| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology | Volume 43 | Number 6 | Pages 981-1001 | 2002
© Oxford University Press 2002
Lamproites from Gaussberg, Antarctica: Possible Transition Zone Melts of Archaean Subducted Sediments
ADVANCED CENTRE FOR QUEENSLAND UNIVERSITY ISOTOPE RESEARCH EXCELLENCE (ACQUIRE), DEPARTMENT OF EARTH SCIENCES, STEELE BUILDING, ST. LUCIA CAMPUS, UNIVERSITY OF QUEENSLAND, ST. LUCIA, BRISBANE, QLD. 4072, AUSTRALIA
Received May 22, 2001; Revised typescript accepted December 17, 2001
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONPetrogenetic models for the origin of lamproites are evaluated using new major element, trace element, and Sr, Nd, and Pb isotope data for Holocene lamproites from the Gaussberg volcano in the East Antarctic Shield. Gaussberg lamproites exhibit very unusual Pb isotope compositions (206Pb/204Pb = 17·44–17·55 and 207Pb/204Pb = 15·56–15·63), which in common Pb isotope space plot above mantle evolution lines and to the left of the meteorite isochron. Combined with very unradiogenic Nd, such compositions are shown to be inconsistent with an origin by melting of sub-continental lithospheric mantle. Instead, a model is proposed in which late Archaean continent-derived sediment is subducted as K-hollandite and other ultra-high-pressure phases and sequestered in the Transition Zone (or lower mantle) where it is effectively isolated for 2–3 Gyr. The high 207Pb/204Pb ratio is thus inherited from ancient continent-derived sediment, and the relatively low 206Pb/204Pb ratio is the result of a single stage of U/Pb fractionation by subduction-related U loss during slab dehydration. Sr and Nd isotope ratios, and trace element characteristics (e.g. Nb/Ta ratios) are consistent with sediment subduction and dehydration-related fractionation. Similar models that use variable time of isolation of subducted sediment can be derived for all lamproites. Our interpretation of lamproite sources has important implications for ocean island basalt petrogenesis as well as the preservation of geochemically anomalous reservoirs in the mantle.
KEY WORDS: lamproites; Pb isotopes; mantle Transition Zone; subducted sediment; anomalous mantle reservoirs
|
INTRODUCTION |
|---|
Lamproites are ultrapotassic mantle-derived volcanic rocks with low CaO, Al2O3 and Na2O contents, high K2O/Al2O3, relatively high Mg content, and extreme enrichment in incompatible elements (Foley et al., 1987
). They are geographically widespread, yet volumetrically insignificant. Unlike kimberlites, which are found mainly in Archaean cratons, lamproites intrude continental crust of varying age (e.g. Graham et al., 1999; Turner et al., 1999), ranging from Archaean (Western Australia) to Palaeozoic (Southern Spain). The oldest known (diamond-bearing) lamproites are of Proterozoic age and occur in Western Australia and India (Pidgeon et al., 1988; Rao et al., 1999). Lamproite magmatism appears to be more common in the Phanerozoic than in the Precambrian, even when the lower preservation potential of ancient examples is taken into account (Mitchell & Bergman, 1991). Gaussberg in Antarctica (56 ± 5 ka) is the youngest lamproite occurrence yet recognized on the Earth (Tingey et al., 1983).
Lamproites, kimberlites, and carbonatites are interpreted to have been derived from geochemically anomalous mantle whose mineralogy, chemistry, and temporal evolution is a matter of continuing debate (e.g. Mitchell & Bergman, 1991; Ringwood et al., 1992; Collerson et al., 2000). The level of incompatible element enrichment in lamproites precludes significant contamination by continental crust (Collerson & McCulloch, 1983), thus they are excellent isotopic proxies for their mantle source. It is generally assumed from Sr, Nd, and Pb isotope compositions that lamproites, kimberlites, and carbonatites form by melting of portions of the sub-continental lithospheric mantle that were enriched by metasomatism and contain phlogopite, clinopyroxene, or K-richterite (Mitchell & Bergman, 1991; Mitchell, 1995; Edgar & Mitchell, 1997). However, the occurrence of ultra-high-pressure syngenetic mineral inclusions in diamonds from kimberlites in China, Brazil and Africa (Wang & Sueno, 1996; McCammon et al., 1997; Stachel et al., 2000), and ultradeep majorite-bearing xenoliths in ‘alnöites’ from Malaita, Solomon Islands (Collerson et al., 2000) suggest that some highly enriched melts may originate in the deep mantle (Ringwood et al., 1992). Furthermore, the presence of primordial noble gas ratios in carbonatites (Sasada et al., 1997; Marty et al., 1998; Dauphas & Marty, 1999) is not consistent with a sub-continental lithospheric mantle origin for these melts.
Continental-derived sediment, which may be subducted along with oceanic lithosphere, is inherently K-rich and contains high-pressure phases such as K-hollandite, which could provide a source component for lamproitic melts. Although most subducted sediment is probably mixed into the convecting mid-ocean ridge basalt (MORB) source mantle, tomographic images (e.g. Van der Hilst et al., 1991; Christensen, 1996; Simons et al., 2000) show that oceanic lithosphere commonly descends to the Transition Zone, and in some subduction zones even penetrates the lower mantle (Van der Hilst et al., 1991; Christensen, 1996). Subduction of oceanic lithosphere has been an integral process of the Earth’s convecting system since the Archaean and it has been estimated that approximately 40 times the volume of the present-day oceanic crust may have been recycled into the mantle over time (Kramers & Tolstikhin, 1997; Kamber & Collerson, 2000). As a result, it is plausible that a considerable amount of subducted material, including continent-derived sediment, could be stored in the Transition Zone and the lower mantle.
