Journal of Petrology

This Article Abstract FREE Full Text (PDF) Supplementary data Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (47) Request Permissions Google Scholar Articles by LE ROEX, A. P. Articles by DAVIS, P. Search for Related Content GeoRef GeoRef Citation Social Bookmarking          What's this?

Journal of Petrology | Volume 44 | Number 12 | Pages 2261-2286 | 2003
© Oxford University Press 2003; all rights reserved

Petrogenesis of Group I Kimberlites from Kimberley, South Africa: Evidence from Bulk-rock Geochemistry

ANTON P. LE ROEX^*, DAVID R. BELL and PETER DAVIS

DEPARTMENT OF GEOLOGICAL SCIENCES, UNIVERSITY OF CAPE TOWN, RONDEBOSCH, 7701, SOUTH AFRICA

^* Corresponding author. Fax: +27-21-650-3783. E-mail: aleroex{at}geology.uct.ac.za

RECEIVED OCTOBER 16, 2002; ACCEPTED JUNE 12, 2003

	ABSTRACT

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
Fresh samples of hypabyssal kimberlite from the five major kimberlite pipes in the Kimberley area of South Africa have been analysed for their bulk-rock major and trace element geochemistry. The geochemical data allow identification of the influence of crustal contamination in certain samples, best illustrated in terms of elevated SiO₂, Al₂O₃, Pb and heavy rare earth element (HREE) contents. Samples devoid of such crustal contamination show coherent major and fluid-immobile trace element variations, whereas fluid-mobile trace elements are scattered. Kimberlites rich in macrocrysts are shown to reflect substantial (up to 35%) entrainment of mantle peridotite, with Ni–SiO₂ and Sc–SiO₂ variations defining mixing trajectories towards garnet lherzolite. The likely primary magma(s) parental to the Kimberley kimberlites is suggested to have a composition of 26–27 wt % MgO, 26–27 wt % SiO₂, 2·2 wt % Al₂O₃ and Mg number 0·86. Subtle differences in chondrite-normalized REE abundance patterns can be explained by small variations in the degree of partial melting within the range 0·4–1·5%, leaving residual garnet. The data are satisfied by melting a source enriched relative to chondrites by a factor of 10 in light REE (LREE), with chondritic or lower HREE abundances. Extended normalized trace element diagrams exhibit significant negative K, Rb, Sr and Ti anomalies that are interpreted to be primary magma characteristics, despite evidence for secondary mobility of K, Sr and Rb. A model is proposed in which fluid or melt from a sub-lithospheric source region precipitates phlogopite en route to metasomatizing the overlying subcontinental mantle lithosphere, imprinting its geochemical signature on a source region previously depleted in HREE relative to primitive mantle. Subsequent 1% melting of the metasomatized source produces a kimberlite with compatible element characteristics strongly influenced by depleted lithospheric peridotite (high Mg number, high Ni, low HREE), but with incompatible elements (and their isotope ratios) characteristic of the deeper source. The similarity of incompatible element ratios (Nb/U, Nb/Th, Ce/Pb) in the kimberlite magmas to those of ocean island basalts from the South Atlantic suggests an ultimate origin in an upwelling mantle plume.

KEY WORDS: kimberlite; geochemistry; Kimberley; petrogenesis

	INTRODUCTION

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
Kimberlite constitutes a rare, highly alkaline, rock type that has in many ways attracted more attention than its relative volume might suggest that it deserves. This is largely because it serves as a carrier of diamonds and garnet peridotite mantle xenoliths to the Earth's surface. Furthermore, its probable derivation from depths greater than any other igneous rock type, and the extreme magma composition that it reflects in terms of low SiO₂ content and high levels of incompatible trace element enrichment, make an understanding of kimberlite petrogenesis important. In this regard, the study of kimberlite has the potential to provide valuable information on the composition of the deep mantle, and melting processes occurring at or near the interface between the cratonic continental lithosphere and the underlying convecting asthenospheric mantle, complementing similar studies on related mafic alkaline magmas of shallower derivation such as melilitite, lamprophyre and carbonatite (e.g. Rogers et al., 1992; Wilson et al., 1995; le Roex & Lanyon, 1998; Bell & Tilton, 2001; Janney et al., 2002). le Roex (1986) argued for a close relationship between upwelling South Atlantic mantle plumes and southern African kimberlite emplacement, and the subsequent recognition of high-pressure (>400 km) mineral phases in some diamond inclusions (e.g. Moore & Gurney, 1985) provides further support for a possible connection between kimberlites and the deep mantle.

Since the early experimental studies of Eggler and Wyllie (Eggler, 1978; Wyllie, 1979, 1980), which clearly demonstrated the important role that CO₂ and H₂O play in kimberlite magma genesis, numerous subsequent studies have developed this theme further and established the importance of low degrees of melting of carbonated peridotite in the generation of magmas of kimberlite composition (Ringwood et al., 1992; Brey & Ryabchikov, 1994; Kesson et al., 1994; Girnis et al., 1995; Dalton & Presnall, 1998; Wyllie & Lee, 1999). Studies concerned with the geochemistry of kimberlites are fewer, and have largely addressed major element (e.g. Clement, 1982; Shee, 1985; Sweeney & Winter, 1999), or radiogenic isotope variations (e.g. Smith, 1983; Fraser et al., 1985–1986; Smith et al., 1985; Nelson, 1989; Taylor et al., 1994; Beard et al., 1998). Of particular importance in this regard was the isotopic study by Smith et al. (1983), which recognized the occurrence in southern Africa of two geochemical varieties of kimberlite: Group I (unradiogenic Sr, Pb; radiogenic Nd) and Group II (radiogenic Sr and Pb; unradiogenic Nd), corresponding broadly to basaltic and micaceous kimberlites, respectively. Fewer studies have focused on the trace element geochemistry of kimberlites with the view to constraining the petrogenetic processes involved in their genesis [some exceptions are those by Fraser & Hawkesworth (1992), Tainton & McKenzie (1994), Taylor et al. (1994), Beard et al. (1998) and Price et al. (2000)].

With the exception of the work by Fraser & Hawkesworth (1992), Tainton & McKenzie (1994) and Price et al. (2000), none of the above studies has attempted to quantify the petrogenetic processes leading to formation of kimberlite magmas. Rather, the focus has been on characterizing the composition and evolution of the mantle source regions involved in terms of incompatible trace element composition, and addressing the question of the sub-lithospheric vs lithospheric mantle origins of the magmas (e.g. le Roex, 1986; Taylor et al., 1994; Beard et al., 1998). Even the well-known Kimberley group of kimberlites have not been accorded any rigorous (quantitative) interpretation of their trace element abundance variations, with most studies being more interested in their mantle xenolith load than the host rocks themselves, although some have provided largely qualitative comment on trace element abundances (e.g. Clement, 1982; Muramatsu, 1983; Muramatsu & Wedepohl, 1985; Shee, 1985). Tainton & McKenzie (1994) included some data from the Kimberley area in their regional quantitative study of kimberlite generation. To date, no attempt has been made to quantify the petrogenetic processes giving rise to these Group I kimberlites, so well exposed in the Kimberley mines, through the use of high-quality trace element data. The close spatial proximity of five major kimberlite pipes and the comparatively fresh, hypabyssal-facies material exposed in the deep mines makes this group of intrusive kimberlites ideal to study with respect to gaining a quantitative understanding of the process involved in their formation and their subsequent evolution en route to the surface.

To ensure an internally consistent dataset for use in quantitative petrogenetic modelling, and to make use of more modern analytical techniques, covering a wider range of trace elements than previously available, the major kimberlite intrusions from the Kimberley area have been resampled and reanalysed. The primary aims of this study are to quantify as precisely as possible the melting process giving rise to the Kimberley kimberlites, to estimate the likely trace element composition and mineralogy of the mantle source region, and evaluate subsequent shallow-level processes that have modified the original primary magma compositions.

