Journal of Petrology Advance Access originally published online on October 1, 2004
Journal of Petrology 2005 46(1):169-190; doi:10.1093/petrology/egh067
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology vol. 46 issue 1 © Oxford University Press 2004; all rights reserved
The Role of Continental Crust and Lithospheric Mantle in the Genesis of Cameroon Volcanic Line Lavas: Constraints from Isotopic Variations in Lavas and Megacrysts from the Biu and Jos Plateaux
1 MAX-PLANCK-INSTITUT FÜR CHEMIE, ABT. GEOCHEMIE, POSTFACH 3060, 55020 MAINZ, GERMANY
2 INSTITUT FÜR MINERALOGIE, SENCKENBERGANLAGE 28, 60054 FRANKFURT/MAIN, GERMANY
RECEIVED AUGUST 26, 2002; ACCEPTED AUGUST 3, 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSAMPLING AND ANALYTICAL...RESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
We present a combined Sr, Nd, Pb and Os isotope study of lavas and associated genetically related megacrysts from the Biu and Jos Plateaux, northern Cameroon Volcanic Line (CVL). Comparison of lavas and megacrysts allows us to distinguish between two contamination paths of the primary magmas. The first is characterized by both increasing 206Pb/204Pb (19·82–20·33) and 87Sr/86Sr (0·70290–0·70310), and decreasing
Nd (7·0–6·0), and involves addition of an enriched sub-continental lithospheric mantle-derived melt. The second contamination path is characterized by decreasing 206Pb/204Pb (19·82–19·03), but also increasing 87Sr/86Sr (0·70290–0·70359), increasing 187Os/188Os (
0·130–0·245) and decreasing Nd (7·0–4·6), and involves addition of up to 8% bulk continental crust. Isotopic systematics of some lavas from the oceanic sector of the CVL also imply the involvement of a continental crustal component. Assuming that the line as a whole shares a common source, we propose that the continental signature seen in the oceanic sector of the CVL is caused by shallow contamination, either by continent-derived sediments or by rafted crustal blocks that became trapped in the oceanic lithosphere during continental breakup in the Mesozoic. KEY WORDS: crustal contamination; CVL; megacrysts; ocean floor; osmium isotopes
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL...RESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The Cameroon Volcanic Line (CVL) comprises a genetically related series of Cenozoic intraplate volcanoes that extend for 1600 km from the island of Annobón (formerly known as Pagalu) in the South Atlantic Ocean to the continental interior of West Africa (Fig. 1). The northern end of the continental part of the CVL is marked by the Cenozoic volcanism of the Biu Plateau, Nigeria. Fitton & Dunlop (1985)
showed that basaltic rocks in the oceanic and continental sectors of the CVL are geochemically and isotopically (87Sr/86Sr) similar and suggested that a line or zone of hot asthenospheric mantle is upwelling underneath the region, partial melting of which has generated parental magmas without any substantial involvement of the overlying lithosphere.
|
This simple picture was challenged by combined Nd, Sr, Pb and O isotope studies of Halliday et al. (1988
, 1990), in which those workers found a distinctive 206Pb/204Pb anomaly in CVL lavas focused at the continent–ocean boundary (c.o.b.), which diminishes over a distance of 400 km to either side. Halliday et al. considered this HIMU Pb isotope signature (high µ 
high 238U/204Pb, leading to time-integrated high 206Pb/204Pb) to be inherited from relatively recent U/Pb fractionation at 125 Ma during impregnation of the uppermost mantle by the St. Helena hotspot when the Equatorial Atlantic opened. The observed Pb isotope heterogeneity of the CVL lavas was therefore proposed to be derived from remelting of variably metasomatized lithosphere rather than reflecting primary asthenospheric source heterogeneity. From a study of peridotite xenoliths Lee et al. (1996) provided evidence that portions of the lithospheric mantle beneath the CVL are isotopically enriched. There is also qualitative evidence for interaction with the continental crust in some evolved lavas of the continental sector based upon large variations in Hf isotopes (Ballentine et al., 1997), 87Sr/86Sr as high as 0·705–0·714 (Marzoli et al., 1999) and the Sr–Nd isotope systematics of lavas and genetically related megacrysts (Rankenburg et al., 2004).
In this study, we examine the respective contributions of crustal contamination and assimilation of sub-continental lithospheric mantle (SCLM) by comparing the isotopic (Sr, Nd, Pb and Os) and trace element variations of Biu and Jos Plateau lavas with the compositions of genetically related megacrysts that grew at mantle depth. We have analysed Sr, Nd and Pb isotopes in 36 whole rocks and 13 megacrysts collected from the Biu and Jos Plateaux, as well as osmium isotopes of a subset of 17 rock samples. The Re–Os isotope system provides an excellent tool for discrimination between assimilation of continental crust or the SCLM. Unlike Sr, Nd and Pb isotope compositions, which may overlap in both continental crust and the SCLM, there is generally a strong contrast in osmium isotopes between the continental crust and the peridotitic SCLM as a result of the compatible behaviour of Os during mantle melting. Whereas continental crust generally has developed variable but high 187Os/188Os ratios over time (e.g. Esser & Turekian, 1993; Esperanca et al., 1997), the SCLM generally has complementary unradiogenic 187Os/188Os ratios (e.g. Walker et al., 1989). Thus, if a melt is contaminated by old crust-derived material, it should have an unusually radiogenic Os isotope signature. In contrast, contaminants derived from the peridotitic SCLM should have unradiogenic Os isotope compositions.
Pyroxenite xenoliths derived from the SCLM may also have a radiogenic Os isotope signature (e.g. Reisberg et al., 1991; Roy-Barman et al., 1996; Lassiter et al., 2000; Pearson & Nowell, 2004). Thus melting of pyroxenite layers or veins in the SCLM has been invoked to explain the ubiquity of elevated Os isotope ratios in ocean island basalt (OIB) (e.g. Hauri & Hart, 1993; Schiano et al., 1997; Lassiter et al., 2000; Hauri, 2002; Kogiso et al., 2004). However, contamination with pyroxenite-derived melts may be distinguished from crustal material based upon other geochemical tracers, such as, for example, Pb isotope and trace element signatures.