This paper, expanding on earlier research on Gaussberg by Collerson & McCulloch (1983) and Williams et al. (1992), uses new major element, trace element, and Sr–Nd–Pb isotope data for an extensive suite of lamproite samples from Gaussberg (collected by K.D.C. during the 1997 Austral summer) to develop a new model for the formation of lamproite magmas.
|
GEOLOGICAL BACKGROUND, SAMPLES, AND PETROGRAPHY |
|---|
Gaussberg is a 370 m high volcanic feature, situated at 67°S, 89°E on the East Antarctic Shield (Fig. 1). The physical, volcanological, and lithological variations in the volcano were mapped by K.D.C. during a 1997 Austral summer expedition to Gaussberg. Several phases of lamproitic activity were recognized. In particular, it was shown that the present
300 m high volcanic construct was built on an earlier eroded volcanic feature. Field observation showed that the volcano erupted sub-glacially, producing pillow lava flows as well as hyaloclastite deposits. Constant pillow flow directions on all points of the volcano indicate that the current cone shape is an artefact of glacial erosion. Thus the original volcano must have been much larger. This conclusion is also supported by the observation that moraine within the adjacent Philippi Glacier is dominated by lamproite fragments. Samples were collected from all representative localities ranging from the base of the volcanic feature to its summit. Details of sample locations are shown in Fig. 2.
|
|
|
The lamproite lavas from Gaussberg range in texture from almost aphyric to hypohaline and are commonly highly vesicular. Microphenocrysts are dominated by subhedral leucite, olivine, and clinopyroxene ranging in size up to 1 mm, with phlogopite in some samples. These sit in a yellow brown glassy matrix containing variable proportions of quench crystals of very fine-grained leucite, phlogopite, apatite, and ilmenite.
Exposed rocks at Gaussberg are generally fresh and particular care was taken to obtain unaltered sample material. All studied samples were examined microscopically and found to be remarkably fresh. They preserve fine petrographic structural and textural detail. Phenocrysts are virtually unaltered. Leucite occurs as 0·5–2 mm hexagonal crystals with characteristic complex twinning. They often contain rings of melt inclusions, which is a common feature of magmatic leucite (Mitchell & Bergman, 1991). Olivines are clear and colourless, and lack signs of significant alteration. They often contain very large melt inclusions and have very well developed crystal faces, implying rapid formation during magma ascent. Clinopyroxene crystals are relatively large (often >2 mm) and show no significant alteration. Devitrification textures in the glass matrix are rare.
|
ANALYTICAL PROCEDURES |
|---|
Whole-rock major element concentrations were determined at the Department of Earth Sciences, University of Queensland by inductively coupled plasma optical emission spectrometry (ICP OES) on a Perkin Elmer Optima 3300DV system. Approximately 50 mg of crushed sample were dissolved in 5 ml of 32% HCl, 3 ml of 70% HNO3, and 2 ml of 50% HF using a microwave oven. After digestion, 39·5 ml of 3·5% boric acid and 0·5 ml of a 1000 ppm Lu internal standard were added. Samples were then placed in the microwave for a further 10 min before analysis. An andesite standard, JA-3 (Imai et al., 1995
), was used as the calibration standard and results for other standards analysed as unknowns are shown in Table 1. Major element totals in Table 2 include BaO, SrO and ZrO2 and assume Fe2+/(Fe2+ + Fe3+) to be 0·63, the average ratio determined by Sheraton & Cundari (1980) for Gaussberg lamproites. Loss on ignition was not determined but major element concentrations are very similar to previously published results (Sheraton & Cundari, 1980).
|
Trace elements were analysed by ICP mass spectrometry (ICP-MS) on a Fisons PQ2+ system at the Advanced Centre for Queensland University Isotopic Research Excellence (ACQUIRE). Sample preparation and analytical procedures used were identical to those of Eggins et al. (1997) except that Tm was not used as an internal standard and W-2, a US Geological Survey diabase standard, was used as the calibration standard. Our preferred concentrations for W-2 and the measured concentrations and relative standard deviations for AGV-1 (an average of 26 analyses of nine digestions analysed over 4 years) are shown in Table 1. Concentrations for W-2 were derived partly by analysing it relative to synthetic standards (Li, Cr, Ni, Ga, Rb, Sr, Y, Zr, Nb, Cs, Ba, Hf, Ta, Pb, Th, and U) or are based on an assessment of published standard data (A. Greig, personal communication, 2001).
All isotope measurements were carried out on a VG 54-30 Sector multicollector mass spectrometer in static mode in the ACQUIRE laboratory. Procedures were identical to those of Wendt et al. (1999). The long-term (5 years) reproducibility of the NBS SRM 987 Sr and La Jolla Nd standards at ACQUIRE is 87Sr/86Sr = 0·710251 ± 17 and 143Nd/144Nd = 0·511861 ± 11, respectively. During the course of this study, the NBS SRM 981 Pb standard yielded an average fractionation per mass unit of 0·0007049 using pyrometer control and reproducibilities of 206Pb/204Pb = 0·0038, 207Pb/204Pb = 0·006 and 208Pb/204Pb = 0·0236.
|
MAJOR ELEMENT CHEMISTRY |
|---|
Lamproites have unusually high K contents, a characteristic more typical of highly evolved melts. However, they also have high MgO contents, which are typical of primitive magmas. The characteristically high K/Al ratio is one of the main criteria for the recognition of lamproites (Foley et al., 1987
). Although this peculiar chemistry partly reflects source composition, an important question remains regarding the nature of the primary melt composition. Complete chemical analyses for 24 Gaussberg lamproite samples are reported in Table 2. Despite representing all major phases of exposed volcanism at Gaussberg, the studied lamproites show little variation in major element chemistry, with very narrow ranges in K2O (11·5–13·5), Al2O3 (9·5–10·6), and MgO (7·5–9·5). Because of the lack of variation in the geochemistry of the Gaussberg lamproites (Table 2) it is difficult to assess the role of fractional crystallization on the basis of major (and trace) element content.