	GEOLOGICAL SETTING AND SAMPLING

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
The Kimberley cluster of kimberlite pipes constitutes the type locality for kimberlite, and as such is an ideal location for studying the geochemistry and petrogenesis of Group I kimberlites. The cluster of kimberlites has been dated at 84 ± 3 Ma (Clement et al., 1979). They intrude Archaean basement gneisses, Ventersdorp lavas and a succession of Mesozoic Karoo sedimentary and volcanic rocks. Present exposure of Karoo rocks in the mines includes intrusive dolerite, carbonaceous shales, siltstone, sandstone of the Ecca and Dwyka groups; vesicular flood basalt lavas of the originally overlying, but now eroded, Stormberg group are found as down-rafted xenoliths in the kimberlites. Approximately 2 km has been removed by erosion in the Kimberley area, resulting in mid-level surface exposure of kimberlite diatremes and penetration to diatreme root zones by mining activity. This degree of exposure provides an abundance of magmatic kimberlite suitable for petrogenetic investigation.

The Kimberley cluster includes five major kimberlite pipes, the locations of which are shown in Fig. 1. These are the De Beers, DuToitspan, Bultfontein, Wesselton and Kimberley mines, the last otherwise known as the ‘Big Hole’. A number of smaller pipes also occur in the Kimberley area, several of which have been mined for diamonds (Wagner, 1914). The smaller pipes were not considered in this study. A kimberlite sill complex is developed around the Wesselton pipe and lies immediately below a major Karoo dolerite sill, which appears to have formed a barrier preventing the intruding kimberlite magma from erupting as a diatreme. Within the root zone of each of the major kimberlite pipes, a number of discrete intrusions can be recognized, each showing distinct mineralogical and textural features. The detailed geology of the individual pipes has been comprehensively described by Clement (1982) and Shee (1985).

View larger version (25K): [in this window] [in a new window]   Fig. 1. Sketch map of the Kimberley cluster of kimberlites, showing location of the major kimberlite pipes sampled in this study. Adapted from Shee (1985).

 
Fifty-four samples of hypabyssal-facies kimberlite were selected from each of the five major pipes, and from the Wesselton Floors sill complex. Access to the Kimberley Mine (Big Hole) is no longer possible and samples of kimberlite were therefore collected from the Colville dumps, which were derived from the floors area at the time of mining of the Big Hole (J. Robey, personal communication, 2000). Both macrocrystic (abundant large, anhedral olivine crystals) and aphanitic (absence of large anhedral olivine crystals) kimberlite was sampled, with attention being paid to selecting only the freshest material available. Samples were selected to cover the range of intrusive phases present, and include the main pipe intrusions and cross-cutting dykes and sills.

	PETROGRAPHY

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
The petrography of the Kimberley kimberlites has been extensively described by previous workers (e.g. Clement, 1982; Shee, 1985), and will not be repeated here. Rather, only those aspects salient to the subsequent discussions are highlighted. In thin section the samples exhibit considerable textural and mineralogical diversity. According to the classification scheme of Skinner & Clement (1979), which is based on the fine-grained matrix mineralogy alone, most of the samples are serpentine monticellite kimberlites with variable amounts of calcite. Modal abundances of the major constituent phases were determined in a subset of the samples (selected to represent the range in textures) by point counting (1000 points); these data revealed that there is a strongly heterogeneous distribution of the constituent phases within individual samples. Seven thin sections of a single hand specimen from the Kimberley pipe (COL1) have olivine macrocryst (anhedral crystals >2 mm in size) proportions that vary from 14·6 to 24·4 vol.%, phlogopite macrocrysts from 0·2 to 2·6 vol. %, and olivine phenocrysts (euhedral crystals <0·5–1 mm in size) from 12·7 to 24·2 vol. %. Modal abundances determined on single thin sections are therefore of limited value, and are not reported here. Rather, samples are simply referred to as macrocrystic where olivine macrocrysts constitute >10 vol. % of the sample, and aphanitic when less. Where present, olivine macrocrysts are anhedral and commonly exhibit strain features (e.g. olivine neoblasts, undulose extinction). They are often fragmented and small fragments can be confused with phenocrystic olivine. Olivine macrocryst abundances range from <5 vol. % to >30 vol. %, and are commonly concentrated into distinct macroscopic flow bands (e.g. COL2; K119/1). Olivine macrocrysts are frequently extensively modified to serpentine through deuteric alteration—always along fractures, commonly pervasively. Other, less common, macrocryst minerals include phlogopite and even rarer garnet, Cr-diopside and ilmenite. When present, phlogopite macrocrysts constitute an insignificant percentage of the samples (<2%; DuToitspan sample K6/55 has 9 vol. %), exhibit strain features such as kink banding and undulose extinction, and may have resorbed outer rims. Olivine is the most abundant phenocryst phase, ranging in size from 0·3 to 0·5 mm, and is typically euhedral in habit. The phenocrysts are commonly serpentinized, either pervasively or along edges, and modal abundances range from <10 vol. % to >30 vol. % (K6/10 has >55 vol. %).

Matrix phases include complex compositionally zoned spinels, perovskite, ilmenite, phlogopite, monticellite, carbonate minerals, apatite and serpentine. Carbonate frequently forms pools or trails in the matrix, and segregationary textures (Clement & Skinner, 1979) are present in some samples (e.g. Wesselton Floors sill sample K5/86). The matrix is generally fine-grained and highly variable on a fine scale, with the level of alteration varying from sample to sample.

Xenolithic fragments include crustal material, derived from the local stratigraphic sequence and rare fragments of mantle peridotite. Both constitute a small fraction of the kimberlites sampled (<10 vol. %). Crustal xenoliths are recognized by the presence of pronounced disequilibrium textures with the groundmass kimberlite; these include fragments of Archaean basement gneiss, Ventersdorp lava and Mesozoic Karoo sequence rocks (basalt, dolerite, shale, mudstone and sandstone; Clement, 1982).

	ANALYTICAL METHODS

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
Samples of fresh hypabyssal kimberlite (of 3–5 kg) were cleaned and passed through a jaw crusher and then powdered with the use of a Sieb swing mill and a carbon-steel vessel. Prior to powdering, all fragments containing visible xenolithic material or carbonate veins were removed by hand picking. Major element abundances were determined by X-ray fluorescence (XRF) using a low dilution fusion technique and a Philips X'Unique wavelength spectrometer. Trace elements were determined by both XRF (on 6 g powdered briquettes) and solution inductively coupled plasma mass spectrometry (ICP-MS). Errors and detection limits associated with the XRF analyses are similar to those reported by le Roex et al. (1981). ICP-MS analyses were determined on a Perkin Elmer ELAN 6000 ICP-MS using the following analytical procedure: 50 mg of -300# sample powder were dissolved in a 3:1 HF–HNO₃ acid mixture in sealed Savilex® beakers on a hotplate for 48 h, followed by evaporation to incipient dryness and two treatments of 2 ml concentrated HNO₃. The final dried product was then taken up in 5% HNO₃ solution containing 10 ppb Re, Rh, In and Bi as internal standards. Standardization was against artificial multi-element standards. Replicate analyses of BHVO-1 typically gave an overall procedural error of better than 3% relative. Accuracy and procedural blank levels were similar to values reported by le Roex et al. (2001). CO₂ was determined for selected samples using the karbonat-bombe method of Birch (1981). Estimated precision for this technique is of the order of 5% relative.

	BULK-ROCK GEOCHEMISTRY

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
Fifty-four of the most visibly fresh kimberlite samples were selected for geochemical analysis from the De Beers pipe (14), the Wesselton pipe (16), DuToitspan (two), Bultfontein pipe (three), the Big Hole (nine) and from the Wesselton Floors sill complex (10). Major and trace element analyses of a selected subset of those analysed are reported in Table 1. In all variation diagrams the geochemical data are plotted as determined, without normalization to a volatile-free basis. The complete dataset may be downloaded from the Journal of Petrology website at http://www.petrology.oupjournals.org.