Geological setting: the Benue Trough and Biu and Jos Plateaux
The continental sector of the CVL has a Y-shaped form (see Fig. 1). Whereas most previous studies considered the Biu Plateau as the end of the NNW branch of the continental sector of the CVL (e.g. Turner, 1978; Fitton, 1980; Halliday et al., 1988; Poudjom-Djomani et al., 1995; Lee et al., 1996; Ballentine et al., 1997; Barfod et al., 1999; Marzoli et al., 2000), the Jos Plateau (located c. 400 km to the NW of the central CVL axis, see Fig. 1) is usually not assigned to CVL volcanism. However, the timing of the Jos Plateau volcanism is very similar to that of the other CVL volcanic centres (Grant et al., 1972). The Biu and Jos Plateau lavas have similar major and trace element chemistry, and Jos Plateau lavas also span a similar range in isotopic compositions, overlapping the data of the CVL as a whole (Rankenburg et al., 2004). We therefore consider the Jos Plateau to be associated with CVL volcanism in the following discussion.
According to Turner (1978), the Biu Plateau was constructed in three stages during two periods of volcanism: (1) an early fissure type eruption; (2) formation of relatively large tephra ring volcanoes and building up of localized thick lava piles (up to 250 m) in the southern part of the plateau. Lavas of this plateau-building stage range in composition from hy-normative basalt to basanite, with K/Ar ages from 5·35 to 0·84 Ma (Grant et al., 1972; Fitton & Dunlop, 1985). Extensive weathering and laterite formation suggest a hiatus after this episode. (3) Resumption of igneous activity with the formation of over 80 NNW–SSE-aligned cinder cones with similar chemistry to the earlier basalts. A rough estimate of the age of the last magmatic period is <50 ka based on diffusional constraints of He in mantle xenoliths of the CVL (Barfod et al., 1999) and >25 ka based on pollen dating of maar sediments from the Biu Plateau (Salzmann, 2000).
As with the Biu Plateau, volcanic activity on the Jos Plateau occurred in two periods and thus the basalts from this region have been divided into an earlier and a more recent group (McLeod et al., 1971). There are no isotopic age determinations available for the older basalts, but Wright (1976) suggested a Paleocene age, roughly synchronous with Benue Trough folding and uplift. The more recent activity formed a group of 22 cinder cones. Radiometric K–Ar ages (Grant et al., 1972) suggest, unlike on the Biu Plateau, continuous volcanism between 2·1 and 0·9 Ma.
The younger volcanics of both the Biu and Jos Plateaux are characterized by abundant inclusions of mantle xenoliths and megacrysts. The megacryst suites of the Biu and Jos Plateaux were described in detail by Wright (1970) and Frisch & Wright (1971), and comprise chemically homogeneous crystals of clinopyroxene (cpx), garnet (gnt), plagioclase (plag) and ilmenite (ilm) with diameters of up to several centimetres, whereas crystals of olivine (ol), amphibole (amph), spinel (sp), apatite (apa), zircon (zr) and blue corundum (cor) are extremely rare.
|
SAMPLING AND ANALYTICAL TECHNIQUES |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL...RESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Major and trace element data were obtained for 27 volcanic rocks from the younger Biu Plateau suite, four rocks from the older, plateau-building suite of the Biu Plateau, and five rocks from the younger Jos Plateau suites (Table 1). The lavas were first coarsely crushed in steel mortars. Selected chips free of obvious xenocrysts or alteration were then powdered in an agate ring-disc mill. The powders were analysed by X-ray fluorescence spectroscopy (XRF) with a Philips PW 1404 instrument at the University of Frankfurt using Li-borate glass discs for major elements and at the University of Mainz using pressed powder pellets for trace elements. Rock powders were commercially analysed at the University of Goettingen, Germany (all samples) and at the Memorial University of Newfoundland, Canada (subset of 17 samples) by inductively coupled plasma mass spectrometry (ICP-MS) following HF–HNO3 acid dissolution [analytical details have been given by Jenner et al. (1990)
]. A subset of 20 samples was additionally analysed for rare earth element (REE) concentrations by inductively coupled plasma atomic emission spectrometry (ICP-AES) following sinter dissolution at the GeoForschungsZentrum in Potsdam (Zuleger & Erzinger, 1988). Comparison of all the datasets revealed problems of the Goettingen ICP-MS laboratory with respect to accurate determination of high field strength element (HFSE) concentrations, most probably because of precipitation of insoluble fluorides from the sample solution (Yokoyama et al., 1999). XRF data are, therefore, reported in Table 1 for Ba, Sr, Nb, Zr, V, Cr and Ni with errors of <5%. All other data in Table 1 are from ICP-MS and ICP-AES analyses with errors on REE, Cs, Rb, Th, U and Pb of <10%, and on Cs, Lu, Ta and Hf of <18%.
|
Pb isotope analyses were carried out on all 36 lavas and 13 megacrysts (Table 2). The lavas were coarsely crushed, sieved and washed in 1N HCl and distilled water to improve surface quality. Handpicked rock chips (
100 mg of the 0·75–1·5 mm fraction) were leached in hot 6N HCl for 1 h and washed ultrasonically in deionized water before dissolution in HF–HNO3. Lead was extracted from the same sample solution that was used for Sr–Nd isotope analyses by anion exchange in mixed HBr–HNO3 media (Abouchami et al., 2000).
|
For Re–Os analysis,
2 g of clean rock chips were handpicked. Complete avoidance of steel tools (which may have considerable concentrations of Os and Re) is not practically possible during the first steps of coarse rock splitting. We used steel hammers in the field and a hydraulic press made of hardened steel to coarsely crush the rock to chips. To reduce possible metal contamination, rock samples were wrapped in thick plastic foil before crushing in the hydraulic press. If small metal particles are abraded during the crushing process, we expect them to reside in the fine-grained fraction of the crushed rocks or alternatively to be plated onto the surface of the rock chips. For Re–Os isotope analyses we therefore used handpicked rock chips as raw material for the subsequent rock processing in the agate ring mill, and discarded the dust fraction. As an additional test, we prepared a duplicate powder for one rock sample with low Os concentration and highly radiogenic 187Os/188Os (Biu 8). The low Os concentration of this sample makes it particularly susceptible to possible contamination during sample preparation. We thoroughly examined the rock chips used to prepare the duplicate powder individually under a binocular microscope to ensure clean, metal-free surfaces before pulverizing the chips in an agate mill. The 187Os/188Os value for the duplicate analysis is similar to, although somewhat higher than, the original analysis, confirming that the measured radiogenic Os signature is a true sample feature.