To assess the role of fractional crystallization, we compare chemistries of worldwide lamproites, which collectively show stronger trends on K2O vs K2O/Al2O3 (Fig. 3a) and K2O vs K2O/CaO plots (Fig. 3b). Compared with Harker diagrams, these plots discriminate more clearly between fractionation trends caused by leucite, clinopyroxene, olivine, and phlogopite (not shown in figures). This is because these minerals have distinctly different K2O/Al2O3 and K2O/CaO ratios and, therefore, distinct trends are evident when either fractionation or accumulation of any of these phases occurred. For example, olivine has a K2O/Al2O3 ratio of zero and a K2O/CaO of close to zero, which means that when olivine is accumulated in a melt these ratios in the melt do not change whereas the K2O content is systematically reduced (vector b in both Fig. 3a and b). It is significant that the compositional trends predicted for accumulation of leucite and phlogopite are not compatible with the trends defined by lamproite data. Therefore, major element variations in lamproites and specifically their K enrichment are not due to the accumulation of the only two plausible K-rich phases. The trends shown in Fig. 3 are, however, consistent with fractionation and/or accumulation of K-poor mineral phases such as clinopyroxene (vector a in both Fig. 3a and b) or olivine (vector c in both Fig. 3a and b). Thus, although the high-MgO and low-Al2O3 contents may be inherited primary features of lamproite melt, part of the compositional range defined by these elements may be due to the accumulation of mafic minerals. By analogy with other lamproites, fractional crystallization cannot be discounted as a process that has affected Gaussberg lamproites. For this reason we will deal mainly with aspects of lamproite chemistry that are insensitive to fractional crystallization.
|
), Yakokut lamproites, Siberia (
) (Mues-Schumacher et al., 1995
) and phlogopite lamproites (•), Leucite Hills lamproites (
) (Mitchell & Bergman, 1991
) (Gibson et al., 1995 |
ISOTOPE GEOCHEMISTRY |
|---|
A number of detailed studies on the isotopic chemistry of lamproites have concluded that such melts originate from a source that had experienced a prolonged history of K, Rb and light rare earth element (LREE) enrichment (Collerson & McCulloch, 1983
; Vollmer et al., 1984; Fraser et al., 1985; Nelson, 1992; Lambert et al., 1995; O’Brien et al., 1995). Foley et al. (1987) concluded that this source was metasomatized, enriched sub-continental lithospheric mantle. However, none of these models provide explanations for the nature of source enrichment and lack an obvious geological link between enrichment of the source and the melting process. Furthermore, ancient metasomatic source enrichment is inherently difficult to test. In this paper we combine the temporal constraints on the lamproite source imposed by the Rb–Sr, Sm–Nd, and U–Pb decay schemes to evaluate the possibility that the Gaussberg lamproites were derived from metasomatized lithospheric mantle and to test alternative explanations.
The isotopic compositions of the Gaussberg lamproites show almost no variation in Pb, Nd, and Sr isotope ratios (Table 3), with no discernible trends in 207Pb/204Pb vs 206Pb/204Pb space (Fig. 4a), 208Pb/204Pb vs 206Pb/204Pb space (Fig. 4b), or 
Nd vs 87Sr/86Sr space (Fig. 4c). This indicates that it is very unlikely that the isotope data represent a mixture of more than one source. In isotope plots, lamproites occupy a region that is more commonly associated with crustal rocks, rather than mantle-derived melts. As mentioned above and discussed in detail below, crustal contamination could only have had subordinate effects on the Pb, Nd, and Sr contents of Gaussberg lamproites (Collerson & McCulloch, 1983). Therefore, their unusual (for mantle-derived melts) isotopic composition reflects that of the mantle source.
|
|
) (Turner et al., 1999
) (Fraser et al., 1985
Testing the sub-continental lithospheric mantle hypothesis
Representative Nd vs 87Sr/86Sr data for a selection of lamproites for which high-quality data are available are shown in Fig. 4c. Lamproites exhibit a very extreme range in Nd from -4 (Wannamaker et al., 2000) to -26 (Fraser et al., 1985). Such compositions require an extended period of source evolution in isolation from the convecting upper mantle. This can be demonstrated when the Nd-isotope evolution of Gaussberg lamproites is compared with that of the MORB source mantle (e.g. Nägler & Kramers, 1998). Using the 147Sm/144Nd decay equation we calculate the Sm/Nd ratio necessary to produce the Gaussberg source from the MORB source mantle for any given time. From a plot of 147Sm/144Nd ratios against time (Fig. 5a) it is clear that the Gaussberg source must be at least 0·6 Gyr old. Taking a very conservative 147Sm/144Nd ratio of 0·078 for the source, isolation from convecting mantle must have occurred 1 Gyr ago. In this example, the 147Sm/144Nd ratio of MORB source was assumed to be 0·238 (Nägler & Kramers, 1998) and the implied enrichment that would have accompanied isolation thus would have lowered the ratio more than three-fold. Using a more realistic enriched source 147Sm/144Nd ratio of 0·15, which represents a much less dramatic LREE enrichment, the time required for isolation is 3 Gyr.