View this table: [in this window] [in a new window]   Table 1: Bulk-rock major and trace element analyses of selected hypabyssal Group I kimberlites from the Kimberley area, South Africa

 
Major elements
Major element compositions are variable, with MgO and SiO₂ correlating positively, and ranging from 14·2 to 34·3 wt % and 20·3 to 35·8 wt %, respectively (Fig. 2a), with corresponding Mg number [atomic Mg/(Mg + Fe²⁺); Fe₂O₃/FeO ratio set at 0·1] ranging from 0·67 to 0·89. Macrocrystic varieties have the highest SiO₂ and MgO contents, with aphanitic varieties having the lowest; there is some overlap in the middle of the range. Al₂O₃ and TiO₂ both show a broad negative correlation with MgO and Mg number, and have similar abundance ranges (0·5 to 5·5 wt %; Fig. 2b and c). CaO describes an excellent negative correlation with SiO₂ (Fig. 2d), ranging from 20 wt % in the most evolved aphanitic samples to 5 wt % in the most macrocryst-rich. There is no correlation between CaO and CO₂, arguing against control by calcite. FeO^* (total Fe as FeO) shows little correlation with MgO in the macrocrystic samples (7–9 wt %), but the two oxides correlate negatively in the aphanitic varieties, with FeO^* reaching 13·6 wt % in the most evolved Wesselton Floors sill samples (Fig. 2e). Alkali elements show a general scatter, although the most evolved aphanitic kimberlite samples from Wesselton Floors sill complex have K₂O abundances that are significantly higher and clearly displaced from the rest of the samples (Fig. 2f). P₂O₅ abundances show a very broad negative correlation with decreasing MgO, ranging from 0·4 to >4 wt %. CO₂ contents are variable, ranging from 1 to 12 wt %; H₂O⁺ (loss on ignition at 950°C minus CO₂) correlates positively with MgO within the aphanitic varieties, and negatively within the macrocrystic varieties, with a maximum reached in kimberlites with 27 wt % MgO (Table 1). There is no systematic difference in major element composition of the kimberlites from the different pipes, although the aphanitic samples from the Wesselton Floors sill complex and Bultfontein tend to have the most evolved (lowest Mg number) compositions.

View larger version (29K): [in this window] [in a new window]   Fig. 2. Selected major element variation diagrams for Kimberley kimberlites. Filled symbols are aphanitic kimberlite; open symbols are macrocrystic kimberlite.

 
Trace elements
Trace element concentrations are highly variable. Ni and Co show a strong correlation with Mg number, with Ni decreasing from 1500 to 350 ppm and Co from 100 to 56 ppm, with decreasing MgO (Table 1). Cr abundances are high (600– 2800 ppm), and show considerable scatter. The high field strength elements (HFSE) and light rare earth elements (LREE) are likewise high in abundance (Zr 160–910 ppm; Nb 75–400 ppm; Th 8–40 ppm; La 64–335 ppm; Ta 3·3–35·6 ppm) and show good mutual correlations (Fig. 3). In contrast, the large ion lithophile elements, although high and variable in abundance (Rb 6·6–158 ppm; Ba 300–3250 ppm; Sr 360–2260 ppm; Pb 0·7–24·8 ppm), do not show strong correlations either amongst themselves or with the HFSE. It is clear from Fig. 3 that the macrocrystic kimberlite samples (and aphanitic samples rich in small olivine phenocrysts) have lowest immobile incompatible element abundances and highest compatible (Ni, Co) element abundances (Table 1). Whereas some element pairs (e.g. Th–La, La–Ce, Zr–Hf) show tightly constrained positive correlations [e.g. Zr/Nb = 2·03 ± 0·3; Th/U = 4·0 ± 0·6; Ba/Nb = 7·9 ± 3·7; La/Th = 8·9 ± 0·8 (Fig. 3); Zr/Hf = 49 ± 4; Table 1], other pairs (e.g. Zr–La, Nb–La, Hf–La) differentiate the Wesselton Floors sills, aphanitic dykes from DuToitspan and Bultfontein, and two samples (K3/158, K3/608) from the De Beers pipe, from the majority of macrocrystic samples which have higher Nb, Zr and Hf for a given La content (Fig. 3). The aphanitic samples from Bultfontein pipe (K8/10, K8/17) and four of the aphanitic Wesselton Floors sill complex samples (K1119/2, K5/1, K5/86, K5/94) are considerably more enriched in incompatible elements than the rest of the sample suite, and show more scatter in their incompatible element abundance ratios (Fig. 3). Although compatible elements such as Ni are low in this group of rocks (<800 ppm), Cr abundances are both higher (2800 ppm) and lower (<1200 ppm) than in the majority of macrocrystic samples (1500 ppm Cr; Table 1).

View larger version (25K): [in this window] [in a new window]   Fig. 3. Variation of selected incompatible trace elements in Kimberley kimberlites. Symbols as in Fig. 2.

 
Figure 4 shows chondrite-normalized rare earth element (REE) patterns for selected samples from the various pipes. All samples are strongly LREE enriched (La/Sm_n = 5·2 ± 0·4; La/Yb_n = 126 ± 27), with La abundances in the range 300–1400 times chondrite, and Lu at 3–10 times chondrite. The aphanitic kimberlite samples (e.g. K8/17) are clearly more strongly enriched in total REE abundances, but retain REE patterns that are sub-parallel to the macrocrystic kimberlites (Fig. 4b). There is no systematic difference in the REE patterns between the kimberlite pipes, but it is clear that, amongst the analysed samples, two sets of REE pattern can be recognized, each including aphanitic and macrocrystic varieties: one group comprises samples with largely sub-parallel patterns (Fig. 4b), differing only in absolute concentration by the same relative amount across all REE (La/Yb_n = 102 ± 26), and the other comprises samples having very similar heavy REE (HREE) abundances, but with a range in LREE abundances (Fig. 4a; La/Yb_n = 163 ± 47).

View larger version (26K): [in this window] [in a new window]   Fig. 4. Chondrite-normalized REE abundances in selected Kimberley kimberlites. (a) shows a selection of samples with diverging LREE patterns, (b) shows a selection of samples with sub-parallel REE patterns. Shaded field shows the effect of 40% accumulation of garnet lherzolite, or fractionation of 40% olivine. Normalizing values from Sun & McDonough (1989). Filled symbols aphanitic; open symbols macrocrystic.

 
When normalized to primitive mantle abundances, the kimberlite samples all show strong, and largely sub-parallel, trace element enrichment patterns, with maximum values being reached in the region of Ta to La (Fig. 5). Superimposed on the smooth, sub-parallel enrichment patterns are a number of negative anomalies, the most important being Ti, Sr, K and Rb. A subset of samples also have strong positive Pb anomalies—these are attributed to crustal contamination and are not shown in Fig. 5, but are discussed further in a later section. The magnitude of these anomalies (expressed as X/X^* for element X, where X^* is the interpolated primitive mantle normalized value assuming a smooth variation between the normalized values of the two adjacent elements) is variable, with Sr/Sr^* = 0·62 ± 0·22, K/K^* = 0·25 ± 0·16 and Ti/Ti^* = 0·46 ± 0·17, but is less variable for Sr and Ti than for K and Rb. In terms of their overall primitive mantle normalized patterns, macrocrystic kimberlite samples and olivine phenocryst-rich aphanitic kimberlite samples show least absolute enrichment, whereas the more evolved aphanitic varieties from Wesselton Floors sills show strongest enrichment, with some being displaced by a substantial amount from the majority of samples (factor of four; Fig. 5). Otherwise, kimberlite samples from the five pipes show no significant, systematic, difference in their primitive mantle normalized patterns.

View larger version (32K): [in this window] [in a new window]   Fig. 5. Primitive mantle normalized trace element patterns in Kimberley kimberlites interpreted to be free of crustal contamination (see text for discussion). Symbols as in Fig. 2: open symbols are macrocrystic samples; filled symbols are aphanitic samples. Normalizing values from Sun & McDonough (1989).

