE, CB3 0ET, UK
2 SCHOOL OF GEOLOGICAL AND COMPUTER SCIENCES, UNIVERSITY OF KWAZULU-NATAL, DURBAN 4041, SOUTH AFRICA
3 COLLEGE OF OCEANIC AND ATMOSPHERIC SCIENCES, OREGON STATE UNIVERSITY, CORVALLIS, OR 97331-5503, USA
4 BRITISH ANTARCTIC SURVEY, c/o NERC ISOTOPE GEOSCIENCES LABORATORY, KEYWORTH, NOTTINGHAM NG12 5GG, UK
5 SCHOOL OF GEOGRAPHY, EARTH AND ENVIRONMENTAL SCIENCES, UNIVERSITY OF BIRMINGHAM, BIRMINGHAM B15 2TT, UK
RECEIVED JULY 16, 2004; ACCEPTED OCTOBER 18, 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONGEOLOGICAL SETTINGGEOCHRONOLOGYWHOLE-ROCK GEOCHEMISTRY OF DYKESDISCUSSIONMAGMA FLOWSUMMARYCONCLUSIONSREFERENCES |
|---|
A suite of mafic dykes from the Underberg region of southern KwaZulu-Natal (South Africa) were intruded at
178 Ma, coincident in age with the major Okavango Dyke Swarm of Botswana, and also coincident with minor Karoo-related intrusions of the northern and central Lebombo. The dykes are all low-Ti–Zr tholeiites, they trend NW–SE and are presumed to continue into the Karoo central area of the Lesotho Highlands. In many respects, the Underberg dykes are similar to the majority of the low-Ti–Zr volcanic and subvolcanic intrusions of the Karoo; however, their 87Sr/86Sr and 
Nd isotope ratios are either ‘Ferrar-like’ (87Sr/86Sr 0·710; Nd < –3) or transitional between Karoo low-Ti–Zr and Ferrar low-Ti magmas. A potential Ferrar source for at least some of the Underberg dykes is supported by anisotropy of magnetic susceptibility analyses of the dyke suite, which demonstrate absolute flow direction from the SE to the NW, consistent with Gondwana reconstructions. The role of crustal contamination and combined fractional crystallization is also demonstrated to have played a key role in the petrogenesis of the Underberg dykes, involving a local upper crust contaminant. However, the composition of the ‘Ferrar-like’ dykes cannot be easily explained by AFC processes, but they do demonstrate that melting of a lithospheric mantle source enriched to a small degree by subduction-derived fluid was also important. KEY WORDS: dyke; basalt; crustal contamination; large igneous province
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGGEOCHRONOLOGYWHOLE-ROCK GEOCHEMISTRY OF DYKESDISCUSSIONMAGMA FLOWSUMMARYCONCLUSIONSREFERENCES |
|---|
Remnants of the Early Jurassic Karoo and Ferrar large igneous provinces (LIPs) are distributed across large parts of southern Africa and East Antarctica (Fig. 1). The geochemistry of the Karoo igneous rocks has been interpreted by a number of workers to indicate either derivation from an enriched lithospheric mantle source (e.g. Erlank, 1984
) or crustal contamination of magmas derived from a lithospheric source. Others have proposed that the Karoo magmas were partial melts of a mantle plume source that were subsequently contaminated by lithospheric mantle components (Cox, 1992; Ellam et al., 1992; Sweeney et al., 1994).
|
The Ferrar province (Fig. 1), although essentially contemporaneous (Riley & Knight, 2001
) with the Karoo volcanic province, is considerably different in chemistry. The erupted rocks of the Ferrar province are characterized by high SiO2, low TiO2, radiogenic 87Sr/86Sr (Kyle, 1980) and negative Nd, which has led several workers (e.g. Antonini et al., 1999) to suggest that processes involving contamination of the magmas by continental crust must have been important (e.g. Faure et al., 1974). Other workers (e.g. Kyle, 1980; Hergt, 2000) maintain that continental crust was not involved in the petrogenesis of Ferrar magmas, supported by their ‘mantle-like’ Os isotope ratios (Molzahn et al., 1996), and that their radiogenic 87Sr/86Sr is a characteristic of their mantle source. Central to the debate concerning the origin of any flood basalt province is establishing whether a mantle plume source existed and determining whether the role of the plume was restricted to conductive heat transfer to the lithosphere, or whether uncontaminated plume-derived magmas were erupted at the surface or intruded at upper crustal levels.
A suite of mafic dykes from the Underberg region of southern KwaZulu-Natal Province of South Africa (Fig. 2) crop out in an area where, based on previous geochemical studies (Elliot et al., 1999), Karoo and Ferrar magmatic provinces may overlap. We present the results of geochemical, geochronological and magma flow studies to constrain the petrogenesis of the dykes and to improve understanding of the relationship between the Karoo and Ferrar flood basalt provinces.
|
|
GEOLOGICAL SETTING |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGGEOCHRONOLOGYWHOLE-ROCK GEOCHEMISTRY OF DYKESDISCUSSIONMAGMA FLOWSUMMARYCONCLUSIONSREFERENCES |
|---|
The intracratonic Karoo sedimentary sequence was deposited, from Permo-Carboniferous to Early Jurassic times, in a basin overlying the Archaean Kaapvaal Craton in the north and on the Proterozoic Namaqua–Natal Belt and Paleozoic Cape Supergroup in the south (Smith, 1990
). The main sedimentary formations, with a total thickness 
6 km, are named, from base to top: Dwyka, Ecca, Beaufort, Molteno, Elliot and Clarens. The Karoo igneous event, with the emplacement of the volcanic Drakensberg Group, marked the end of the Karoo basin evolution. This event consisted in the continental-scale eruption of tholeiitic basalt lavas and the intrusion of numerous dykes and sills of similar composition at 183 ± 1 Ma. Present-day outcrops of Karoo volcanic rocks are concentrated in the central area of the Karoo (predominantly Lesotho), the Lebombo rifted margin and NE Botswana, with local concentrations occurring elsewhere in South Africa, Botswana, Namibia and western Dronning Maud Land (Antarctica) (Fig. 1). Intrusive rocks of the Karoo volcanic province are more widespread, with a vast sill complex (Fig. 1) distributed over much of South Africa and Namibia (Chevallier & Woodford, 1999), and major dyke swarms occurring in Botswana (Okavango and south Botswana) and southern Lebombo (Rooi Rand).