For Re–Os analyses, the rock powder was digested in a sealed quartz vessel together with a mixed 185Re/190Os isotope tracer and concentrated HCl–HNO3 (2:3) for 16 h in a Perkin Elmer high-pressure asher operating at 10 MPa N2 overpressure and 300°C (Brügman et al., 1999). Osmium was extracted into liquid bromine and purified by micro-distillation following the method of Birck et al. (1997). Rhenium was separated and purified from the residue using ion exchange extraction (Morgan & Walker, 1989). Os and Re were subsequently loaded onto Pt filaments with a mixed Na(OH)–Ba(OH)2 emitter. The concentrations and isotopic compositions reported in Table 2 were measured at the Max-Planck-Institut, Mainz, by thermal ionization mass spectrometry in negative ion mode (N-TIMS) using a Finnigan MAT-262 system. The effects of fractionation during Os runs were corrected for by normalizing the Os isotope ratios to 192Os/188Os = 3·0827 (Luck & Allègre, 1983). Six procedural blanks for Os ranged from 0·16 pg to 1·45 pg with 187Os/188Os between 0·23 and 0·39, resulting in corrections on sample 187Os/188Os of <2%, and corrections on sample Os concentrations of <2%. Measured Re procedural blanks were 11–47 pg Re, resulting in corrections to Re sample concentrations of up to 20%. The Mainz in-house Os standard yielded 187Os/188Os of 0·10703 ± 20 (2
error, n = 5). The reproducibility of 187Os/188Os ratios based upon duplicate sample dissolutions was <2·7% of the mean, with lower errors associated with higher Os concentrations of the samples, but still significantly worse than the standard reproducibility. This variability may have several reasons. Osmium is a strongly chalcophile element and the Os budget in a sample may be dominated by Os contained in small sulphide globules (Roy-Barman et al., 1998), which may have heterogeneous isotopic compositions. The specific distribution of sulphide globules within aliquots of the sample powder therefore may account for the poor sample reproducibility observed. The same effect may also be responsible for the relatively poor reproducibilities of Os and Re concentrations, which were <24% and 5·4% of the mean. Similar variability in Os concentrations and isotopic compositions in individual flows has previously been reported by, for example, Alves et al. (1999). Minor sample contamination by disaggregated xenoliths may also contribute to sample heterogeneity.
Analytical details and results for major element, trace element and Sr–Nd isotope analyses of the megacrysts have been given by Rankenburg et al. (2004). For Pb isotope analyses of megacrysts (10 cpx and three feldspars; Table 3) 300 mg of cpx or 100 mg of plagioclase, respectively, were handpicked under a binocular microscope in dark and bright field. Grains were then leached twice in hot 2·5N HCl for 20 min, then in cold 5% HF for 15 min in an ultrasonic bath, rinsed with cold 2·5N HCl to remove fluoride complexes and finally rinsed in deionized water. During a microscopic reexamination all grains with visible reaction rims were removed to ensure 100% optically pure separates. Grains were then dissolved in Teflon beakers using HF–HNO3. Samples were run on a Finnigan MAT 261 multicollector TIMS instrument in static mode. All data are reported after fractionation correction of typically 0·116% per a.m.u. as determined by contemporaneous runs of the NBS981 standard. External 2 reproducibility (n = 29) of the standard was 387 ppm, 614 ppm and 861 ppm for 206Pb/204Pb, 207Pb/204Pb and 208Pb/204Pb ratios, respectively. Total procedural lead blanks were <50 pg.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL...RESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Major and trace elements
Most of the samples from the Biu and Jos Plateaux are classified as basalts, and include both nepheline- and hypersthene-normative compositions. The older plateau-building rocks of the Biu Plateau range from basanites to trachybasalts. In contrast, the younger cinder cones have more variable chemistry ranging from basanite to phonolite. Only the younger suites of the Biu and Jos Plateaux contain xenoliths (peridotites, pyroxenites and crustal rocks) and megacrysts. Although the cinder cones are only small volcanic structures that produced <2 km3 lava (Turner, 1978
), rock samples from a single location can be chemically heterogeneous. For example, two samples from a single cone (Dam) are basanite and trachybasalt, respectively. Ilmenite megacrysts from a rock of basaltic trachyandesitic composition (Miringa) contain melt inclusions of a more evolved trachyandesitic composition. This highlights the importance of magma mixing processes in the genesis of the younger rocks.
Typical primitive mantle normalized trace element patterns for the Biu and Jos Plateau samples are shown in Fig. 2. Although all patterns are similar to typical OIB patterns with positive Nb and negative Pb anomalies, they are overall more enriched in incompatible elements than typical OIB [see the St. Helena basalt SH68 (Thirlwall, 1997) for comparison in Fig. 2]. The melt inclusions found in ilmenite megacrysts are highly enriched in incompatible trace elements but depleted in the heavy rare earth elements (HREE), possibly indicating garnet fractionation in the genesis of the melts. Two samples seem to be affected by late-stage alteration because of their unusual low U/Nb (Hilia 2), or low K/Rb (Hizshi), respectively. The phonolite pattern (not shown) has a large negative Ti anomaly pointing to fractionation of Fe–Ti-oxides, whereas the lack of a europium anomaly indicates that plagioclase was not a major fractionating phase in phonolite genesis. Excluding the two altered samples and the phonolite, La/Yb and Zr/Nb ratios are similar for Biu and Jos Plateau lavas and are in the range of 17·6–63·8 and 2·8–4·1, respectively (Table 1). La/Yb and Zr/Nb ratios are comparable with ratios reported for HIMU-type OIB worldwide (e.g. Sun & McDonough 1989; Weaver, 1991; Thirlwall, 1997).
|
Sr, Nd, Pb and Os isotopes of lavas and megacrysts
Figure 3 shows a compilation of Sr–Nd isotope compositions of the more primitive rocks of the CVL along with our own analyses of volcanic rocks and associated megacrysts from the Biu and Jos Plateaux (Rankenburg et al., 2004
; Table 2). We chose an arbitrary cutoff at MgO >5 wt % to exclude the rare continental CVL phonolites and trachytes that fractionated within the continental crust (Marzoli et al., 1999) and to highlight the variations within the more primitive group of the lavas.