 

	PETROGENESIS

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
Lying at one compositional extreme of the broad spectrum of mafic–ultramafic mantle-derived magmas, kimberlites are furthest removed from the familiar petrogenetic processes of basalt generation. Additionally, the hybrid nature of kimberlite magmas has complicated the detailed understanding of their petrogenesis. One of the most significant obstacles is considered to be the difficulty in identifying what constitutes a true ‘primary’ kimberlite magma (Mitchell, 1986). Furthermore, it is possible that ‘primary’ kimberlite magma is itself generated by complex petrogenetic processes.

The roles of late-stage deuteric alteration, fluid migration, and crustal and mantle assimilation are all important complications that need to be carefully evaluated before the mantle source region characteristics and the primary petrogenetic processes giving rise to the parental kimberlite magma can be investigated in detail. Through a systematic approach to removing, as best as possible, the effects of identifiable later stage processes, we show below that a number of important constraints can be placed on the petrogenesis of the Group I kimberlites from Kimberley.

Low-temperature alteration
The high volatile contents of kimberlite magmas and their concentration into any residual fluids upon partial solidification results in several effects that may influence the whole-rock chemistry of kimberlite. Petrographic studies of the matrix of kimberlites reveal the widespread presence of a late crystallizing assemblage dominated by cryptocrystalline serpentine (‘serpophite’) and calcite. Amoeboid pools of this residual fraction exist in various degrees of segregation from the predominantly silicate-oxide groundmass (Skinner & Clement, 1979; Clement & Reid, 1989). In extreme cases this segregation may lead to distinct carbonatitic bodies, such as those observed in the Benfontein (Dawson & Hawthorne, 1973) and Wesselton sills (Mitchell, 1984) or the Premier Mine (Robinson, 1975). The water-rich character of these high-temperature fluids results in pervasive deuteric alteration effects, particularly the extensive serpentinization of olivine macrocrysts and phenocrysts.

This late-stage, high-temperature, fluid movement has the potential to perturb the initial concentrations of fluid mobile elements such as the alkaline and alkaline earth elements (Rb, K, Ba, Sr, etc.), U and Pb. That such a process has occurred in the Kimberley kimberlite intrusions is well illustrated by contrasting the behaviour of immobile incompatible elements such as Zr, Nb, La, Hf and Th, all of which display coherent correlations (Fig. 3), with that of the mobile elements, which tend to display more scatter when plotted against an immobile element such as La (e.g. K₂O, Fig. 2f; Rb and Ba, Fig. 3e and f). Since in the absence of phlogopite crystallization, the bulk partition coefficients for K, Rb and La during crystallization should all be significantly less than one in kimberlite magma, positive correlations would be expected between these elements (even in the presence of minor perovskite, where the La partition coefficient of >1 would result only in a change of slope). Although not shown, Sr and Pb show similar variations to K and Rb, whereas U, also known to be fluid mobile, describes a very good correlation with La, indicating that its concentration has remained robust. The abundance variations of these elements, although still informative given their high absolute concentrations relative to their degree of variability, need to be used with caution in evaluating the petrogenetic history of kimberlite magmas.

Crustal contamination
A characteristic feature of most kimberlite magmas is the abundance of xenolithic fragments, and particularly crustal fragments, dislodged from the wall rock en route to the surface. The Kimberley kimberlites are no exception, and crustal fragments are ubiquitously present and evident both in hand specimen and in thin section. Although every care was taken to avoid such fragments when crushing the samples, it was impossible to be sure that none were inadvertently crushed. Moreover, given the low melting temperature of much of the local crustal material (shales, siltstone), it is more than likely that some xenolithic fragments were fully or partially assimilated by the kimberlite magma prior to solidification.

The influence of crustal assimilation on the geochemistry of the Kimberley kimberlites is well illustrated by a plot of SiO₂ vs MgO (Fig. 2a). All known upper-crustal rocks in the Kimberley stratigraphic section have higher SiO₂ and much lower MgO than kimberlite, and it is clear that a subset of samples are displaced from the main trends towards low MgO and elevated SiO₂. Another characteristic feature of crustally contaminated mafic magmas is a positive Pb anomaly (Pb/Pb^* >> 1) on a primitive mantle normalized diagram (e.g. le Roex et al., 2001). Figure 6a shows the primitive mantle normalized trace element patterns of a subset (for clarity) of the same suite of samples with elevated SiO₂ shown in Fig. 2a, and highlighted in the insert to Fig. 6b, and it is clear that all have substantial positive Pb anomalies. On a plot of Ce/Pb vs SiO₂ there is a broad negative correlation, with this same group of sample plotting at the high-SiO₂, low-Ce/Pb end of the correlation. As crustal rocks generally have higher Pb contents than mafic magmas, positive Pb anomalies can be generated through crustal contamination, but, equally, the known mobility of Pb in the weathering environment can also lead to such anomalies. As kimberlite magmas have Pb concentrations of a similar order to those in crustal rocks (10–20 ppm, Rudnick & Fountain, 1995), the cause of this positive anomaly is ambiguous. The presence of a significant positive Pb anomaly (Pb/Pb^* >1·25) in all rocks displaced from the general trend in Fig. 6b (insert) to elevated SiO₂ (but low MgO), argues for a relationship to crustal contamination. However, some samples have minor positive Pb anomalies without obvious disturbance of their SiO₂ content, suggesting also a role for late-stage fluid mobility. The variable abundances of K, Rb and Ba are similarly evident in otherwise very similar trace element patterns (Figs 2f and 3e, f), and these variations are likewise interpreted to reflect a combination of their known fluid mobile behaviour (deuteric alteration) and crustal contamination.

View larger version (26K): [in this window] [in a new window]   Fig. 6. (a) Primitive mantle normalized trace element patterns for selected crustally contaminated and uncontaminated kimberlite samples. Positive Pb anomaly and raised HREE abundances in crustally contaminated samples should be noted—the latter inferred from the raised SiO2 content [shown as filled symbols in insert to (b)]. Only one pattern of an uncontaminated sample is shown for clarity—further examples of the patterns of uncontaminated kimberlite are shown in Fig. 5. (b) Gd/Lu vs SiO2, illustrating depressed Gd/Lu ratio in crustally contaminated samples (shown as filled symbols in insert). Normalizing values from Sun & McDonough (1989).

 
Kimberlite magmas have particularly steep REE patterns, with low HREE abundances, only slightly higher than chondritic. In contrast, crustal rocks tend to have rather flat HREE patterns with absolute abundances (10 x chondrite) greater than kimberlite (Rudnick & Fountain, 1995), and contamination by such material will therefore have a greater relative impact on the HREE. This crustal influence is evident in the disturbance (shallowing) of the HREE slopes of the SiO₂- and Pb-enriched samples highlighted in Fig. 6, and the associated decreased Gd/Lu ratio in these samples (Fig. 6b). In subsequent discussions and figures, and particularly in evaluating petrogenetic models, all samples showing one or more of the above-mentioned features have been excluded as having had their compositions compromised by such open-system processes. A total of 19 out of the original 54 samples were excluded on this basis.

Aphanitic kimberlites
Although the constancy of inter-element ratios breaks down in the most evolved aphanitic kimberlites, the otherwise similar trace element ratios shown by the Kimberley kimberlites (e.g. U/Th, Zr/Nb, La/Sm, Zr/Hf) suggest derivation from a common, broadly homogeneous, mantle source region. The role for crystal–liquid fractionation in the evolution of the Kimberley kimberlite magmas is illustrated by a plot of Ni vs Mg number (Fig. 7a), as is the relationship of the aphanitic kimberlites to the macrocrystic variety. The observed continuum of compositions across the range of aphanitic kimberlite samples, extending from the least MgO-rich macrocrystic samples, is qualitatively consistent with dominant control by olivine in a continually evolving magma, rather than through partial melting, which would lead to broadly constant Ni and Mg number (e.g. Hanson & Langmuir, 1978).