The Karoo volcanic province has been divided into high- and low-Ti–Zr basalts (Cox et al., 1967), with low-Ti–Zr subprovinces dominating overall and high-Ti–Zr basalts widespread in northern Botswana, Zimbabwe (Jones et al., 2001), the Tuli syncline and the northern Lebombo (Duncan et al., 1990; Sweeney et al., 1994). Outside of the Karoo magmatic province, coeval high-Ti basalts crop out in the Falkland Islands (Mitchell et al., 1999) and the Ahlmannryggen region of western Dronning Maud Land (Harris et al., 1990; Riley et al., 2005).
The Ferrar magmatic province extends over 3500 km from the Theron Mountains of Antarctica (Fig. 1) to SE Australia and Tasmania. The volume of the entire province is relatively small (<500 000 km3) compared with other large igneous provinces (typically >106 km3) and a significant part of this volume is contained in the Dufek–Forrestal layered mafic intrusion (Fleming et al., 1997). The basalts of the Ferrar province are entirely of the low-Ti–Zr type and are typically high SiO2 compared with Karoo compositions. The linear outcrop pattern (Fig. 1) of the Ferrar magmatic province is sub-parallel to the proto-Pacific margin of Gondwana, which has led some workers (e.g. Hergt et al., 1991) to attribute the chemistry of the Ferrar magmas to enrichment of their mantle source by subduction-derived fluids. Cox (1988), however, attributed the linear outcrop pattern to a similar-shaped heat source (a hot line). Elliot et al. (1999) suggested the Ferrar magmas originated from a unique focus and that the present-day outcrop is the result of long distance magma transport from this point source. They suggested that magma transport was ultimately controlled by an active rift system initiated in the Early Jurassic and that the point source for the magmas was a thermal anomaly at the Weddell Sea triple junction (Fig. 1c) (Elliot & Fleming, 2000), which was also adjudged to be responsible for the Karoo low-Ti basalts of southern Africa.
The synchroneity of the Ferrar and Karoo provinces is well recognized (e.g. Encarnación et al., 1996; Pálfy & Smith, 2000) as a result of detailed 40Ar–39Ar and U–Pb geochronology. A recent compilation of Karoo–Ferrar high-precision age determinations corrected to a common standard and common monitor age (Riley & Knight, 2001) confirm an overlap between the Karoo and Ferrar provinces, although the Karoo peak of 183 ± 2 Ma is 3 Myr older than the Ferrar peak of 180·3 ± 2·2 Ma.
Underberg dykes
A suite of dykes that trends approximately NW–SE and crops out in the southern KwaZulu-Natal Province of SE Africa (Fig. 2), near the town of Underberg, has received very little attention. This dyke suite extends SE from the Karoo central area of Lesotho toward the coast (Fig. 2). The dykes are fine- to medium-grained dolerites and occur as rubbly outcrops, with coarser-grained weathered centres and finer-grained, sometimes glassy, intact margins. The dykes are typically >8 m width, with three dykes >45 m width, forming a clear bimodal population (Fig. 3a). Across the region, the dykes have a very uniform trend, with a strike orientation in the range 130–140° (Fig. 3b). This is different from the Okavango and southern Botswana dyke swarms, which trend consistently 110–120°. The Underberg dykes intrude sedimentary sequences of the Triassic Beaufort Group and the overlying Stormberg Group (Molteno, Elliot, Clarens formations), which consist of fine- to medium-grained sandstones, shales and mudstones (Turner, 1999). The Beaufort Group is overlain by the Molteno Formation, which consists of coarse, blue–grey sandstones. The overlying Elliot Formation consists of fine-grained red or purple shales and mudstones. The Early Jurassic Clarens Formation lies above the Elliot Formation and is made up of fine-grained cream sandstones.
|
The emplacement style of Karoo intrusions varies with stratigraphic height through the sedimentary sequence. At the base, thin sills dominate in the Dwyka and Ecca groups. Sills and sheets reach their maximum development in thickness and abundance within the Beaufort Group and are frequently transgressive and basin-shaped (Chevallier & Woodford, 1999
). Cross-cutting dykes are also present in the Beaufort Group, but become more prominent in the overlying Molteno, Elliot and Clarens formations, which form the upper part of the sedimentary sequence.
Petrography
The dolerite dykes of the Underberg region are typically fine to medium grained and are feldspar-phyric. Rare phenocrysts of augite and unaltered olivine can also be identified in hand specimen. The plagioclase phenocrysts are set in a groundmass of smaller plagioclase, augite and Fe–Ti oxides. The dykes typically have an intergranular and/or subophitic texture involving euhedral plagioclase laths and anhedral or equant augite crystals. At least two-thirds of the samples contain olivine, which in most cases is altered or corroded, although relatively fresh cores are preserved in some cases and some grains retain their euhedral habit. Olivine replacement minerals include iddingsite, carbonate and serpentine.