|
), relative to MORB source mantle (dashed line) of Kramers & Tolstikhin (1997)
|
The 147Sm/144Nd ratios that appear to characterize the Gaussberg source are significantly lower than those reported for sub-continental lithospheric mantle (McDonough, 1990). Xenoliths interpreted as being samples from the sub-continental lithospheric mantle have variable but substantially higher 147Sm/144Nd and 143Nd/144Nd ratios than are required in the Gaussberg source. Thus, if Gaussberg lamproites originated in the lithospheric mantle, their source would have to represent a portion that is not normally sampled in xenolith populations. Foley et al. (1987) have suggested that the source of lamproites could be enriched lithospheric veins. If such veins formed in the garnet stability field, where heavy REE (HREE) would be retained in the residue, they could, over time, evolve to 143Nd/144Nd ratios similar to those of Gaussberg. A separate thermal event would then be required to explain why these veins melted after >1 Gyr of stability in the lithospheric mantle.
Constraints imposed by the Rb–Sr isotope system indicate that the Gaussberg source must have undergone a prolonged period of Rb enrichment to explain the radiogenic Sr isotope characteristics of the lamproites (Collerson & McCulloch, 1983). An estimate of the Sr evolution of the upper mantle was calculated using an initial 87Sr/86Sr at 4·5 Ga of 0·69897 (BABI; Papanastassiou & Wasserburg, 1969) and 87Rb/86Sr ratio of 0·08923 (Allègre et al., 1983) and a linear decrease in the 87Rb/86Sr ratio with time to the present-day depleted upper-mantle 87Sr/86Sr value of 0·702706 (Allègre et al., 1983). If the Gaussberg source separated from the upper mantle 1 Gyr ago it would have had a 87Rb/86Sr ratio of 0·53 (Fig. 5b, analogous to Fig. 5a). Isolation at 3 Ga would have required a ratio of 0·22 and would be consistent with derivation of the source by partial melting in the garnet stability field, followed by isolation in the sub-continental lithospheric mantle, as indicated by the Sm/Nd systematics.
Isotopic evolution of Sr and Nd depends on the respective parent/daughter ratios, and time. In the case of the Gaussberg source, the parent/daughter ratio is not known nor is the time over which the source evolved. Although difficult to constrain, modelling of the Sr and Nd isotope evolution using estimated parent/daughter ratios shows that the Gaussberg source could have been directly derived from the upper mantle, but only if it remained isolated in the sub-continental lithospheric mantle for a period in excess of 1 Gyr.
This scenario is not compatible with Pb isotopes. The Pb isotopic composition of Gaussberg is very unusual for contemporary mantle-derived rocks. In 207Pb/204Pb vs 206Pb/204Pb space (Fig. 4a), Gaussberg lamproites plot above terrestrial evolution lines, but to the left of the meteorite isochron. This apparently contradicts the first Pb paradox (i.e. future paradox), which states that average marine sediment and MORB have evolved to the right of the meteorite isochron (Allègre, 1969; Kramers & Tolstikhin, 1997). Importantly, all ocean island basalts (OIB) plot to the right of the meteorite isochron (Kamber & Collerson, 1999). Only certain crustal rocks plot close to Gaussberg Pb. These are generally Archaean granitoids that have undergone U loss, associated with later high-grade metamorphism (e.g. Montgomery & Hurley, 1978).
The very unusual Pb isotope composition of the Gaussberg lamproites places strong constraints on any model of lamproite source evolution, because 207Pb/206Pb evolution is non-linear as a result of the difference in half-life between 238U and 235U. A temporal model can therefore be developed to calculate whether, and at what time, the Gaussberg source separated from a specific reservoir such as the MORB source mantle. Such a temporal model can then be evaluated together with the Nd and Sr isotopic constraints.
The temporal evolution of Pb isotopes in the MORB source mantle and in various crustal reservoirs has been modelled by Kramers & Tolstikhin (1997). Pb isotope evolution of a source that separated from the MORB source mantle at a specific time can be calculated using the Kramers & Tolstikhin (1997) 207Pb/204Pb and 206Pb/204Pb values as the initial values in the decay equations with the parent/daughter ratio as the only remaining unknown variable. The 238U decay equation can be solved for a given time to yield the present-day Gaussberg 206Pb/204Pb ratio. From this analysis the 238U/204Pb (µ) for the source can be derived (Fig. 5c). Invariably, we find that when this µ is applied to the 235U decay equation, the corresponding 207Pb/204Pb ratio is lower than that of Gaussberg and plots to the left of the meteorite isochron but below terrestrial models. Solving the 235U decay equation produces isotope compositions that plot above terrestrial mantle models but to the right of the meteorite isochron. As an example, we show the results of such calculations using a separation age of 2·5 Ga in Fig. 5d. As stated above, 207Pb/204Pb ratios are much lower than the observed values if the 238U decay scheme is used for calculation of µ. Conversely, the 235U decay equation yields 206Pb/204Pb ratios that are too high. This finding reflects the fact that it is impossible to obtain evolution paths that converge close to the Gaussberg value. Thus the Gaussberg source could not have originated from 2·5-Gyr-old MORB mantle if it evolved with a single-stage µ. This is true irrespective of the proposed time of isolation, because the U/Pb ratio calculated via the two independent decay schemes do not converge over the age of the Earth (Fig. 5c).