View larger version (21K): [in this window] [in a new window]   Fig. 7. (a) Ni vs Mg number and (b) FeO* vs MgO in Kimberley kimberlites. Olivine fractionation curves were calculated assuming equilibrium crystallization, for olivine = 0·36 (Herzberg & O'Hara, 2002) and DNi = 124/MgO – 0·9 (Hart & Davis, 1978). Grey circle shows composition of inferred primary magma composition (see text for further discussion). Symbols as in Fig. 2.

 
It is clear from a plot of FeO^* vs MgO (Fig. 7b) that, whereas the macrocrystic kimberlite samples have relatively constant FeO^* content, there is a distinct change in slope shown by the less MgO-rich aphanitic Bultfontein and Wesselton Floors sill samples. The rapid increase in FeO^* with decreasing MgO is clearly not consistent with olivine-only control (Fig. 7b), but is broadly consistent with olivine plus calcite (abundant Fe-free groundmass phase) control. Furthermore, whereas some HFSE (Ti, Nb, Ta) describe coherent changes with Mg number and MgO across the range of kimberlite compositions, supportive of simple olivine control, this does not hold for Zr, La or Th, for which the correlation breaks down in the most evolved aphanitic Wesselton Floors sills kimberlite samples. This decoupling requires processes more complex than simple olivine control in the origin of at least some of the highly evolved aphanitic Kimberley kimberlites. Given that this breakdown in correlation occurs only in the most evolved samples from sills, or cross-cutting late-stage dykes, we suggest that it might result from complex crystallization processes within a stagnant magma body, where the final residual melt, in equilibrium with late-stage minor accessory phases such as apatite, monazite, zircon and phlogopite (e.g. Mitchell, 1995), in addition to abundant olivine and calcite, is expelled from the host crystalline matrix.

Primitive mantle normalized plots show that the highly aphanitic kimberlites have overall trace element patterns uniformly displaced to higher concentrations, relative to those shown by the macrocrystic kimberlites, by a factor of up to four across the range from La to Yb (Fig. 5). Such sub-parallel enrichment is more readily achieved by fractionation processes en route to solidification than by partial melting. Given that for a garnet lherzolite residue, the bulk D values for these elements are of a similar order of magnitude to the likely degree of melting (F <2%, e.g. Dalton & Presnall, 1998), the absolute concentrations are not particularly sensitive to degree of melting. For example, a change in degree of melting by a factor of four (e.g. from 2% to 0·5%), will cause only a two-fold increase in highly incompatible element abundances (e.g. La), but more importantly, in the presence of residual garnet, less incompatible elements such as Yb will remain largely constant or even decrease in abundance (this is illustrated in detail in a later section; see Fig. 9). To generate a uniform three- to four-fold enrichment across the full spectrum of incompatible elements (including HREE) through partial melting processes in the garnet stability field is thus unlikely, and is more readily achieved through crystal fractionation processes in the absence of garnet. However, to achieve even a 2·5-fold enrichment in incompatible elements, starting from a Mg number of 0·86 and assuming completely incompatible element behaviour (bulk D <0·001), requires at least 60% fractional crystallization. To avoid Ni being completely depleted if olivine is the dominant component of the fractionating assemblage (Fig. 7a), phases other than olivine (with lower D^Ni) to decrease the bulk D (e.g. calcite) must also be involved in the process. The geochemical evidence is broadly consistent with the aphanitic Kimberley kimberlites being the result of extensive crystal–liquid fractionation, via complex congelation crystallization in stagnant magma bodies, of a parental kimberlite magma more MgO-rich than the least evolved aphanitic sample analysed.

View larger version (34K): [in this window] [in a new window]   Fig. 9. Rare earth element forward melting model. Source composition calculated from sample Col 1 (filled symbols) assuming 1·5% equilibrium melting. The range of calculated melt compositions from this source for F = 0·4–1·5% is shown by the dark shaded field. Source mineralogy and melting proportions are given in Table 2. Partition coefficients are given in Table 3. The diagonally shaded field illustrates the impact of different amounts of residual garnet (2% and 5%) on the HREE in the calculated source, and of variable degrees of melting (0·5–1·5%) on the LREE in the calculated source. Chondritic values from Sun & McDonough (1989). The field of Kaapvaal Craton garnet lherzolites is from Gregoire et al. (2003). (See text for further discussion.)

 
Mantle entrainment and macrocrystic kimberlite
Olivine macrocrysts, ubiquitous to most kimberlite magmas, are commonly believed to represent disaggregated mantle-derived xenoliths in view of their anhedral habit similar to that observed in peridotite nodules, and the common deformation features (e.g. Clement, 1982; Shee, 1985) that are not expected to be sustained in a liquid. This qualitative view is supported by the occasional presence of peridotitic garnet, phlogopite and Cr-diopside xenocrysts. The rarity of orthopyroxene has been argued to be a consequence of reaction with, and complete resorption by, the highly silica-undersaturated host kimberlite magma (Shee, 1985). The origin of olivine macrocrysts is central to determining the composition of the primary Kimberley kimberlite magma(s) and is investigated further by considering the geochemical impact of mantle entrainment and partial assimilation.

Figure 8 shows the variation of Ni and Sc with respect to SiO₂ in the Kimberley kimberlites, with macrocrystic and aphanitic kimberlites being distinguished. Also shown are representative compositions for olivine, orthopyroxene, garnet and Cr-diopside from mantle xenoliths from the Kimberley region (Gregoire et al., 2003). It is clear that the macrocrystic kimberlite samples define a trend that is at an angle to that defined by the aphanitic kimberlite samples and does not extend directly towards olivine (as would be expected if the olivine macrocrysts represented accumulated olivine) but requires a component poorer in Ni. In the Sc–SiO₂ diagram the need for a phase other than olivine, and specifically one containing Sc, is equally evident. The trend defined by the macrocrystic kimberlite samples is best accounted for by entrainment of garnet lherzolite with an average composition of 57% olivine, 25% orthopyroxene, 12% clinopyroxene and 6% garnet. By application of the lever rule, up to 35% entrainment of garnet lherzolite is required to account for the spectrum of macrocrystic kimberlite compositions, which is in good agreement with the observed abundances of olivine macrocrysts.

View larger version (24K): [in this window] [in a new window]   Fig. 8. Ni and Sc vs SiO2 in Kimberley kimberlites. Aphanitic kimberlites with <25 wt % SiO2 are not shown. Representative compositions of olivine, orthopyroxene, garnet and clinopyroxene are from Bultfontein garnet lherzolite (Gregoire et al., 2003). The position and composition of the marked garnet lherzolite was determined by use of the lever rule. Symbols as in Fig. 2.

 
Primary magma composition
Accepting the above arguments implies that the primary kimberlite magma composition, prior to lherzolite entrainment and partial assimilation, would lie at the low-SiO₂ and -MgO end of the macrocrystic Kimberley kimberlites, and at the high-SiO₂ and -MgO end of the aphanitic kimberlites, corresponding to the break in slope between the macrocrystic and aphanitic kimberlite varieties on many variation diagrams (e.g. SiO₂–MgO, Fig. 2a; FeO–MgO, Fig. 7b). This composition has 26–27 wt % SiO₂, 26–27 wt % MgO, Mg number 0·86, 2·2 wt % Al₂O₃, 2·2 wt % TiO₂, 1–2 wt % K₂O, 12 wt % CaO, 7 wt % CO₂ and 8 wt % FeO^* as read from variation diagrams (samples most similar to this compositional point include K119/2, K6/55 and K8/115). It is noteworthy that the SiO₂, MgO and Al₂O₃ composition of this inferred primary kimberlite magma corresponds closely to that predicted for 0·6–0·7% melting by interpolation of the experimental data of Dalton & Presnall (1998). The CaO content of the inferred primary magma (12 wt %) is, however, significantly lower than that of the experimental kimberlite melt (18 wt %, Dalton & Presnall, 1998).