Sample numbering
Sample stations all have the prefix of SA. (e.g. SA.13). Samples from the same locality collected for geochemistry have the suffix .1 (e.g. SA.13.1), whereas samples for anisotropy of magnetic susceptibility (AMS) have the suffix .2 (e.g. SA.13.2), or .3, if two margins of the dyke were sampled. The grid reference (latitude–longitude) is the same for all samples from the same locality.
|
GEOCHRONOLOGY |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGGEOCHRONOLOGYWHOLE-ROCK GEOCHEMISTRY OF DYKESDISCUSSIONMAGMA FLOWSUMMARYCONCLUSIONSREFERENCES |
|---|
Previous studies
Karoo volcanic province
Early geochronology studies of the mafic lavas and sills from the Karoo Province largely relied upon whole-rock K–Ar and Rb–Sr data (e.g. Allsopp et al., 1984
; Fitch & Miller, 1984). The K–Ar whole-rock method is now known to produce widely inaccurate results for volcanic rocks that have undergone even low-grade metamorphism (Walker & McDougall, 1982).
More recently, geochronology studies have been carried out using the 40Ar/39Ar and U–Pb methods. Duncan et al. (1997) reported a detailed 40Ar/39Ar incremental heating study of 32 mafic and silicic volcanic rocks (plagioclase mineral separates and whole-rock cores) from South Africa, Namibia and Antarctica. A 2 km lava succession in Lesotho yielded a close grouping of ages, such that the entire section was adjudged to have been erupted within 0·5 Myr at 183 Ma. Basaltic and rhyolitic volcanic rocks from the Lebombo–Mwenezi region have revealed a slightly broader age range, with two rhyolites yielding ages of 178·1 ± 0·6 and 179·7 ± 0·7 Ma and several mafic rocks giving ages between 181·2 ± 1·0 and 184·2 ± 0·6 Ma. Mafic sills, lavas, and dykes have also been analysed from the Marienthal area of Namibia, central and eastern South Africa, and KwaZulu-Natal, all of which yielded ages indistinguishable from the main period of Lesotho and Lebombo volcanism, 182 ± 1 Ma. Recent work by Riley et al. (2004) on the U–Pb (sensitive high-resolution ion microprobe; SHRIMP) geochronology of rhyolites from the Lebombo has improved the dating of this area. This work, combined with the existing 40Ar/39Ar data, indicates that the 12 km succession of volcanic rocks in the Lebombo rift was erupted in 1–2 Myr at 182 Ma.
The Okavango dyke swarm (Fig. 1) has been dated [40Ar/39Ar (whole-rock and plagioclase) geochronology] by Elburg & Goldberg (2000) and Le Gall et al. (2002), who have suggested an emplacement age of 178–179 Ma.
Ferrar magmatic province
Dating of the Ferrar magmatic province of East Antarctica has previously yielded a broad spread of ages (90–308 Ma; Elliot et al., 1985), although a ‘preferred’ age of 180 ± 5 Ma has been advocated as the best estimate (Elliot et al., 1985). Recent, high-precision ages for magmatic rocks of the Ferrar province also demonstrate a short-lived episode of magmatism. Heimann et al. (1994) determined 40Ar/39Ar ages in the range 176·4 ± 0·6 to 177·2 ± 0·5 Ma from different stratigraphical levels within the Kirkpatrick basalts. Very similar ages were reported by Elliot et al. (1999), in the range 174·9 ± 0·5 to 177·4 ± 0·5 Ma, for a number of sills along the Transantarctic Mountains.
Encarnación et al. (1996) used U–Pb geochronology on zircon and baddeleyite to determine the age of dolerite sills from the central Transantarctic Mountains and Victoria Land (Fig. 1), which yielded ages of 183·4 ± 1·4 and 183·8 ± 1·6 Ma, respectively. 40Ar/39Ar geochronology by Fleming et al. (1997) on feldspar and biotite separates from five dolerite sills yielded a range of plateau and total gas ages from 179·4 ± 0·7 to 181·0 ± 0·7 Ma. The age of emplacement of the Dufek Intrusion (Fig. 1) has been determined using U–Pb techniques on zircon separates (Minor & Mukasa, 1997), which gave crystallization ages of 182·7 ± 0·4 and 183·9 ± 0·4 Ma for a felsic dyke from the Dufek Massif and a capping granophyre intrusion, respectively.
Many of the recently published ages from the Karoo and Ferrar provinces were recalculated by Riley & Knight (2001) using a common neutron fluence monitor (MMhb-1) and a common age for that monitor (523·1 ± 2·6 Ma; Renne et al., 1998). Most ages changed very little (<1 Myr) when recalculated, with the exception of the 40Ar/39Ar ages of the Ferrar intrusive and volcanic rocks produced by the Ohio State University laboratory (Heimann et al., 1994; Fleming et al., 1997; Elliot et al., 1999). The published data from these workers yielded ages of 176–177 Ma, using the neutron fluence monitor, MON-4 and an assigned age for MMhb-1 of 513·5 Ma. These ages were recalculated by Riley & Knight (2001) to 180–181 Ma, which makes them consistent with other 40Ar/39Ar ages and U–Pb data.