Hence, if Gaussberg Pb evolved from the upper mantle, a multi-stage history must be invoked, such as that previously proposed by Williams et al. (1992). A similar multi-stage history was also proposed by Nelson (1992) for the Ellendale lamproite of Western Australia. For example, a source starting with upper-mantle Pb isotope ratios at 2·5 Ga would have had to evolve with a µ of 26 for 0·6 Gyr before experiencing a strong decrease in µ to 1·55 for the remaining 1·9 Gyr. This is graphically shown in Fig. 5e. Similar two-stage evolution histories can be calculated for different separation times, provided that the separation occurred before 2·2 Ga.
The 2·5 Ga model requires the initial fractionation event to increase the µ from the upper mantle value of 6·68 to 26. Although extreme, a melting event could cause such a change in the U/Pb ratio and could also have increased the Rb/Sr ratio and reduced the Sm/Nd ratio. However, the Pb isotope data require a subsequent event of massive U depletion at 1·9 Ga. It is unlikely that melt extraction could have removed more than 90% of the U from the source without complementary Pb depletion. Furthermore, melt extraction would also reduce the Rb/Sr ratio and increase the Sm/Nd ratio, a situation that is not consistent with a source that has experienced K and LREE enrichment. Therefore, if the Gaussberg source separated from the MORB mantle in the late Archean, a third event of relatively recent metasomatism would be required. This would have strongly enriched K and LREE but also flooded the source with Sr, Nd and Pb, which would probably exhibit MORB source mantle isotope compositions. The evolution of such a source would become increasingly difficult to model, relying on increasingly older ages of isolation from the MORB source mantle and increasingly dramatic enrichment and depletion events, as was suggested by Williams et al. (1992). In the following, we propose a new and more elegant solution to explain the isotope evolution of the source of the Gaussberg lamproites.
A single-stage mantle evolution model for the lamproite source
The complexity of lithospheric mantle models is largely caused by the high 207Pb/206Pb ratio of the Gaussberg lamproites. It therefore appears more likely that Gaussberg Pb evolved for a significant length of time early in the Earth’s history in a high-µ environment before reduction of µ in a U/Pb fractionation event. Isotopically similar evolutions are typically found in high-µ Archaean gneisses that have experienced substantial U loss induced by subsequent metamorphism (e.g. Montgomery & Hurley, 1978).
Continental crustal Pb has had a major influence on the isotope evolution of Pb in all terrestrial silicate reservoirs, including the MORB source mantle (Kramers & Tolstikhin, 1997). For example, recycling of U and Pb from the upper crust into the MORB source mantle as subducted sediment explains the second terrestrial Pb paradox (Kramers & Tolstikhin, 1997; Collerson & Kamber 1999) and has strongly influenced the 206Pb/204Pb isotope evolution of the mantle (Kramers & Tolstikhin, 1997). According to Ringwood et al. (1992) it may be possible for some subducted sediment to remain isolated as high-pressure mineral phases in the Transition Zone or lower mantle. This provides a mechanism for Pb that evolved for a considerable time in continental crust to be isolated and stored for a considerable length of time in the mantle. Thus, continental-derived material can evolve in a closed system until a subsequent melting event. Slab dehydration driven by metamorphism at the time of subduction is a plausible mechanism for fractionation of parent/daughter ratios (e.g. Becker et al., 2000).
We now test whether this hypothesis could provide a model for the Gaussberg magma source. Kramers & Tolstikhin (1997) modelled the Pb isotope evolution of terrestrial sediment that has been subducted into the mantle throughout Earth’s history. We use this modelled sediment reservoir, termed ‘erosion mix’, as an approximation to the Pb isotope composition of terrestrial sediment to test whether the Pb isotope evolution of the Gaussberg source can be reconstructed from it. We calculate from ‘erosion mix’ the µ required to produce Gaussberg Pb isotope ratios via the 238U and the 235U decay equations, analogous to the calculation for Fig. 5c. The solutions converge at 2·6 Ga (Fig. 5f). This indicates that the Gaussberg Pb isotopes can indeed be derived from ‘erosion mix’ with initial 206Pb/204Pb, 207Pb/204Pb and 208Pb/204Pb values of 14·32, 15·04 and 33·74 with a µ of 6·4 and a 232Th/238U (
) of 5·2 from 2·6 Ga. If the Gaussberg source was derived from subducted sediment, only relatively small changes in µ and are necessary. Relative to the original ‘erosion mix’, a lowering of the original µ from 13·8 to 6·4 and an increase of the original from 3·9 to 5·2 is required. This could have occurred in a number of ways:
- if the subducted material represented a mixture between silicate sediment and an appreciable amount of carbonate (carbonate is enriched in Pb, is low in U and has a high Th/U ratio);
- if accessory phases such as monazite, which concentrate Th, were stable during subduction (hence increasing the Th/U ratio);
- if high O2 fugacity prevailed during subduction; oxidation of U4+ to the more soluble U6+ could have caused U to be lost relative to both Th and Pb, thereby reducing µ and increasing .
Trace element data reported for eclogites that are interpreted to have undergone subduction zone dehydration-induced metamorphism (Becker et al., 2000) show a large range in µ (1–30) and (0·2–6·2). The Gaussberg source model µ and lie within this range. More importantly, the model-derived of 5·2 for the Gaussberg source is in good agreement with the calculated from 230Th/232Th isotope results (Williams et al., 1992).