The high Mg number of the proposed primary Group I Kimberley kimberlite magma suggests equilibration against a highly refractory mantle source composition (Fo₉₅ olivine) and would require that early phenocrysts would be very Fo-rich (Fo₉₅, assuming an olivine–melt K^D_Fe–Mg = 0·36, Herzberg & O'Hara, 2002). These compositions match those found in cratonic mantle xenoliths (Fo_92–94, Gregoire et al., 2003) and the most Mg-rich phenocrysts in the analysed sample suite (Fo₉₅). The calculated Ni content of such a primary magma would be 950–1000 ppm, assuming a refractory garnet lherzolite source with 2500 ppm Ni, and using partition coefficients and data from Hart & Davis (1978), Hirschmann & Ghiorso (1994) and Gregoire et al. (2003), consistent with the inferred composition plotted in Fig. 7a.

Partial melting
The role of partial melting and source mineralogy in controlling the compositions of kimberlite magmas, and thus ultimately providing information on the source region composition of these unusual magmas is of considerable interest. Experimental evidence (Wyllie, 1980, 1987; Eggler, 1987; Kesson et al., 1994; Dalton & Presnall, 1998; Wyllie & Lee, 1999) has shown that kimberlite magmas can be produced by partial melting of carbonated garnet lherzolites at P >5 GPa. The study by Dalton & Presnall (1998) in the system CaO–MgO–Al₂O₃–SiO₂–CO₂ has shown that at 200 km depth, melting of carbonate-bearing garnet lherzolite gives rise to a continuum of melt compositions ranging from carbonatite (0·3% melting) to kimberlite (1% melting). Although this result does not necessarily imply that kimberlite forms by partial melting of carbonated peridotite, the experiments do suggest that, at the point of segregation from its source, the kimberlite magma would approximate a 1% partial melt. Such low degrees of melting are consistent with the strong enrichment in incompatible elements in kimberlites, and the relative depletion in HREE. We note that 1% melting implies that the CO₂ content of the mantle source would be 700 ppm, assuming a primary magma concentration of 7 wt % CO₂ (see above). Modelling of REE abundances in Group II kimberlites and lamproites led Tainton & McKenzie (1994) to the conclusion that 0·3–0·4% melting of garnet lherzolite can account for the HREE abundances in Group II kimberlite. In their study of the Group II Finsch kimberlite, Fraser & Hawkesworth (1992) suggested that the coupled Sm/Nd and Nd-isotope systematics could be accounted for by 1% partial melting of a garnet lherzolite mantle source.

What is less clear is the role of accessory phases, and small variations in the degree of melting, in controlling the trace element inventory of these extreme magma compositions. It was noted above that, in terms of chondrite-normalized REE patterns, two groups of kimberlites can be recognized within the Kimberley cluster, one having slightly greater La/Yb_n ratio (163 ± 47) and divergent LREE patterns, the other sub-parallel patterns with La/Yb_n = 102 ± 26. Figure 4 illustrates this difference, which cannot be attributed to olivine fractionation–accumulation or to garnet lherzolite entrainment–assimilation, as neither process is able to effect the necessary fractionation of the LREE from the HREE. For example, the shaded region in Fig. 4b illustrates the effect of 40% lherzolite addition or olivine removal (in detail, olivine addition would cause uniform depletion in all REE, whereas lherzolite entrainment would dilute the LREE to a slightly greater extent than the middle REE (MREE) and HREE in view of their low, but not insignificant, abundances in clinopyroxene and garnet, respectively, but these differences are negligible relative to the scale of the diagram). It is evident that the sub-parallel displacement of the REE patterns in Fig. 4b is consistent with up to 40% olivine or lherzolite entrainment, which is not able to effect the necessary fractionation shown by the divergent patterns in Fig. 4a. By contrast, it can be shown that the LREE–HREE fractionation evident in the latter group of samples can be accounted for by a four-fold difference in the degree of partial melting (at low F; 0·4–1·5% melting). This is illustrated in Fig. 9 by forward modelling of the REE abundances as follows: the mantle source composition capable of giving rise to the REE pattern shown by sample COL1 has been calculated assuming 1·5% non-modal equilibrium melting of garnet lherzolite (source mineralogy and melting parameters are given in Table 2; partition coefficients are given in Table 3). Melting of this same source over the range 1·5–0·4% (shaded region in Fig. 9) gives chondrite-normalized REE patterns effectively equivalent to the range in LREE enrichment seen in the Kimberley kimberlites, i.e. at the degrees of melting in question, the 2% residual garnet in the source region is sufficient to strongly buffer the HREE whereas the LREE behave as incompatible elements and vary by a factor of 2·5–3. Clearly, the absolute degrees of melting used above are model dependent and are not unique, but are consistent with experimental evidence for the likely degree of melting giving rise to kimberlite magmas (e.g. Dalton & Presnall, 1998).

View this table: [in this window] [in a new window]   Table 2: Parameters used in the melting calculations and calculated source composition illustrated in Fig. 10

 

View this table: [in this window] [in a new window]   Table 3: Partition coefficients used in quantitative modelling

 

View larger version (30K): [in this window] [in a new window]   Fig. 10. Multi-element forward melting model for a close to primary Kimberley kimberlite magma composition. Error bars reflect the negligible effect of ±15% (range 30%) garnet lherzolite entrainment on the plotted composition. An evolved aphanitic kimberlite composition is shown for comparison. The source composition is calculated from the close to primary kimberlite composition assuming 1% equilibrium non-modal melting. Normalizing values from Sun & McDonough (1989). Source mineralogy, melting mode and calculated source composition are given in Table 2. Partition coefficients are given in Table 3. The field of Kaapvaal garnet lherzolites and the composition of a deformed garnet lherzolite BD2421 are from Gregoire et al. (2003) and are shown for comparison. (See text for further discussion.)

 
The above model (which uses realistic source mineral modes, published partition coefficients and reasonable estimates of melting degree) is sufficient to account for the variable LREE–HREE fractionation seen in the Kimberley kimberlites, and demonstrates that the coherent REE geochemistry conforms to expected silicate melt partitioning, and illustrates that exotic REE-bearing accessory phases occurring in many Bultfontein mantle xenoliths (e.g. Haggerty, 1983; Erlank et al., 1987; Waters & Erlank, 1988) are not required in the source region. All that is required is an enriched mantle source, with the LREE enriched by a factor of 8 x chondrite and the HREE by a factor of 0·4xchondrite. Greater residual garnet has the impact of raising the required HREE abundances in the source; for example, 5% residual garnet results in the source HREE abundances increasing to 1xchondrite, for the same degree of melting (Fig. 9). The calculated REE composition of the source falls within that shown by Kaapvaal craton garnet lherzolite xenoliths (Fig. 9).

Other than the overall strong enrichment in incompatible elements, the most striking features of the trace element compositions of the Kimberley kimberlites are the marked negative anomalies in Ti, K, Sr and Rb when normalized to primitive mantle abundances (Fig. 5). These anomalies could result variously from (1) late-stage element mobility, (2) crystal–melt fraction processes en route from source to surface, (3) partial melting processes, or (4) could be inherited from the source region. Although it is possible that more than one have played some role, when considering all the evidence available the last alternative is thought to dominate; our reasons for reaching this conclusion are developed systematically below.

Interpretation of the negative K and Rb anomalies is complicated by the demonstrated mobility of the alkali elements. However, given the ubiquitous occurrence of such anomalies in all ‘fresh’ Group I kimberlites, and indeed most highly alkaline magmas such as nephelinites, melilitites, lamprophyres, basanites and carbonatites where late-stage alteration might not be as significant (e.g. Rogers et al., 1992; Späth et al., 1996, 2001; Class & Goldstein, 1997; le Roex & Lanyon, 1998; le Roex et al., 2001), coupled with the scattered but systematic differences in K/K^* between individual pipes, we argue that the negative K and Rb anomalies are a primary feature of these Group I kimberlite magmas, although in some instances the absolute magnitudes (scatter) might have been affected to a degree by late-stage alteration processes.