Underberg dykes
There are no reported ages for the dyke suite near Underberg (Fig. 2), although Fitch & Miller (1984) reported both K–Ar and 40Ar/39Ar ages for several dykes that intrude the Lesotho Formation lavas at Monontsha Pass in NE Lesotho (Fig. 1). Based on dyke strike these dykes may well be a continuation of the Underberg dyke suite and yielded ages in the range 159–210 Ma, with preferred ages quoted at 159, 165 and 193 Ma for three olivine-bearing dykes. Encarnación et al. (1996) have dated (U–Pb) a granophyre from the New Amalfi sheet (Fig. 1) at 183·7 ± 0·6 Ma. The sheet is believed to be related to the Lesotho lavas and is fed by the Elephant's Head dyke, which crosscuts the lowermost lavas of the Lesotho lavas and is subparallel to the Underberg dykes.
This study
Three samples (SA.3.1, SA.7.1, SA.19.1) were selected for 40Ar/39Ar geochronology from across the Underberg region, providing the best possible spread based on the range of geochemical compositions.
Analytical methods
Samples were either 5 mm diameter whole-rock cores or copper-foil wrapped mineral separates, packaged in evacuated quartz vials, and were irradiated in the Oregon State University TRIGA Reactor for 6 h at 1 MW power on May 23, 2003. The neutron flux was measured using standard FCT-3 biotite, 28·03 Ma (Renne et al., 1994). Reactor temperatures can reach 270°C. Additionally, samples were baked at 195°C for 48 h during extraction line pump-down to 10–9 torr.
Depending on sample composition, the incremental heating experiment started in the range 400–600°C and was typically complete by 1400°C, using a Heine low-blank resistance furnace with a Ta–Nb crucible and Mo liner. Each heating step was of 20 min duration with an additional 5 min cooling and continued removal of active gases with St101 Zr–Al and St172 Zr–V–Fe getters.
A MAP 215-50 rare gas mass spectrometer source at 3000 V, and equipped with a Johnston MM1-1SG electron multiplier at 2050 V, was used for analysis. During the 15 min analysis time per mass peak height, data were collected for 10 cycles of masses 35–40 for baselines and peak tops. Data were reduced and age calculations completed using ArArCALC v2.2 software for 40Ar/39Ar geochronology (Koppers, 2002).
Results
The 40Ar/39Ar results are presented in Table 1 and Fig. 4. Plagioclase separated from sample SA.3.1 yielded an excellent mid-temperature, four-step plateau (176·4 ± 1·2 Ma; Fig. 4a) comprising 70% of total gas released, which is the preferred age. There is slight 40Ar loss (first step only) and 39Ar recoil at high temperature. The isochron is concordant at 176·1 ± 4·4 Ma. Plagioclase separated from sample SA.7.1 also yielded an excellent mid- to high-temperature plateau (176·1 ± 1·2 Ma; Fig. 4b) comprising 78% of gas released, which is the preferred age. Again there is slight 40Ar loss (first step only) and 39Ar recoil (step 2). The isochron is concordant at 177·1 ± 1·2 Ma. Whole-rock sample SA.19.1 provided a good low- to mid-temperature, four-step plateau (181·7 ± 0·7 Ma; Fig. 4c) comprising 54% of total gas released. There is evidence of 40Ar loss at low temperature and 39Ar recoil at high temperature. There is a ‘secondary’ plateau (steps 7–9) that gives a mean age of 175·3 ± 1·2 Ma. The corresponding isochron for these high-temperature steps is 172·1 ± 2·2 Ma, and an initial 40Ar/36Ar ratio of 428 (distinct from the atmospheric value of 296). There is therefore a possibility that this dyke contains excess 40Ar, which may be responsible for the older step ages. On this basis, SA.19.1 could be contemporaneous with SA.3.1 and SA.7.1.
|
|
|
WHOLE-ROCK GEOCHEMISTRY OF DYKES |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGGEOCHRONOLOGYWHOLE-ROCK GEOCHEMISTRY OF DYKESDISCUSSIONMAGMA FLOWSUMMARYCONCLUSIONSREFERENCES |
|---|
Analytical techniques
Powders for geochemical analysis were prepared from 2–3 kg of fresh rock. Samples were reduced to pass a 1700 µm sieve using a hardened steel fly press. The powders were produced using an agate Tema-mill. Sr and Nd isotope compositions were measured at the NERC Isotope Geosciences Laboratory (Keyworth, UK) on a Finnigan-MAT 262 mass spectrometer. Rb–Sr and Sm–Nd analysis followed procedures described by Pankhurst & Rapela (1995)
. Sr isotope composition was determined in multidynamic peak-jumping mode. During the period of analysis, 22 analyses of the Sr isotope standard NBS987 gave a value of 0·710259 ± 0·000008 (2
errors). Nd-isotope composition was determined in static collection mode. Twenty-four analyses of the in-house J&M Nd isotope standard gave a value of 0·511196 ± 0·000022 (2 errors); reported 143Nd/144Nd values were normalized to a value of 0·511130 for this standard, equivalent to 0·511864 for La Jolla.
Major and selected trace element whole-rock analysis was by standard X-ray fluorescence (XRF) techniques at the Department of Geology, University of Keele, following the methods described by Floyd (1985). Higher precision trace element abundances were determined by inductively coupled plasma mass spectrometry (ICP-MS) at the University of Durham. The analytical methods, precision, and detection limits have been detailed by Ottley et al. (2003).