|
206Pb/204Pb ratios of older and younger lavas from the Biu Plateau overlap and range from 19·03 to 20·33 with a mean of 19·75 (Fig. 4). 206Pb/204Pb ratios in lavas from the Jos Plateau are more restricted and range from 19·26 to 19·68. 207Pb/204Pb ratios in older and younger Biu lavas range from 15·62 to 15·69. 207Pb/204Pb in Jos lavas are slightly higher for a given 206Pb/204Pb than Biu lavas. 208Pb/204Pb ratios in older and younger Biu lavas range from 39·18 to 40·35, whereas Jos lavas range from 39·30 to 39·64. In contrast, megacrysts from both the Biu and Jos Plateaux range to more radiogenic Pb isotope compositions than their associated host lavas. Biu megacrysts are in the range of 20·05–20·90 in 206Pb/204Pb, 15·71–15·76 in 207Pb/204Pb and 39·77–41·07 in 208Pb/204Pb (n = 10). Jos megacrysts are in the range of 19·82–20·04 in 206Pb/204Pb, 15·67–15·70 in 207Pb/204Pb and 39·56–39·79 in 208Pb/204Pb (n = 3).
|
(= 232Th/238U) to the NHRL, and therefore is different from typical HIMU compositions such as St. Helena, Tubuai or Mangaia. (c, d) Same data as in (a) and (b) but plotted using the delta notation of Hart (1984)In combined plots of 207Pb/204Pb, 208Pb/204Pb,
7/4, 8/4, Nd and 87Sr/86Sr vs 206Pb/204Pb (Figs 4 and 5) whole-rock samples from the Biu and Jos Plateaux fall within the fields for oceanic and continental CVL rocks previously reported by Halliday et al. (1988, 1990), Lee et al. (1994) and Marzoli et al. (1999, 2000). Samples with 206Pb/204Pb >19·8 lie close to the northern hemisphere reference line (NHRL) as defined by Hart (1984). As far as can be concluded from the restricted dataset, megacrysts from a single location cluster on a general trend towards enriched compositions. For example, all four cpx and three plagioclase megacrysts analysed from the Miringa volcano (Biu Plateau) plot within a narrow range only slightly greater than analytical uncertainty. The isotopic trends defined by the megacrysts as a whole overlap with the isotopic compositions of the basalts, but extend to considerably more radiogenic Pb isotope compositions. However, the megacrysts are different from typical HIMU OIB compositions such as St. Helena (SH), Tubuai (T) or Mangaia (M), because of their higher 208Pb/204Pb (Fig. 4b) and 87Sr/86Sr (Fig. 5b) ratios for a given 206Pb/204Pb.
|
Osmium concentrations in the lavas vary from 19 to 350 (mean 125) ppt (Table 2), and are typical of those found in OIB (Shirey & Walker, 1998
). Rhenium concentrations, on the other hand, vary from 64 to 396 (mean 158) ppt and are low when compared with average mid-ocean ridge basalt (MORB) (926 ppt) or OIB (377 ppt) (Righter & Hauri, 1998). Cu/Re ratios range between 1·06 x 105 and 7·96 x 105. Cu/Re ratios higher than the primitive mantle value of 1·07 x 105 (McDonough & Sun, 1995) and low Re concentrations might be related to degassing of subaerial erupted lavas, as suggested by Bennett et al. (2000) and Lassiter (2003) for Hawaiian tholeiites. Re and Os concentrations do not correlate with silicate-compatible elements such as Mg or Ni, or with chalcophile elements such as Co or Cu. This indicates that there is no direct relationship between Os abundance and the degree of sample differentiation. However, Re and Os concentrations are fairly well correlated (r2 = 0·57).
The measured 187Os/188Os ratios of the lavas range from 0·1234 to 0·2450. 187Re/188Os ranges from 3·3 to 27·4. Given the young age of the rocks (5·35–0·84 Ma for older Biu Plateau lavas, other samples <50 ka), age corrections are generally small and within the analytical reproducibility. On a 187Os/188Os vs 1/[Os] diagram, the volcanics do not form a significant correlation. However, low concentrations (<70 ppt) are generally associated with the highest 187Os/188Os ratios (0·1384–0·2450). There is a broad negative correlation between 187Os/188Os ratios and 206Pb/204Pb (Fig. 6), with radiogenic Os associated with unradiogenic Pb compositions. Gannoun et al. (2001) recently found a similar trend in Os–Pb isotope space in CVL lavas from the continental sector. There is no significant trend in Os isotope composition in lavas with 206Pb/204Pb >19·8. 187Os/188Os ratios range between 0·1266 and 0·1400, similar to typical OIB values. The two samples from this subset with the highest 187Os/188Os (0·1384 and 0·1400) are again those with the lowest Os concentrations (48 and 66 ppt, respectively).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL...RESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Origin of megacrysts
To use the composition of the megacrysts as a tracer for the composition of CVL magmas at depth, we have to demonstrate the genetic link between them. Megacrysts have been argued to be either genetically related (e.g. Green & Ringwood, 1967
; Irving, 1974; Irving & Frey, 1984; Liu et al., 1992; Schulze et al., 2001) or xenocrystic phases (e.g. Righter & Carmichael, 1993; Davies et al., 2001) in the past and much of the discussion in the literature is based upon comparison of isotope systematics or trace element concentrations of lavas and megacrysts. However, petrological and petrographic evidence may be better suited to constrain the origin of megacryst suites, because isotopic compositions and trace element budgets of the host lavas may be modified subsequent to megacryst formation. Therefore simple comparison of the isotopic range of megacrysts and host lavas does not provide a suitable means to prove or disprove their cogeneity.
A complete discussion of our model for the genesis of megacrysts in alkaline basalts from the Biu and Jos Plateaux has been given by Rankenburg et al. (2004) and the reader is referred to that work for more petrological detail. In summary, major and trace element covariations of the megacrysts are consistent with their derivation via fractional crystallization from an evolving alkali basaltic liquid. Megacrysts do not represent xenocrystic phases derived from disaggregated peridotite- or pyroxenite-xenoliths, because the latter are more calcic and have significantly more depleted (lherzolite), or more enriched (pyroxenite) trace element patterns for a given mg-number than cpx megacrysts.