If the primary kimberlite magma(s) to the Kimberley kimberlites had no initial negative K anomaly, then by reference to Fig. 5, and abundances of elements of similar incompatibility to K, K₂O content of such primary magma(s) would have been of the order of 6–7·5 wt %. To reduce this concentration to the observed 1–2 wt % would require >50% phlogopite-only fractionation. Phlogopite is a rare, minor matrix phase in these samples (sample K6/55 excepted), and in the absence of experimental evidence to the contrary, substantial phlogopite fractionation is unlikely to be the cause of the significant relative depletion in K. Consequently, the negative K anomaly is interpreted to be a feature of the primary kimberlite magma(s).

Rogers et al. (1992) attributed the strong negative K anomaly in Western Cape melilitites to equilibration with residual phlogopite, whereas Späth et al. (2001) showed that residual amphibole is required in the lithospheric source of the Chyulu Hills nephelinites, and argued that absolute K contents of these magmas were too low to have been in equilibrium with phlogopite. As K is a stoichiometric component of phlogopite, the K content of a melt in equilibrium with residual phlogopite will be constant as long as phlogopite is residual, at a concentration dependent only on the proportion of phlogopite entering the melt (Späth et al., 2001). If mantle phlogopite contains 9 wt % K₂O (e.g. Gregoire et al., 2002), and has a melt mode of 50–70% (Wass & Rogers, 1980; Greenough, 1988), the melt in equilibrium with phlogopite should contain at least 36 000 ppm K. The K content of the Kimberley kimberlite magmas is <16 000 ppm, too low to have been in equilibrium with residual phlogopite. Furthermore, Ulmer & Sweeney (2002) have argued that phlogopite is not stable at the kimberlite liquidus at pressures greater than 4–5 GPa. Consequently, although residual phlogopite during melting is capable of generating the requisite relative depletion in K, low absolute K abundances in the kimberlite magma coupled with experimental evidence suggest that the negative K and Rb anomalies are intrinsic features of the mantle source and are not a result of buffering against residual phlogopite.

The magnitude of the negative Ti anomaly (Ti/Ti^* = 0·2–0·8) is greatest in the evolved aphanitic samples from Bultfontein, DuToitspan and Wesselton Floors sills, but does not correlate with degree of differentiation as measured by Mg number or incompatible element abundances. Moreover, the absence of negative Nb and Ta anomalies in the Kimberley kimberlites argues against ilmenite or perovskite fractionation as a cause of the anomaly. Although there is a slight increase in La/Nb ratio with decreasing Ti/Ti^* ratio across the full sample suite, the increase is controlled largely by the variation within the evolved Wesselton Floors sill samples. Similarly, the negative Sr anomaly does not appear to correlate with any index of differentiation, although, as for Ti, the aphanitic kimberlites from Bultfontein, DuToitspan and Wesselton Floors sills have slightly larger anomalies (Sr/Sr^* 0·5 compared with 0·7). The origin of these anomalies appears therefore to relate either to partial melting process or existed in the source prior to melting. Consequently, we propose that the Ti and Sr anomalies are a feature of the primary kimberlite magma(s).

To investigate the trace element characteristics of a possible source to the Kimberley kimberlites, the modelling approach used for the REE can be extended to the full range of incompatible elements as illustrated in Fig. 10. In this forward modelling exercise a value of 1% melting is assumed to give rise to the trace element abundances characteristic of a primary Kimberley kimberlite magma, which in turn allows the calculation of an inferred source composition. A more sophisticated modelling approach utilizing inversion techniques as developed by Hofmann & Feigenson (1983) is not justified, given the absence of a suite of suitable primary magma compositions, and the magnitude of the assumptions involved in trying to back-calculate primary magma compositions from samples variably affected by lherzolite entrainment and/or fractionation. The required source composition was then calculated using melting parameters and partition coefficients given in Tables 2 and 3. On the grounds of arguments presented above, phlogopite is not included as a residual mantle phase. The assumed Kimberley primary magma composition was chosen to reflect least evolved aphanitic kimberlite and least MgO-rich macrocrystic kimberlite. Given the high absolute trace element abundances and log scale used, the effect of 15% change in abundances (e.g. from minor lherzolite entrainment or olivine fractionation or accumulation) is negligible, as shown by the size of the error bars in Fig. 10, and does not affect our arguments.

The calculated model source using the above-mentioned parameters has highly incompatible elements enriched by a factor of 2–5 x primitive mantle abundances, and, depending on residual modal garnet content, the less incompatible elements depleted by a factor of 0·2–0·5 x primitive mantle abundances. It is clear that the negative K, Rb, Sr and Ti anomalies in the kimberlite magmas are transferred to the calculated source in the absence of phases capable of buffering their concentrations. Of interest is that the predicted source plots within the field of garnet lherzolites from the Kaapvaal craton, as determined by Gregoire et al. (2003).

The absence of a negative Nb (+ Ta) anomaly in the Kimberley Group I kimberlites supports the contention that equilibration against an oxide phase (ilmenite, rutile, armalcolite, lindsleyite) is not the cause of the negative Ti anomaly. However, the partition coefficient for Ti in clinopyroxene is sensitive to pressure; although Adam & Green (1994) showed a decrease with increasing pressure (at least to 3 GPa), it increases with increasing ^IVAl content of clinopyroxene (Hill et al., 2000), and it increases substantially with the carbonate content of the melt (Blundy & Dalton, 2000). Given its high CO₂ content, primary kimberlite magma is likely to have a significant carbonate component, and the Ti–cpx partition coefficient could be considerably higher than that assumed in Table 3 (appropriate for basanite petrogenesis). Furthermore, Baker et al. (1995) have shown that at low melt fractions just above the solidus, Ti clinopyroxene–melt partition coefficients are larger than at high degrees of melting. The negative Ti anomaly is thus conceivably due to a Ti–cpx partition coefficient greater than used in these calculations—a value of 0·35 would account for the observed anomaly at the assumed degree of melting, without recourse to a source anomaly. A broad negative correlation between Ti/Eu ratio (or Ti/Ti^*) and La/Yb ratio supports a melting control on this anomaly. However, we note that negative Ti (and Sr) anomalies are nearly ubiquitous in analysed mantle xenoliths from the Kaapvaal craton (Gregoire et al., 2003)

From the above discussion, the extreme trace element compositions of the Group I Kimberley kimberlites can be accounted for by low degrees of melting (1%) of a moderately enriched garnet lherzolite source not unlike that found as xenoliths in the host kimberlites. Although non-unique, this model is intended to illustrate that simple silicate–melt partitioning and a moderately enriched source region can account for the overall trace element abundances in Group I kimberlite magmas, at degrees of melting consistent with experimental evidence. It may nevertheless also be possible to generate the requisite major and trace element characteristics by more complex processes that we have not explored here, such as percolative zone refining or chromatographic effects. If phlogopite is initially present in the source region, it appears not to survive the melting episode and is fully consumed into the melt. Alternatively, phlogopite is sufficiently randomly distributed in the source region that it contributes to some melt fractions and not others before accumulation and mixing of the incipient magmas into a volume sufficiently large to escape to the surface; in essence, K might have an intermediate behaviour between a stoichiometric component and an incompatible trace element.

	DISCUSSION

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
The trace element geochemistry of the Kimberley Group I kimberlites is surprisingly coherent given the hybrid nature of kimberlite magmas, and the extensive deuteric alteration suffered by most. Semi-quantitative modelling of partial melting processes indicates that many features of the trace elements are adequately explained by silicate melt equilibration at low degrees of melting (1%) of garnet lherzolite. Source compositions required to give rise to the average Kimberley Group I kimberlite are enriched in highly incompatible trace elements, but not excessively so (2–5 x primitive mantle abundances for the most incompatible elements), but HREE are depleted relative to chondrite and primitive mantle values, indicating a prior depletion event. Such a source composition falls well within the observed range of moderately metasomatized mantle from the region (Gregoire et al., 2003), although isotopic criteria suggest that the continental lithospheric mantle is not the ultimate source of incompatible trace elements (e.g. Smith, 1983).