Classification
The major and trace element and Sr–Nd isotope data are reported in Table 2. The analysed samples are subalkaline and range in composition from basalt to basaltic andesite (Fig. 5) and on the basis of their CIPW norms (Yoder & Tilley, 1962) they can all be classified as quartz tholeiites (Table 2), although two samples (SA.19.1, SA.20.1) are on the boundary between quartz and olivine tholeiites. When the data for the Underberg dykes are plotted against MgO (wt %) as an index of differentiation (Fig. 6), Ni is strongly correlated with MgO, suggesting olivine control during magmatic differentiation. Al2O3 increases as MgO decreases, suggesting that plagioclase fractionation is not important until MgO contents fall below 7 wt %. Many of the major and trace elements exhibit compositional trends typical of tholeiites, with negative correlations of Fe2O3, TiO2, Zr and Y with MgO (Fig. 6).
|
|
|
In common with other studies of Mesozoic and older flood basalt provinces we primarily use the incompatible high field strength elements (HFSE; Ti, Zr, Y, Nb) as discriminants between magma types. These elements are considered largely immobile during low-temperature alteration processes (e.g. Peate, 1997
) and ratios between them are not significantly modified by moderate amounts of fractional crystallization or susceptible to variations in the degree of partial melting (e.g. Luttinen & Furnes, 2000). Zr can be used as an effective index of differentiation in magmas that do not crystallize zircon. These discriminants are subsequently supported by variations in 87Sr/86Sr, 143Nd/144Nd, rare earth element (REE) and other major and trace elements and their ratios.
The Underberg dykes are all low-Ti–Zr (LTZ) tholeiites with TiO2 <1·5 wt % and Zr <150 ppm (Fig. 7), consistent with Sweeney et al.'s (1994) classification of a Karoo low-Ti magma type. Geochemically they overlap with the majority of low-Ti basalt types of the Karoo, including the once-contiguous Dronning Maud Land (Fig. 1) lavas from Kirwanveggen, Heimefrontfjella and Vestfjella (Harris et al., 1990; Luttinen & Furnes, 2000). They also overlap with the principal field of Ferrar tholeiites (Antonini et al., 1999). The Underberg dykes have SiO2 contents in the range 48·8–54·5 wt % and have typically low Mg numbers (48·9–68·8; mean of 55·4; Table 2).
|
Cr and Ni contents are varied (Cr 125–837 ppm; Ni 49–267 ppm). The Underberg dykes are all light REE (LREE) enriched, with (La/Lu)N ranging from 2·2 to 5·3, and typically have flat middle REE (MREE) to heavy REE (HREE) chondrite-normalized patterns (Fig. 8). Most samples have a distinctive negative Eu-anomaly as a result of plagioclase fractionation. The multi-element variations in Fig. 9 are characterized by distinct troughs at Ta–Nb, P and Ti, when normalized to primitive mantle; however, four samples (SA.11.1, SA.13.1, SA.19.1, SA.20.1; dashed lines in Fig. 9) have flatter multi-element patterns with no marked troughs at Ta–Nb and P.
|
|
The Underberg dykes show a significant range of variation in a diagram of Th/Yb vs Nb/Yb (Fig. 10) where they are plotted relative to the mid-ocean ridge basalt (MORB)–ocean island basalt (OIB) mantle array and compared with a restricted (because of a lack of good Th data) range of Karoo and Ferrar dolerites. The majority of the Underberg dykes plot toward the low-Ti field of the Vestfjella lavas. Four samples (SA.11.1, SA.13.1, SA.19.1, SA.20.1) plot within the MORB–OIB array, which are the same four samples with no marked Nb–Ta anomaly (Fig. 7). The remaining samples trend toward higher Th/Yb values, indicating an increased contribution from continental crust and/or partial melts of subduction-modified lithosphere. Three samples (SA.3.1, SA.5.1, SA.6.1) plot toward even higher Th/Yb values, and have amongst the highest SiO2 contents and are isotopically ‘Ferrar-like’. They plot close to the Ferrar field and also the field of average global subducting sediment (GLOSS).
|
Therefore, central to the discussion regarding the petrogenesis of the Underberg dyke suite is resolving the contributions from the continental crust and a lithospheric mantle source enriched by subduction-derived fluids.
The Underberg dykes display significant variations in both 87Sr/86Sr and Nd (at 180 Ma). 87Sr/86Sr varies from 0·7044 to 0·7090, whereas Ndi varies from –3·8 to 1·1 (Fig. 11). The two samples with positive Ndi values (SA.11.1 and SA.13.1) also have the least radiogenic initial 87Sr/86Sr values (0·7045) and plot within the OIB array (Fig. 10). The two other samples which plot in the OIB array (Fig. 10) have very similar 87Sr/86Sr values (0·7044–0·7046), but have marginally negative Ndi values (–0·7 to –0·8). Rocks with positive Nd values are rare in the Karoo volcanic province and are absent from the Ferrar magmatic province. Dykes of the Rooi Rand (Duncan et al., 1990), central Lebombo (Sweeney et al., 1994), Falkland Islands (east–west trending; Mitchell et al., 1999), Ahlmannryggen, Dronning Maud Land (Riley et al., 2005) and rare rocks from Vestfjella (Luttinen & Furnes, 2000) and Kirwanveggen (Harris et al., 1990) also have positive Nd, with the Ahlmannryggen Group 3 dykes (Riley et al., 2005) being the most depleted (MORB-like with Ndi +9) in the entire Karoo volcanic province.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGGEOCHRONOLOGYWHOLE-ROCK GEOCHEMISTRY OF DYKESDISCUSSIONMAGMA FLOWSUMMARYCONCLUSIONSREFERENCES |
|---|
The key questions concerning the petrogenesis of almost all continental flood basalt (CFB) magma types are centred on the relative contributions from sub-lithospheric mantle sources (mantle plume, ambient asthenosphere), lithospheric mantle and continental crust. Low-Ti–Zr tholeiites are the predominant rock type throughout the Karoo volcanic province (e.g. Sweeney et al., 1994
; Marsh et al., 1997). These are isotopically and elementally distinct from oceanic basalt compositions, which has led many workers to highlight the important role of the subcontinental lithospheric mantle (SCLM) in Karoo magma petrogenesis (e.g. Hawkesworth et al., 1984, 1999; Ellam & Cox, 1991; Saunders et al., 1992; Sweeney et al., 1994).