Megacrysts could potentially represent phenocryst phases from an earlier magmatic event unrelated to recent CVL volcanism. However, cpx–garnet intergrowths preserve magmatic textures and record magmatic temperatures of 1400°C (Rankenburg et al., 2004). The absence of cooling features, such as recrystallization, diffusional gradients or exsolution lamellae, commonly found in pyroxenite xenoliths rules out the interpretation that the megacrysts were precipitated by an earlier magmatic event and stored in cooler lithosphere for a significant amount of time. Based upon modelling of concentration profiles of fast diffusing species observed in cpx–garnet megacryst intergrowths (Rankenburg et al., 2004), we concluded that the megacrysts were not stored within the SCLM for more than a few hundred years. Therefore, we propose that the megacrysts derive from magmas related to the recent (5 Ma to present) magmatism on the Biu and Jos Plateaux.
Pressure estimates for crystallization of primitive cpx and garnet megacrysts based upon phase relations in alkaline magmas (Bultitude & Green, 1971), the pMelts code (Ghiorso et al., 2002) and cpx–liquid thermobarometry (Putirka et al., 1996) are 1·7–2·3 GPa. Some ilmenite megacrysts intergrown with more evolved cpx compositions contain tiny trachyandesite melt inclusions. Assuming that the melt inclusions and evolved cpx were in equilibrium, we calculate Peq and Teq of 1·36 GPa and 1160°C (Putirka et al., 1996). This estimate is consistent with the temperature calculated from the most evolved cpx–garnet intergrowth, which yields 1100°C. The crust–mantle boundary beneath the Biu and Jos Plateaux is constrained from seismic data to depths of 28–30 ± 6 km or 0·9 ± 0·2 GPa (Poudjom-Djomani et al., 1995). This requires that the megacrysts grew well below the crust–mantle boundary, and we can therefore take the megacrysts as probes of melt evolution within the subcontinental lithospheric mantle.
Although plagioclase is typically considered a low-pressure phase, there is one experiment on a natural amphibolite that produced garnet + cpx + ab-rich plagioclase coexisting with an andesitic melt at 1·8 GPa (Rushmer, 1993). The compositions of experimentally produced crystals (cpx, garnet, plag) in that experiment match those of the megacrysts closely, thus permitting plagioclase precipitation within the mantle. One feature of such high-pressure plagioclase is its low anorthite content, consistent with the observed composition of the plagioclase megacrysts.
Mixing arrays in Sr–Nd–Pb isotope space
In plots of Nd–Pb and Sr–Pb isotopes (Fig. 5) the lavas define triangular fields and can be explained by mixing of three different end-member compositions forming two linear arrays. The junction of both arrays is defined by a cluster of lavas with 206Pb/204Pb 19·82, 207Pb/204Pb 15·64, 208Pb/204Pb 39·53, Nd 7·0 and 87Sr/86Sr 0·70290. We will refer to this composition in the following discussion as component ‘A’. CVL lavas with a similar Sr–Nd–Pb isotope composition have mantle-like 
18O values of 5·5
, whereas samples with higher 87Sr/86Sr extend to higher 18O values (Halliday et al., 1988). La/Yb, Zr/Nb, Ce/Pb and K/U ratios of component ‘A’ lavas closely match the ratios reported for HIMU-type OIB (e.g. Thirlwall, 1997). Because component ‘A’ is prominent in the most primitive samples and is common to both trends, we suggest that component ‘A’ has a sub-lithospheric, plume-like origin.
The second end-member (component ‘B’) is best represented by the most radiogenic cpx megacryst of each plateau. An important observation from Fig. 5 is that the megacrysts extend the mixing array towards component ‘B’ defined by the lavas to considerably higher 206Pb/204Pb, 207Pb/204Pb, 208Pb/204Pb, 87Sr/86Sr and lower Nd. Another observation is that cpx megacrysts from the Biu and Jos Plateaux do not overlap in all isotopic systems. Jos cpx megacrysts are characterized by a lower 206Pb/204Pb for a given Nd and 87Sr/86Sr when compared with Biu megacrysts (Fig. 5).
All rocks from the Biu and Jos Plateaux with 206Pb/204Pb <19·8 seem to fan out to compositions that lie within the field defined by literature data for Nigerian basement rocks [granulites, gneisses, migmatites and granites analysed by Halliday et al. (1988), Dickin et al. (1991) and Dada et al. (1995)]. We will refer to this composition in the following discussion as component ‘CC’ (see Figs 4 and 5). It is important to note that the megacrysts do not plot on this trend. In the following section, we evaluate the nature of the three components ‘A’, ‘B’ and ‘CC’ by comparison of the lava compositions of the Biu and Jos Plateaux with the isotopic compositions of genetically related megacrysts that grew within the lithospheric mantle.
Comparison of megacrysts and host lavas
As pointed out above, the megacrysts in Figs 4 and 5 extend to more radiogenic lead isotope compositions than their host lavas. In principle, this might be related to radiogenic ingrowth after formation of the megacrysts, or alternatively diffusional re-equilibration of the megacrysts with more radiogenic wallrocks. If the more radiogenic signature of the megacrysts was related to radiogenic ingrowth, we would expect a strong correlation of Pb isotope ratios with their parental U/Pb ratios. However, we observe that cpx and plagioclase megacrysts from a single volcano are indistinguishable in Pb isotope composition despite their very different µ values (0 and 8; see Lee et al., 1996). We therefore rule out the possibility that the radiogenic Pb signature of the megacrysts is due to late-stage radiogenic ingrowth.
Diffusional re-equilibration with surrounding radiogenic wallrock could in theory explain the observed offset in lead isotope space. However, given the relatively short time of possible storage in the SCLM, the slow diffusion coefficient of Pb in cpx (Cherniak, 2001) and the large size of the megacrysts (up to several centimetres across), we have to invoke 206Pb/204Pb ratios of the assimilated component that are far outside the range reported for mantle xenoliths from the CVL (Lee et al., 1996). Diffusive re-equilibration of the cpx and plagioclase megacrysts with radiogenic lithospheric mantle therefore cannot account for their higher Pb isotope signatures than the lavas. Because neither radiogenic ingrowth nor diffusive re-equilibration subsequent to megacryst formation can account for the radiogenic Pb isotope signatures of these phases, the Pb isotope compositions of cpx and plagioclase megacrysts most probably reflect the source compositions of the magmas from which they grew. Therefore, we conclude that either the megacrysts grew from melts that had been contaminated by a radiogenic component, or the host lavas were contaminated with an unradiogenic component after megacryst crystallization.