Above, we demonstrated that an extreme 147Sm/144Nd ratio would be required if the Gaussberg source had evolved directly from depleted mantle. However, the Nd of subducted sediment (approximated again by ‘erosion mix’) at 2·6 Ga was significantly negative (–5), because it contained unradiogenic Nd that evolved in the continental crust (Nägler & Kramers, 1998). Thus a much less extreme 147Sm/144Nd ratio of 0·155 is required to produce Gaussberg Nd isotope ratios from the ‘erosion mix’ at 2·6 Ga with an initial 143Nd/144Nd ratio of 0·50923. To achieve the calculated 147Sm/144Nd ratio of the Gaussberg source from the ‘erosion mix’ (147Sm/144Nd ratio of 1·05 at 2·6 Ga; Nägler & Kramers, 1998) the Sm/Nd ratio must have increased. During subduction, slabs undergo LREE depletion, which increases the 147Sm/144Nd ratio. This is seen in ultra-high-pressure rocks such as jadeite quartzites (ultra-high-pressure metasediments) from the Dabie Mountains, China, that have a range of 147Sm/144Nd ratios from 0·11 to 0·16 (Liou et al., 1997). This does not conflict with the LREE-enriched nature of Gaussberg lamproites because their REE characteristics are likely to predominantly reflect melting processes.
It is not difficult to produce the observed 87Sr/86Sr ratios in a mildly enriched isolated mantle reservoir over a period of 2–3 Gyr. Therefore, even less enrichment is required if the source started with a more radiogenic Sr isotope composition (such as that of subducted sediment). Because the Sr isotope composition of weathered and eroded continental material is likely to have been heterogeneous (as it is today) no quantitative treatment of the source evolution in Rb–Sr is warranted. In summary, producing the Gaussberg source Nd and Sr isotope signature from continental-derived sediment in the Transition Zone is a simpler solution than enrichment of sub-continental lithospheric mantle and is consistent with 207Pb/206Pb evolution of the lamproites.
|
TRACE ELEMENT CHEMISTRY |
|---|
In this section we evaluate whether our model is consistent with the trace element chemistry of Gaussberg lamproites. Gaussberg lavas have very unusual trace element chemistry as they show extreme, but irregular enrichment in incompatible elements (Table 2; Fig. 6a). However, they show very limited variation in both major and trace element concentrations (Table 2). It is probable that fractional crystallization has affected the trace element chemistry to some degree, but this mechanism cannot be used to explain the extraordinary enrichment in incompatible elements, the shape of REE patterns (Fig. 7), or the general trace element systematics discussed below. Nevertheless, where possible we use combinations of elements that have very similar partition coefficients (in MORB melting) to examine features that have been least affected by fractional crystallization. This allows us to study the effect of crustal contamination (if any) on Gaussberg lamproites, and to evaluate features of the source chemistry.
|
Continental contamination
Gaussberg lamproites are considerably more enriched in incompatible elements (Fig. 6a) than average continental crust (Hofmann, 1988). This does not, per se, preclude the possibility of crustal contamination. There are, however, a number of direct and indirect lines of evidence against significant crustal contamination:
- comparison of MORB-normalized REE plots of Gaussberg lamproites (Fig. 7) and average continental crust (Hofmann, 1988) shows that continental crust is enriched in LREE but not nearly to the extent of the lamproites. If contamination had occurred, the LREE concentration of the lamproites would be reduced! At the same time, the lamproites are depleted in HREE relative to both continental crust and MORB. Contamination would therefore readily increase their HREE content. In other words, the slope of the normalized REE pattern would flatten. In addition, it is unlikely that crustal contamination would have affected all the sampled flows equally. The slopes of individual lamproite samples, however, do not deviate from the average to any significant degree (Fig. 7), which indicates insignificant contamination.
- A worst-case scenario is calculated where we assume that the uncontaminated lamproitic melt had a Lu content of zero and that the Lu content of Gaussberg lamproites is directly proportional to crustal contamination. Even in this hypothetical scenario, contamination could only account for 30% of Gaussberg Pb.
- If the Gaussberg Pb isotopic signature were to be explained by 30% crustal contamination, the composition of that crust would have to plot well outside the known terrestrial Pb isotope space, at very high 207Pb/206Pb ratios seen only in lunar samples.
- Most of the other known lamproites also plot above the depleted mantle and to the left of the meteorite isochron in 207Pb/204Pb vs 206Pb/204Pb space (Fig. 4a). This implies that their sources evolved in a very similar way to Gaussberg. It is very unlikely that they experienced crustal contamination to a similar degree and by crust with a similarly unusual isotope composition.
We therefore conclude, in agreement with Collerson & McCulloch (1983) that the chemistry of Gaussberg lamproites is a reflection of source chemistry and source mineralogy with no significant crustal contamination.
Inferred source chemistry
Gaussberg lamproites have very unusual trace element characteristics, with extreme but irregular enrichments in incompatible trace elements (Fig. 6a). These unusual geochemical features are likely to reflect particular source features. In this section we review aspects of the chemistry of the lamproites, using relative element abundances that have not been strongly influenced by fractional crystallization, to elaborate the nature of the Gaussberg source. We first discuss the abundance of the moderately to highly incompatible elements Sr, Pb, Nb, and Ta, as these show significant anomalies in Gaussberg lamproites (Fig. 6a). Second, we will discuss their superchondritic Nb/Ta ratio, a feature that sets them apart from most terrestrial volcanic rocks. Third, we discuss the contents of U, Th and Cs relative to K and Rb. Finally, we discuss the high concentrations of the compatible elements Mg, Ni, and Cr of lamproites in the context of a subducted sediment source for lamproites.