The Ferrar magmatic province is geochemically distinct (high SiO2, high 87Sr/86Sr, low Nd) from the Karoo volcanic province, which has led to the interpretation that the Ferrar magmas either were extensively contaminated by the continental crust (Faure et al., 1982; Antonini et al., 1999) or were derived from an enriched lithospheric mantle source (Kyle, 1980; Hergt, 2000), where the enrichment was related to earlier subduction along the Gondwana margin. Although the Karoo and Ferrar magmatic provinces are broadly contemporaneous (Pálfy & Smith, 2000), they are generally regarded as being distinct provinces derived from separate magma sources. However, Elliot & Fleming (2000) have identified Ferrar-like lavas (Golden Gate) in the centre of the Karoo volcanic province, which led to the inevitable conclusion that the two provinces may overlap.
The multi-element patterns in Fig. 9 include four samples with relatively flat trends (dashed lines) compared with the remaining dykes, which are characterized by pronounced negative anomalies in Nb, Ta, P and Ti, typical of arc-related basalts. Puffer (2001) noted that some Karoo basalts (and other continental flood basalts) have arc-like trace element abundances, which can be attributed to the melting of pre-existing subduction-modified mantle.
Analysis of the geochemical and isotopic data indicates that the petrogenetic issues central to understanding the Underberg dykes are the role of crustal contamination, the modification of the lithospheric mantle by subduction-derived fluids and links to the Ferrar magmatic province.
Role of crustal contamination
Interaction between an ascending mafic magma and the surrounding continental crust is inevitable at some point prior to eruption or intrusion, although in some cases the chemical changes to the magma may be insignificant. Any attempts to model the effects of crustal contamination have to consider the varied composition of the local crust and the composition of the uncontaminated parent magma.
Potential proxies for the upper crust in the KwaZulu-Natal Province include the granitic basement of the regionally extensive Mesoproterozoic Namaqua–Natal Belt or hornfels of the Ecca Group. Granitic basement of the Namaqua–Natal Belt has been sampled from deep boreholes (2·5 km). The granitic basement has negative Nd values (–4·4) and strongly radiogenic 87Sr/86Sr (0·7618) at 180 Ma (Eglington & Armstrong, 2003), and Ecca Group hornfels from the Insizwa intrusion has similar Nd values (–4·3) and 87Sr/86Sr values of 0·710 at 180 Ma (Lightfoot & Naldrett, 1984).
Potential primary magma compositions could include the most isotopically depleted rocks of the Underberg suite (e.g. SA.13.1; Table 2), which are akin to the Rooi Rand dykes of the central Lebombo (Duncan et al., 1990), considered to be amongst the most depleted rocks in the Karoo of South Africa. The most depleted rocks in the entire Karoo province are from the Dronning Maud Land region of Antarctica (Riley et al., 2005). They have Nd values of +9, 87Sr/86Sr values close to 0·7035, and could also form a potential primary composition.
Assimilation with fractional crystallization (AFC)
There is, in part, a weak correlation between 87Sr/86Sri and MgO (Fig. 6), suggesting that combined assimilation and fractional crystallization (AFC) could have affected some of the Underberg dykes. However, much of the variation in 87Sr/86Sr (0·7045–0·7075) occurs at fairly uniform MgO (6–6·6 wt %; Fig. 6), indicating that AFC was not the primary process ressponsible for the geochemical variation in the Underberg dykes; however, the remainder of the dykes do show co-variation in MgO and 87Sr/86Sr (Fig. 6), which could be accounted for by AFC processes. AFC was modelled for the Underberg dykes using the energy-constrained recharge–assimilation–fractional crystallization (EC-RAFC) model of Spera & Bohrson (2004), as well as the AFC equations of De Paolo (1981). EC-RAFC differs from standard AFC models in that it tracks the isotopic and trace element composition of the melt, cumulate, country rock partial melts and enclaves during simultaneous recharge, assimilation and fractional crystallization (Fowler et al., 2004).
The depleted composition from Dronning Maud Land (Riley et al., 2005) is an unsuitable end-member for AFC models of the Underberg dyke suite, given its very high Nd value (+9). However, the most depleted rocks of the Underberg suite have Ndi values of +1 and 87Sr/86Sri values of 0·7045, and provide a more feasible end-member for the remainder of the Underberg suite; this is akin to the composition of the Rooi Rand dykes.
Using both Namaqua–Natal Belt granitoid and the Ecca Group hornfels as potential crustal contaminants, AFC processes were modelled using an Underberg depleted composition as the magma end-member (Fig. 11). Employing the AFC model of De Paolo (1981), the Namaqua–Natal granitoid is an unsuitable contaminant composition (Fig. 11); however, the Ecca Group hornfels provides a more suitable contaminant composition to fit the Underberg dyke suite. Using the geologically reasonable parameters set out in Table 3, several of the Underberg dykes could reasonably be interpreted as the products of AFC involving a hornfels contaminant, although the levels of contamination using the De Paolo model are probably too high. Employing the EC-RAFC model of Spera & Bohrson (2004), and using the Ecca Group hornfels as a contaminant composition, generates an appropriate trend for the Underberg dyke suite with geologically reasonable contamination values. The EC-RAFC model was run with an initial magma liquidus temperature estimate of 1210°C and a temperature interval of 7·5°C (full details are given in Table 4). The samples with the most negative Nd values (<–3) are not well represented given an equilibration temperature of 1000°C. However, in general terms, an AFC model involving the most depleted Underberg composition and a local contaminant could generate some of the variation observed in the Underberg suite.
|
|
Lithospheric mantle enrichment
Mantle metasomatism is believed to be the key process responsible for the enrichment of the lithospheric mantle (e.g. Menzies & Hawkesworth, 1987
). The lithospheric mantle beneath the Archaean craton of southern Africa has been extensively modally metasomatized, evidenced from a detailed study of peridotite xenoliths (Erlank et al., 1987).