Evidence for assimilation of SCLM in the genesis of lavas from the Biu and Jos Plateaux
There is evidence for a combined assimilation–fractional crystallization (AFC) process in the lavas lying along the trend from component ‘A’ to ‘B’, i.e. in lavas with 206Pb/204Pb >19·8 (Fig. 7). In a plot of Ni concentration vs CaO/Al2O3 ratio (Fig. 7a) we identified two samples with elevated Ni contents. Because these two samples also show high modal abundance of disequilibrium [
(Fe–Mg) <0·3] olivine phenocrysts, we conclude that these two melts have accumulated olivine phenocrysts in a late stage of their history. After correcting the MgO contents of the lavas for olivine accumulation, we see clear correlations between decreasing MgO and increasing 206Pb/204Pb ratios (Fig. 7b) and decreasing Nd, respectively (Fig. 7c). This suggests that the temporal AFC progression proceeded from ‘A’ to ‘B’ and not vice versa.
|
Assimilation and fractional crystallization are almost always coupled because the heat needed to generate an anatectic wallrock melt must be balanced by cooling and the latent heat of crystallization (cumulate formation) of the magma that is being contaminated (e.g. DePaolo, 1981
). More sophisticated energy-constrained AFC calculations (Bohrson & Spera, 2001; Spera & Bohrson, 2001) all yield ratios of anatectic melt added to mass of cumulates formed <1. Although assimilation may not lead to a decrease in MgO (depending on the MgO content of the partial melt), fractional crystallization will, and this is likely to be the dominating process seen in Fig. 7.
The P–T conditions of formation of the Biu and Jos Plateaux megacrysts indicate that the megacrysts formed within the underlying SCLM (Rankenburg et al., 2004). We therefore conclude that component ‘B’ is also derived from the SCLM and not from the continental crust. This assumption is further supported by the Os isotope systematics of the lavas. The SCLM and melts derived therefrom on average should have subchondritic to chondritic 187Os/188Os ratios. Average continental crust, on the other hand, is highly radiogenic [e.g. 187Os/188Os of 1·05 (Peucker-Ehrenbrink & Jahn, 2001)]. Assimilation of SCLM or continental crust therefore should lead to different trends in 187Os/188Os when plotted versus indices of magma differentiation. Figure 6 shows that 187Os/188Os ratios of lavas with 206Pb/204Pb >19·8 scatter around typical OIB values of 0·130. The sample with the highest 206Pb/204Pb (Mir) also shows somewhat elevated 187Os/188Os of 0·1384. However, this sample also has very low Os concentration (48 ppt). Therefore it remains unclear whether Mir defines a general trend towards more radiogenic 187Os/188Os, possibly because of minor incorporation of a HIMU component, or if Mir is affected by small amounts of crustal contamination, too small to be detected by the other isotopic systems. The latter conclusion is supported by the fact that Mir shows small fragments of felsic material in thin section. However, all samples with 206Pb/204Pb >19·8 are within the range of typical OIB basalts and do not extend to the radiogenic values seen in samples with 206Pb/204Pb <19·8 (187Os/188Os up to 0·2450). The lack of a significant trend towards ‘crustal’ 187Os/188Os further argues against a crustal origin for component ‘B’.
Component ‘B’ also does not appear to be a typical HIMU melt, because this component is different from the latter in that it has higher 208Pb/204Pb and 87Sr/86Sr, and lower 187Os/188Os for a given 206Pb/204Pb (Figs 4–6). Lee et al. (1996) reported radiogenic Pb (206Pb/204Pb = 21·0) in a harzburgitic SCLM xenolith from the Biu Plateau. Therefore, we propose that component ‘B’ represents metasomatically enriched SCLM. Plume-derived melts (component ‘A’) assimilated variable quantities of enriched SCLM or SCLM-derived melts or fluids en route to the surface. Megacrysts extend to more contaminated compositions than the magmas because they derive from magmas that ponded in the lithosphere for longer periods, thus allowing greater assimilation of SCLM-derived material. The tiny melt inclusions found in ilmenite megacrysts provide a suitable parental liquid for the most evolved megacrysts in terms of their highly enriched major and trace element chemistry (Rankenburg et al., 2004). However, an isotopic study of the melt inclusions was not possible because of their small size. We previously suggested that the compositions of the melt inclusions are not sampled at the surface as individual lavas because they mix with fresh batches of primitive magma, which trigger ascent and eruption of the magma–megacryst mixture to the surface (Rankenburg et al., 2004).
We have shown in Fig. 5 that cpx megacrysts from the Biu and Jos Plateaux do not overlap in all isotopic systems. However, the similar trace element patterns of primitive cpx megacrysts from the Biu and Jos Plateaux (Rankenburg et al., 2004) suggest a common source magma for both areas. Therefore, the three Jos Plateau cpx megacrysts seem to define a different contamination vector in combined Nd–Sr–Pb isotope space than Biu megacrysts. Assuming that the process of megacryst genesis is similar in both areas, we propose that the lithospheric mantle underlying the Biu Plateau is more radiogenic in Pb isotopes compared with that beneath the Jos Plateau, but similar in Sr–Nd isotopes. Additional analyses of peridotite xenoliths from the Biu and Jos Plateaux are needed to better constrain the regional variations in lithospheric mantle composition beneath the continental sector of the CVL.