The concentration of Pb in Gaussberg lamproites is high. This manifests as a prominent positive spike on a MORB-normalized trace element variation diagram (Fig. 6a). Sr, on the other hand, is significantly under-abundant. These anomalies could not have originated during melting, because Sr and Pb have similar compatibility during dry mantle melting (Hofmann, 1988). If the source of Gaussberg lamproites was in the sub-continental lithospheric mantle (which is depleted as a result of the production of continental crust), a very specific enrichment event is required where Pb is preferentially enriched and Sr depleted relative to elements with similar compatibility. It is possible to fractionate elements with similar partition coefficients, for example during unmixing of two immiscible melts or fluids (Veksler et al., 1998). Gaussberg source lithospheric mantle would have required a very specific metasomatism by a fluid or melt with relatively high Pb but low Sr content. This complex and specific history, although not impossible, is very difficult to test.
Our model, on the other hand, assumes that the source of Gaussberg lamproites originated from subducted, continental-derived sediments. Therefore we compare the lamproites with an estimate of present-day Global Subducting Sediment (GLOSS; of Plank & Langmuir, 1998). When normalized to GLOSS, the lamproites no longer have a negative Sr anomaly (Fig. 6b), because sediments inherit a negative Sr anomaly from continental crust. Moreover, in this normalization, Pb has a negative anomaly (Fig. 6b). This is because normalization to GLOSS removes the effect of inheritance of relatively high Pb concentrations from continental crust in sediment. This feature points to the influence of subduction-dehydration on the chemistry of the Gaussberg source before its storage in the deep mantle. In this regard, Pb is a highly mobile element (Brenan et al., 1995) and is readily lost from slabs via dehydrating fluids.
Prominent negative anomalies are present for Nb and Ta in a MORB-normalized trace element diagram (Fig. 6a). Conventionally, the negative Nb and Ta anomalies (often associated with negative Ti anomalies) of lamproites are explained by residual rutile in the lithospheric mantle (Mitchell & Bergman, 1991). If this explanation is applied to Gaussberg lamproites it implies that the residue is strongly enriched in rutile. If we assume that the Th/Nb ratio of the Gaussberg source was similar to that of MORB (0·052), and that Nb was retained in rutile to produce the Th/Nb ratio of the lamproites (0·245), the source would have retained 4·5 times the Nb content of the lamproites. Because rutile is the main Ti carrier, a linear correlation between Nb and Ti can be expected. However, in Gaussberg lamproites Ti does not show a negative anomaly in a MORB-normalized pattern (Fig. 6a) and is not correlated with Nb. Titanium in the Gaussberg source was either decoupled from Nb or the negative Nb and Ta anomalies are primary features of a source other than the sub-continental lithospheric mantle. We favour the latter explanation and propose that the low Nb and Ta contents were inherited from continental crust. When normalized to GLOSS, Nb and Ta show positive anomalies (Fig. 6b). This again reflects the process of dehydration during subduction where Nb and Ta are compatible and therefore not significantly lost relative to U and Th (Kamber & Collerson, 2000; Rudnick et al., 2000).
The Gaussberg lamproites have superchondritic Nb/Ta ratios (17·2–20·8), which are extremely rare in terrestrial volcanic rocks. This is not likely to be an analytical artefact for several reasons. First, the analytical protocol in the ACQUIRE laboratory consistently reproduces the Nb and Ta concentrations of standard rocks (Table 1). For example, the propagated 1-sigma standard deviation on the Nb/Ta ratio of AGV-1 from the concentration data (Table 1) is 17·14 ± 0·28. Second, reproducibility of the Nb/Ta ratio is even slightly better than those of the Nb and Ta contents. For example, the average Nb/Ta ratio of AGV-1 by ACQUIRE ICP-MS analysis is 17·14 ± 0·27. Third, the only potential deviation from true values is towards lower Nb/Ta ratios because the less abundant element Ta is more sensitive to sample carry-over and occasional high aberrant blanks (or beaker memory). Thus, we regard the superchondritic Nb/Ta ratio of the lamproites to be a real feature.
Virtually all OIB and MORB have chondritic to slightly subchondritic Nb/Ta ratios (Kamber & Collerson, 2000). Because Nb and Ta are not significantly decoupled during melting, the Gaussberg source must have had a superchondritic Nb/Ta ratio. Nb/Ta ratios in the sub-continental lithospheric mantle range from chondritic for depleted samples (Jochum et al., 1989) to crustal (i.e. significantly subchondritic) for enriched samples (McDonough, 1990). Samples of the sub-continental lithospheric mantle that are interpreted to have undergone metasomatism contain spinel with reaction rims enriched in Nb and Ta (Bodinier et al., 1996). These rims dominate the Nb and Ta budget, but they consistently have low Nb/Ta ratios. Therefore, melts of enriched sub-continental lithospheric mantle should have low Nb/Ta ratios.
Nb/Ta ratios of sediments are also low (e.g. GLOSS = 14·2; Plank & Langmuir, 1998) and melting of unmodified sediment should lead to melts with low Nb/Ta ratios. However, Nb and Ta are fractionated in subduction zones, either during dehydration (Kamber & Collerson, 2000) or during partial melting with residual rutile (Rudnick et al., 2000). This is inferred from the fact that eclogitic metabasalts have substantially higher Nb/Ta ratios (33·7 ± 3·5; Kamber & Collerson, 2000) than their protoliths. High-P, medium-T metasediments from the Swiss Alps (Henry et al., 1996) also show strong unidirectional Nb/Ta fractionation, with some dehydrated samples recording a Nb/Ta increase from 11 in the original sediment to 18 in their metamorphosed equivalent. Therefore, the high Nb/Ta of Gaussberg lamproites may well be a source feature that records Ta loss during subduction-zone processes. On a global scale, lamproites exhibit a large range in Nb/Ta from 14 to 20 (Gibson et al., 1995; Lambert et al., 1995). In our model this reflects variable extent of dehydration of heterogeneous sediment.