A key process, which has been suggested in the petrogenesis of many CFB provinces, is the contamination of ‘primary’ basaltic magmas with partial melts of enriched lithospheric mantle (e.g. Ellam & Cox, 1991; Luttinen & Furnes, 2000), with contamination taking place before the magmas reach crustal level. As discussed above, the trace element patterns of many of the Underberg dykes (and other Karoo rocks) are characterized by negative anomalies in the HFSE, typical of arc-related basalts. As there was no active continental margin at the time of Karoo magmatism, the arc-like signature must be a remnant of pre-existing subduction. To categorize the amount of subduction-derived fluid added to the mantle, as well as the degree of partial melting, it is helpful to examine incompatible trace element ratios, such as La/Yb and Ba/Nb (Guo et al., 2005). In addition to being reliable indicators of partial melting and subduction fluid addition, the ratios La/Yb and Ba/Nb are not affected by small amounts of fractional crystallization of olivine and clinopyroxene. Guo et al. (2005) developed a mixing and partial melting model for these trace elements to simulate the petrogenesis of potassic magmas from SE Tibet. The model assumes that the magmas were generated in two stages: (1) mixing of depleted mantle and subduction-derived fluid; (2) partial melting of the resultant enriched mantle. The starting composition is taken as N-MORB source mantle, which has trace element abundances equivalent to 0·1 times N-MORB (Sun & McDonough, 1989), as MORB magmas are 10% partial melts of N-MORB source mantle. The composition of the subduction-derived fluid is less straightforward, as the abundances of La, Yb, Ba and Nb vary significantly (e.g. Tatsumi & Hanyu, 2003); therefore Guo et al. (2005) used two different fluid compositions in their models, one with high abundances (Fluid A), the other with low abundances (Fluid B) (Fig. 12). The mixing model between depleted mantle and the two fluid compositions are shown in Fig. 12 for non-modal batch partial melting models and the Underberg dykes are plotted. The Underberg dykes have low values of both Ba/Nb and La/Yb and plot close to the origin of the mixing curves of Guo et al. (2005). The mixing model using the Fluid A composition (high concentrations of La, Yb, Nb, Ba) indicates that the Underberg dykes have a very low content of the subduction-derived fluid, typically <0·2% (Fig. 12a). The mixing model using the Fluid B composition (low concentrations of La, Yb, Nb, Ba) shows that the Underberg dykes also have a relatively small contribution from the subduction-derived fluid, typically <10% (Fig. 12b), although clearly much greater than the Fluid A model. The non-modal partial melting model indicates a degree of mantle partial melting of 15%. The samples that show the greatest contribution from a subduction-derived fluid are those with Ba/Nb values >60 and include the samples SA.3.1, SA.5.1 and SA.12.1, which also have amongst the highest 87Sr/86Sri values (0·7075–0·7100) and are ‘Ferrar-like’ in composition.
|
Links to Ferrar magmatism
Karoo and Ferrar tholeiites are considered geochemically distinct and geographically separate, with overlap known only in a few areas of pre-break-up Gondwana. Brewer et al. (1992)
identified both Karoo- and Ferrar-like rocks in the Theron Mountains of Coats Land, Antarctica (Fig. 1), although they are chronologically distinct (193 Ma for Ferrar; 176 Ma for Karoo). Other workers have also speculated that Ferrar-like magmatic rocks might extend beyond the accepted limits of the Ferrar province (Fig. 1). Elliot & Fleming (2000) noted that the low-Ti, Golden Gate lavas of northern Lesotho (Fig. 1) are geochemically similar to the low-Ti tholeiites of the Ferrar province. The Golden Gate lavas have similar 87Sr/86Sr and Nd isotope ratios to the Ferrar intrusions (Fig. 11), which are distinctive when compared with other Karoo volcanic rocks. Elliot & Fleming (2000) concluded that the Golden Gate lavas were derived from the same source as the Ferrar, and suggested the interfingering of Ferrar-derived lavas with the lower units of the Karoo flood basalt sequence in Lesotho.
A geochemical study of the Mesozoic dolerite dykes of the Falkland Islands by Mitchell et al. (1999) also identified Karoo- and Ferrar-like rock types. An east–west-trending dyke group has compositional similarities (Fig. 11) to the Rooi Rand dyke swarm of the Karoo (Duncan et al., 1990), whereas a north–south-trending dyke group is geochemically similar to Ferrar magmatic rocks (Fig. 11). Mitchell et al. (1999) also noted that rare samples of the north–south-trending suite are geochemically akin to the Kraai River volcanic rocks of southern Lesotho.