Models for the origin of enriched SCLM
Lee et al. (1996) reported enriched isotopic signatures (87Sr/86Sr = 0·704156, Nd = –0·4, 206Pb/204Pb = 21·00) in an SCLM-derived harzburgite xenolith collected on the Biu Plateau. Models for the enrichment of the SCLM as a result of underplated plumes or through metasomatism by asthenosphere-derived melts have been previously proposed by, for example, Ringwood (1982), Hawkesworth et al. (1984), Stein & Hofmann (1992) and Halliday et al. (1995). Halliday et al. (1990) suggested that the high 206Pb/204Pb anomaly focused at the CVL c.o.b. was inherited from relatively recent U/Pb fractionation at 125 Ma during impregnation of the uppermost mantle by the St. Helena hotspot when the Equatorial Atlantic opened. A better estimate of the timing of enrichment of the lithosphere underlying the Biu Plateau might be 147 Ma, based upon the earliest period of magmatic activity in the northern Benue Trough (Coulon et al., 1996). The overall CVL HIMU signature was therefore proposed to be derived from recent radiogenic ingrowth in the lithospheric mantle with variable but high U/Pb.
If the radiogenic Pb isotope signature of the CVL was derived from a lithospheric mantle source that was variably fractionated in U/Pb at 147 Ma, we would expect the present-day samples to lie along a corresponding 147 Ma isochron in 206Pb/204Pb–207Pb/204Pb isotope space. However, we observe that component ‘A’ and the enriched lithospheric component ‘B’, as represented by megacryst compositions, both cluster close to the NHRL, which defines an apparent ‘age’ of 1·77 Ga [using a gradient of 0·1084 and an intercept of 13·491 from Hart (1984)]. The 147 Ma reference line indicated in Fig. 4a has a considerably shallower slope than the NHRL. Internal differentiation of the SCLM at 147 Ma therefore cannot account for the Pb isotope signatures observed in the lavas and megacrysts of the Biu and Jos Plateaux. An alternative model is that the radiogenic Pb isotopes in the sub-CVL SCLM could derive from HIMU-type magmas derived from the fossil St. Helena hotspot, which at the time of opening of the South Atlantic was located in the region occupied by the present c.o.b. However, as already pointed out by Halliday et al. (1990), there is a discrepancy between the isotopic data for typical HIMU and other CVL lavas, in that the former have lower 208Pb/204Pb and 87Sr/86Sr for a given 206Pb/204Pb and do not form an appropriate mixing end-member for formation of the megacryst source magmas.
We propose that the enriched isotopic signature of the SCLM simply represents older (Proterozoic?) lithosphere that was subjected to multiple metasomatic events rather than a single overprint in the Mesozoic. On average, long-term metasomatism produced Pb isotope signatures lying along the NHRL. Sr and Nd isotopes, however, may be more radiogenic compared with typical MORB because of the elevated Rb/Sr and U/Pb and low Sm/Nd ratios of the percolating melts. Regardless of the origin of the radiogenic Pb isotopes in the sub-CVL SCLM, the presence of mantle xenoliths with elevated 206Pb/204Pb indicates that portions of the SCLM beneath the Cameroon Volcanic line do possess isotopic characteristics similar to our proposed component ‘B’. The correlation of Pb and Nd isotopes with MgO in the lavas with 206Pb/204Pb >19·8 strongly suggests that this component is related to assimilation of such material in the SCLM.
Evidence for crustal contamination of lavas from the Biu and Jos Plateaux
Lavas that do not lie on a mixing trend between primary composition ‘A’ and the inferred lithospheric mantle contaminant ‘B’ form broad trends pointing towards a third end-member characterized by high 87Sr/86Sr, low Nd, low 206Pb/204Pb, high 7/4, high 8/4 and high 187Os/188Os (Figs 4–6). Previous publications reported evidence for interaction with the continental crust in some evolved phonolites and trachytes of the continental sector based upon 87Sr/86Sr ratios as high as 0·705–0·714 (Marzoli et al., 1999) and large variations in Hf isotopes (Ballentine et al., 1997). The Biu Plateau lies mainly on granite, gneisses and charnockitic rocks of the Pan-African basement complex of Nigeria, and on the Cretaceous Bima sandstones of the Benue trough in the north and west. Radiometric ages of Nigerian basement rocks have been presented by, for example, Van Breemen et al. (1977), Dickin et al. (1991), Dada (1998) and Kröner et al. (2001). Ages cluster around 3·5, 3·1–3·0, 2·7–2·5, 2·1–1·8 Ga and there is a major tectonometamorphic imprint on most Nigerian basement rocks in Pan-African (600 Ma) times. On average, these rocks are characterized by high 87Sr/86Sr, low Nd, low 206Pb/204Pb, high 7/4 and high 8/4, and therefore provide a suitable candidate for composition ‘CC’. Megacrysts or xenoliths with these compositions have not been observed. We therefore suggest that the trend towards component ‘CC’ observed in the lavas was imposed on the magmas at shallower depths, after formation of the megacrysts. This may be due to assimilation of very shallow SCLM or the continental crust.
Os isotopes provide a means for distinguishing if component ‘CC’ derives from the SCLM or the continental crust. If the SCLM is responsible for the trends observed, we expect a trend towards lower or constant 187Os/188Os with increasing 87Sr/86Sr or decreasing 206Pb/204Pb, respectively. Contamination with continental crust, on the other hand, should lead to significantly higher 187Os/188Os. Os isotopes are generally correlated with 206Pb/204Pb in lavas with 206Pb/204Pb <19·8 (Fig. 6), trending towards higher 187Os/188Os with decreasing 206Pb/204Pb. The 187Os/188Os ratios in lavas falling along the ‘A–CC’ trend extend to ratios much higher than observed in the vast majority of OIB and mantle derived xenoliths. Because Nigerian basement rocks are also characterized by high 7/4 ratios, crustal contamination should also result in increasing 7/4 in the lavas falling along the ‘A–CC’ trend. Figure 4 shows our data (symbols) along with the mean composition of local Pan-African continental crust. The lavas with 206Pb/204Pb <19·8 clearly point towards the composition of continental crust. The trend towards elevated 187Os/188Os and 7/4 with decreasing 206Pb/204Pb strongly suggests that component ‘CC’ derives from assimilation of continental crust.