Further features of Gaussberg lamproites are relative depletions in Cs, Th, and U (and Pb) compared with GLOSS (Fig. 6b). These can also be explained by subduction metamorphism, because Cs, Th and U are among the most fluid-mobile during dehydration (Brenan et al., 1994). They are strongly enriched in arc magmas (Pearce & Peate, 1995) whereas complementary depletions are seen in high-pressure metabasalts (Becker et al., 2000) and metasediments (Liou et al., 1997). In view of inter-element differences in distribution behaviour (i.e. lacking the chemical similarity of Nb and Ta) and redox sensitivity (U), no quantitative treatment is warranted.
Lamproites are not only strongly enriched in incompatible elements but also have substantial concentrations of compatible elements. The high concentrations of MgO, Ni, and Cr are difficult to explain if the lamproite source consisted entirely of deeply subducted sediment. The isotopic and incompatible trace element evidence does not, however, require a pure metasediment source. Pb, Nd, and Sr isotope compositions can be reconciled with a substantial contribution of basaltic melt from melting of surrounding peridotite. There is the added possibility that eclogitic oceanic crust with which the metasediment was associated also melted and contributed towards the compatible element budget of the final lamproite melt. Further deliberations on the issue of compatible element (including MgO) contents of lamproites are at present not warranted because it is unclear what compositions can be expected for melting of peridotite and eclogite at Transition Zone pressure. However, Os-isotope systematics could probably discriminate between a peridotite and eclogite origin of the compatible elements in lamproites.
|
MELTING ENVIRONMENT |
|---|
Knowledge of potential melting environments and the properties of melting assemblages are important components of any model for lamproite formation. Gaussberg lamproites require isolation of their source from the convecting mantle for at least 2 Gyr. Plausible physical environments are restricted to the sub-continental lithospheric mantle, the Transition Zone, and the lower mantle because only they can withstand entrainment into convecting upper mantle on a time scale >0·5 Gyr. In this section we discuss physical (i.e. pressure and temperature) and mineralogical controls on potential lamproite sources provided by high-pressure and -temperature experiments on relevant compositions.
Results of experiments at lithospheric pressures and temperatures (Mitchell, 1995; Sato et al., 1997; Edgar & Mitchell, 1997; Konzett et al., 1997) suggest that an assemblage rich in phlogopite, K–Ti-amphiboles, clinopyroxene, apatite, and a mixture of exotic K–Ba–Zr–Nb titanates could be the source of lamproites. These mineral phases are potentially stable throughout the sub-continental lithospheric mantle possibly extending into the diamond stability field (Sato et al., 1997; Konzett et al., 1997), and could be represented by MARID (mica, amphibole, rutile, ilmenite, and diopside) xenoliths found in kimberlites (Dawson & Smith, 1977). Experimental evidence for the upper pressure and temperature stability of such paragenesis is incomplete. If such a source is envisaged for Gaussberg, melting could have occurred by decompression or thermal perturbation (by a plume).
If we assume a geothermal gradient equivalent to a surface heat flow of 44 mW/m2 it seems possible that an enriched sub-continental lithospheric mantle source could melt as a result of decompression (Sato et al., 1997). However, decompression melting proceeds to varying degrees, leading to fractionation trends in major and trace elements. Such trends are not seen in Gaussberg lamproites. This is supported by the lack of evidence for the surface expression of sub-continental lithospheric mantle thinning or erosion (e.g. rifting, basin formation) or other likely features of ascent of hot asthenosphere beneath the sub-continental lithospheric mantle. Rather, the East Antarctic shield in the vicinity of Gaussberg supports the burden of an ice sheet.
Alternatively, melting could have been caused by thermal perturbation from a plume head. This, too, seems an unlikely scenario for Gaussberg lamproites. Because plume heads incorporate, upon ascent through the asthenosphere, material from surrounding mantle their melting products (i.e. OIB) invariably represent mixtures of deep and shallow mantle material. This is clearly expressed in the Pb-isotope systematics of OIB, yet no mixing trends are seen in the chemical and isotopic composition of Gaussberg lamproites. Furthermore, the Pb isotope composition of all plume-related OIB is so distinctive (Fig. 4a) that a plume source for Gaussberg can be excluded. In addition, Gaussberg is situated 500 km from the closest expression of plume volcanism at Heard Island (Collerson & McCulloch, 1983). As volcanism occurred contemporaneously at Heard Island and Gaussberg it is very unlikely that the two are related. Therefore in agreement with Collerson & McCulloch (1983) and Tingey et al. (1983), we envisage a separate isolated magmatic event.
In the deep mantle, only the highly viscous lower mantle or the Transition Zone could contain lamproite source material. We have argued that the source of lamproites is deeply subducted sediment. Mantle tomographic studies show that oceanic lithospheric slabs containing a small amount of subducted sediment can accumulate in the Transition Zone (e.g. Simons et al., 1999). The important factor that governs the behaviour of slabs at Transition Zone depths is buoyancy, which is controlled by the relative densities of mineral phases. Experimental studies on crustal compositions (Irifune et al., 1994