Several of the Underberg dykes (SA.3.1, SA.5.1, SA.6.1) have 87Sr/86Sr (0·7090–0·7100) and Nd (–3·8 to –14·5) values close to those of the Ferrar magmas (Fig. 11), although the sample with Nd = –14·5 is more akin to the low-Nd rocks of CT1 (Vestfjella; Luttinen & Furnes, 2000) and the basalts of the southern Lebombo (Sweeney et al., 1994), and may reflect the incorporation of Precambrian LREE-enriched lithospheric material. The similarity to Ferrar rocks is also highlighted in Fig. 10, which shows that these three samples plot close to the Ferrar field and are distinct from the other Underberg dykes. A similar pattern is observed in Fig. 13, where Nb/Nb* is plotted against 87Sr/86Sri. The Underberg dykes are plotted relative to a restricted (because of the absence of published Th data) range of Karoo and Ferrar rock types. The variation in Nb/Nb* relative to 87Sr/86Sri shows a trend towards lower Nb/Nb* values, indicating an increased contribution from the continental crust or melts from subduction-modified lithospheric mantle. The samples with the lowest Nb/Nb* and highest 87Sr/86Sri overlap with the field of Ferrar rock types.
|
(ThN x LaN)] vs 87Sr/86Sr. The Nb anomaly can be used to test the role of sediment contamination in mantle-derived magmas. The DML Groups 2 and 3 field are from Riley et al. (2005)The presence of Ferrar-like rocks in the Underberg region of southern KwaZulu-Natal, coupled with the ‘Ferrar-like’ Golden Gate lavas (Elliot & Fleming, 2000
), suggests that there is a clear Ferrar component within the Karoo volcanic province. The composition of lava sequences from Moshesh's Ford and Kraai River, as well as several of the dykes from the Underberg suite, are intermediate between a Rooi Rand-like composition and a Ferrar-like composition, which prompts the suggestion that some of the variability could be explained by mixing of Ferrar and Karoo magma compositions. To test the possibility of Ferrar-sourced magmas being transported into the Karoo province, samples were selected for AMS analysis to determine the magma flow direction
|
MAGMA FLOW |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL SETTINGGEOCHRONOLOGYWHOLE-ROCK GEOCHEMISTRY OF DYKESDISCUSSIONMAGMA FLOWSUMMARYCONCLUSIONSREFERENCES |
|---|
Anisotropy of magnetic susceptibility (AMS) data
Background
Anisotropy of magnetic susceptibility (AMS) measurements have been shown to be a very sensitive technique for determining the flow direction of mafic magmas (e.g. Knight & Walker, 1988
; Ernst & Baragar, 1992; Cañón-Tapia et al., 1996; Liss et al., 2002). With this aim, we collected oriented block samples from the margins of well-exposed dykes where the dyke–wall-rock contact was observed and surface weathering minimal. Of the dykes sampled for geochemistry only six possessed two exposed margins. Block samples were oriented in the field using a magnetic compass, and later drilled in the laboratory. The retrieved cores were measured on an AGICO KLY3 Kappabridge at the School of Geography, Earth and Environmental Sciences (University of Birmingham) to determine the three principal axes of magnetic susceptibility kmax, kint, kmin. These principal axes define an ellipsoid of magnetic susceptibility, the shape of which is described by the magnetic lineation [L = (kmax – kmin)/kmean] and foliation [F = (kint – kmin)/kmean]. The shape of the ellipsoid is indicated by the angle µ = tan–1 L/F (the angle from the F axis to a point on a plot of L vs F, ranging from 0° for purely oblate to 90° for purely prolate), whereas the strength of the anisotropy is indicated by the total anisotropy (H = L + F). AMS data together with dyke orientation and width data are presented in Table 4.
AMS fabrics
The AMS fabrics predominantly fall into two distinct types, as follows.
(1) Type A fabrics are characterized by kmin axes clustering approximately normal to the dyke wall, and kmax and kint forming a magnetic foliation sub-parallel to the intrusion plane (e.g. Fig. 14, sample SA.1.3). Such ‘normal’ AMS fabrics are interpreted to be the product of magma flow, with kmax parallel to flow (e.g. Knight & Walker, 1988; Ernst & Baragar, 1992; Rochette et al., 1992). Normal magnetic fabrics account for 55% of the AMS fabrics measured in this study.
|
Type B fabrics are characterized by a cluster of kint axes plotting approximately normal to the intrusion plane (e.g. Fig. 14, sample SA.14.3). Such ‘intermediate fabrics’ (Rochette et al., 1999
) have been variously related to the mixing of differing magnetite grain sizes (Rochette et al., 1992), magmatic shear and associated rotation of ellipsoidal magnetite grains (Dragoni et al., 1997), or the vertical compaction of a static magma column (Park et al., 1988). In addition to the normal and intermediate fabrics, two samples (SA.8.2 and SA.8.3) display abnormal AMS fabrics that are characterized by strongly prolate or oblate fabrics, where only one principal axis shows a good clustering of data, with the other two axes being widely distributed. These fabrics are denoted type C in Fig. 14 (samples SA.8.2 and SA.8.3).
Discussion and interpretation of AMS fabrics within the Underberg dyke suite
Interpretation of magma flow direction from the AMS fabrics obtained from the Underberg dykes is complicated by the presence of both normal and intermediate magnetic fabrics within the same dyke. However, a similar distribution of normal and intermediate fabrics was encountered in dykes within the Koolau complex, Hawaii (Knight & Walker, 1988), where both normal and intermediate fabrics displayed similarly oriented kmax axes, the orientations of which were sub-parallel to observed macroscopic magma flow indicators along the dyke margins. Knight & Walker (1988) concluded that the orientation of kmax in most of the dykes approximated the magma flow direction. This combination of AMS and field data indicates that intermediate fabrics can represent a simple switch between the kint and kmin axes, whereas kmax (the magnetic lineation) remains parallel to magma flow, as predicted by numerical modelling of sheared ellipsoidal grains (Dragoni et al., 1997) or some models for the proportional mixing of single-domain and multi-domain m