To quantify the amount of crustal contamination in the lavas, we used the mean composition of Nigerian basement rocks found in the literature (Halliday et al., 1988; Dickin et al., 1991; Dada et al., 1995), which is similar to mean upper continental crust (Taylor & McLennan, 1995), but significantly more enriched in the REE. Sr, Nd, Pb and Os concentrations of the uncontaminated end-member were taken as the mean of type ‘A’ lavas. Mixing can successfully be modelled via bulk-rock assimilation. The results for Pb, Sr and Nd isotopes are given in Fig. 8 and are consistent with assimilation of 8% crust in the most contaminated lavas.
|
There have been only a limited number of Re–Os isotopic studies on lower- and upper-crustal rocks. As yet, there are no 187Os/188Os data available for Nigerian basement rocks. Therefore, modelling of Os isotopes is not as straightforward as for Sr, Nd and Pb isotopes. The continental crust, however, generally has 187Re/188Os ratios ranging between 10 and 100 (e.g. Esser & Turekian, 1993
; Saal et al., 1998; Peucker-Ehrenbrink & Jahn, 2001). For the purposes of this paper, we modelled present-day 187Os/188Os ratios of Nigerian basement rocks using this range of Re/Os ratios and mean ages of the basement rocks ranging from 0·6 to 3·5 Ga (Dada, 1998). As a result we expect present-day 187Os/188Os ratios of the Nigerian basement rocks ranging from 0·225 to 6·1. If we assume a mean age of 2 Ga for the basement rocks and a mean upper-crustal 187Re/188Os ratio for these rocks of 34·5 (Peucker-Ehrenbrink & Jahn, 2001), we calculate a present-day 187Os/188Os ratio of 1·28. A mixing scenario for these three cases is included in Fig. 6. In Pb–Os isotope space the lavas from the Biu and Jos Plateaux with 206Pb/204Pb ratios <19·8 fall within the field of crustal contamination. Most of the samples in Fig. 6 scatter around a mixing trajectory calculated for an assimilant with 187Os/188Os of 1·28. Although mixing calculations including Os isotopes do not allow quantitative estimations of the amount of crustal contamination, the negative correlation of Os and Pb isotopes, with Os isotopes extending to radiogenic values greatly exceeding those found in primitive OIB, strongly supports the hypothesis that lavas with 206Pb/204Pb <19·8 were contaminated by local Pan-African continental crust.
Implications for crustal contamination in the oceanic CVL
The above discussion strongly suggests that crustal contamination has affected some continental CVL lavas, particularly those lavas with 206Pb/204Pb <19·8. However, it has previously been noted that the ranges of Sr, Nd, Hf and Pb isotopes in CVL lavas from the continental and oceanic sectors are similar. This observation has been used to argue against a significant role for crustal contamination in the continental CVL lavas (Fitton & Dunlop, 1985; Halliday et al., 1988; Ballentine et al., 1997). Most lavas from the oceanic portion of the CVL have Pb isotope compositions that plot near the NHRL. However, a subset of samples from Principe and Annobón extend to anomalously high 7/4 and 8/4 values (Fig. 4c and d). If the high 7/4 and 8/4 values (and high 187Os/188Os) in the Biu and Jos Plateau lavas reflect crustal contamination, this raises the question of how such features can be present in samples from the oceanic portion of the CVL.
Historically, the ‘continental’ signature present in lavas from both the continental and oceanic portions of the CVL has been interpreted as reflecting the presence of an EM-type component in the plume source of these lavas. However, most EM-type lavas such as those from the Society and Cook–Austral Islands are characterized by relatively unradiogenic Os isotopes (187Os/188Os 0·135), in contrast to the crust-contaminated Biu and Jos Plateau lavas. It is possible that the mantle source of CVL lavas contains an EM component that is coincidentally similar in composition to the local Pan-African crust and that this component is responsible for the ‘continental’ signature in some oceanic CVL lavas. However, we propose that blocks of Pan-African crust may have become embedded in the oceanic crust during initiation of continental breakup. Consequently, contamination of mantle-derived magmas at shallow levels by continental crust may be responsible for the ‘continental’ signature in lavas from the oceanic portion of the CVL as well as in those from the continental sector.
There is growing evidence that fragments of continental crust may be more prevalent in the ocean basins than previously believed. Fragments of continental crust have been drilled or dredged at several locations along the Mid-Atlantic Ridge (e.g. Bonatti et al., 1996; Belyatsky et al., 1997; Pilot et al., 1998). Schaltegger et al. (2002) have also reported ancient continental zircon xenocrysts in basalts from Iceland and Mauritius, and suggested the presence of either rafted or shallowly subducted continental crust under these islands. In the Indian Ocean, fragments of continental crust have been recovered from portions of the Kerguelen Plateau (Ingle et al., 2002), and the isotopic and trace element compositions of many Kerguelen Plateau basalts appear to record assimilation of continental crust (Frey et al., 2002; Ingle et al., 2003). Thus even in the ocean basins the possibility of continental crustal contamination of lavas cannot be excluded a priori. Not all isotopic variations in OIB necessarily reflect variations in mantle composition.
At present, there is no direct evidence for the presence of continental fragments underneath the islands of Principe or Annobón. Future Os isotope studies of lavas from these islands should provide one test for the crustal contamination model. Gannoun et al. (2001) reported 187Os/188Os values up to 0·1876 in basalts from the oceanic portion of the CVL. If high 187Os/188Os values correlate with high 7/4 and 8/4 values on these islands, as in the case for the Biu and Jos Plateau lavas, this would further strengthen our contention that local crustal contamination is the source of the anomalous Pb isotope signature in some Principe and Annobón lavas. Additional petrographic studies should be used to determine the presence or absence of xenocrystic zircons or other fragments of continental crust in oceanic CVL lavas.
Finally, we note that the refractory ‘LOMU’ component in Mid-Atlantic Ridge basalts described by Douglass et al. (1999) and Douglass & Schilling (2000) is isotopically similar to Pan-African continental crust. Douglass et al. (1999) proposed the ‘LOMU’ component to be delaminated subcontinental lithospheric mantle dispersed into the upper mantle during the breakup of Gondwana. However, the most pronounced ‘LOMU’ compositions sampled in oceanic basalts are found in tholeiites from the Aphanasey Nikitin Rise (Douglass et al., 1999), which were recently proposed by Borisova et al. (2001) to be contaminated via shallow assimilation of continental crust derived from cratonic Gondwanian lithosphere based upon combined major and trace element and isotopic data. We propose that the South Atlantic ‘LOMU’ signature in general may not correspond to refractory delaminated SCLM, but instead reflects shallow assimilation of continental