Numerous studies of metasomatized xenoliths from the lithospheric mantle beneath the Kaapvaal craton have demonstrated the importance of a range of exotic phases such as K-richterite, lindsleyite, mathiasite and armalcolite in addition to more common silicate minerals such as diopside and phlogopite (e.g. Haggerty, 1983; Erlank et al., 1987; Waters & Erlank, 1988). Collectively, these metasomatic minerals act as hosts to elements such as Rb, Ba, Sr, K, Nb, Ta, LREE and Ti. Two important observations in this regard are: (1) that one would not expect such a metasomatized source to show relative depletion in elements such as Rb, K, Sr and Ti as appears to be required by the Kimberley kimberlites; (2) the absence of any Nb or Ta anomaly in the kimberlites argues against any of the titanium-bearing phases remaining in the residue after kimberlite melt extraction. The experimental observation that near-solidus liquids in the peridotite– CO₂ system can be kimberlitic (e.g. Dalton & Presnall, 1998; Wyllie & Lee, 1999), and the attendant implication that kimberlite could be formed by the melting of carbonated peridotite, suggests that metasomatism of the source region was effected by a CO₂-rich fluid. Direct evidence for such carbonatite metasomatism has been found, for example, in mantle xenoliths from Tanzania (Rudnick et al., 1993). Assuming 7% CO₂ in the primary kimberlite magma and 1% partial melting would indicate a source CO₂ content of 700 ppm.

A compositional characteristic of carbonatite magmas is their relative depletion in HFSE and alkali elements (Ti, Hf, Zr, K, Rb), and whereas some of these traits are seen in the predicted source, i.e. low Ti, K and Rb, the source is not obviously depleted in Zr or Hf. A logical mineral phase to deplete a metasomatic fluid in K, Rb and Ti is phlogopite. However, experimental evidence on the stability of phlogopite in the presence of carbonated fluids is conflicting. Wendlandt & Eggler (1980) showed phlogopite + magnesite stable at 1200°C at pressures of 4–5 GPa, and Eggler & Wendlandt (1979) showed phlogopite stable at 5·5 GPa at sub-solidus temperatures, coexisting with lherzolite + magnesite + vapour at 1200°C and X_CO2 = 0·3, but with a considerably reduced phlogopite phase volume relative to 3 GPa. In contrast, Ulmer & Sweeney (2002) argued that phlogopite is not stable at near-solidus temperatures above 4–5 GPa in the presence of carbonate, with X_CO2 = 0·37. Although recognizing the conflict in the experimental results, at least common to both sets of experiments is that abundant CO₂ appears to decrease the phlogopite phase volume at high pressures (>4–5 GPa). The strong negative Rb and K anomalies in the inferred mantle source region are therefore interpreted to suggest that phlogopite was involved in the early stages of deep mantle metasomatism, causing a depletion in Rb, K and Ti in the metasomatic fluid, which in turn imposed negative K, Rb and Ti anomalies on the overlying incipient kimberlite source region. Melting of this latter metasomatized source to give rise to the Kimberley Group I kimberlites would result in the observed negative anomalies being transferred to the primary magmas. Although the details of the precise metasomatic history still require working out, the overall model is broadly in agreement with the scenario proposed previously for kimberlite generation by Girnis et al. (1995).

With the exception of their strong HREE depletion, the primitive mantle normalized trace element patterns for these Group I kimberlites are not very different from those of most other alkaline rock types. Figure 11 shows a typical range of compositions of primitive nephelinite (Chyulu Hills, Kenya), melilitite (Western Cape, South Africa) and ultramafic lamprophyre (northwestern Namibia), and it is clear that the major difference is the significantly lower HREE abundances in the Group I kimberlites. Negative Ti and Sr anomalies are more subdued or absent in the nephelinite, melilitite and lamprophyre magmas, whereas the negative K anomaly is comparable with that in the kimberlite magmas. The similar absolute enrichment in highly incompatible elements suggests derivation from source regions that have experienced a similar degree of metasomatic enrichment, whereas the depletion in HREE abundances indicates generation from a source that was previously more strongly depleted—we interpret this as evidence for involvement of deep cratonic continental lithospheric mantle in Group I kimberlite genesis.

View larger version (19K): [in this window] [in a new window]   Fig. 11. Primitive mantle normalized trace elements in average Wesselton kimberlite compared with the field of other highly alkaline rock types, such as nephelinite, melilitite and lamprophyre. Source of data: nephelinites (Chyulu Hills, Kenya) from Späth et al. (2001); melilitites (Western Cape, South Africa) from P. E. Janney & A. P. le Roex (unpublished data, 2003); ultramafic lamprophyres (northwestern Namibia) from le Roex & Lanyon (1998).

 
Smith (1983) argued on isotopic grounds that the source of Group I kimberlites from southern Africa lies in the asthenospheric mantle. le Roex (1986) proposed a plume origin for the kimberlites on the basis of hotspot tracks and trace element and isotopic tracers. Following the arguments of Hofmann et al. (1986), Nb/U and Ce/Pb ratios determined in the present study confirm the sub-lithospheric [ocean island basalt (OIB)-like] origin for the incompatible trace element complement of these kimberlites (Fig. 12). Other incompatible trace element ratios (e.g. La/Th <12; Nb/Th <15) also illustrate their similarity to OIB compositions, and distinction from depleted South Atlantic asthenospheric [mid-ocean ridge basalt (MORB)] mantle (La/Th >12, Nb/Th >15, le Roux et al., 2002; A. P. le Roex unpublished data, 2003). Furthermore, these trace element ratios are clearly different from those of Group II kimberlites believed to represent direct lithospheric melts (e.g. Smith, 1983), and Karoo flood basalts which, although their origin is more contentious, are also argued by many (e.g. Hawkesworth et al., 1984) to represent melts of the sub-Gondwana lithospheric mantle (Fig. 12).

View larger version (25K): [in this window] [in a new window]   Fig. 12. Ce/Pb and Nb/U ratios in Kimberley kimberlites. Field for ocean island basalt (OIB) from the South Atlantic is from le Roex (1985), le Roex et al. (1990) and A. P. le Roex (unpublished data, 2003). Mid-ocean ridge basalt (MORB) field from the South Atlantic is from le Roux et al. (2002), Karoo flood basalt field is from Marsh et al. (1997) and Group II kimberlite field is from A. P. le Roex (unpublished data, 2003). Crustal values from Rudnick & Fountain (1995). Symbols as in Fig. 2.

 
The HREE, Mg number and Ni requirement for derivation from a source that had experienced prior geochemical depletion suggests that the Kimberley Group I kimberlites do not represent direct melts of an upwelling plume mantle, but reflect a two-stage process, described above, in which fluids derived from an upwelling mantle plume initially metasomatize the overlying (depleted) sub-continental lithospheric mantle, and thus introduce their geochemical signature into the sub-continental lithospheric mantle source of the kimberlite magmas. There is a substantial difference between the inferred thickness of the cratonic mantle root beneath Kimberley (300 km, James et al., 2001) and the deepest xenoliths in kimberlite magmas (200 km). This suggests that about 100 km of lithospheric mantle may have been affected by porous flow that resulted ultimately in kimberlite eruption. Similar flow fronts have been documented in the exposed continental mantle section of the Ronda massif in Spain (Lenoir et al., 2002). The partial equilibration with subcontinental lithosphere can also explain differences in Sr, Nd and Pb isotope composition between mafic–ultramafic lavas erupted in shallow off-craton regions in southern Africa and cratonic kimberlites (Janney et al., 2002).

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
Analysis of fresh, hypabyssal Group I kimberlite from the major kimberlite pipes in the Kimberley area, South Africa, shows that their trace element variations are remarkably coherent, and allows a number of constraints to be placed on their petrogenesis, as follows.

1. Whereas the HFSE and REE abundances appear undisturbed by post-melting processes, the alkali and alkaline earth elements have suffered disturbance through late-stage residual fluid segregation and deuteric alteration processes. Some samples show clear evidence for crustal contamination in their raised SiO₂, Al₂O₃, HREE and Pb abundances.
   
2. The abundant olivine macrocrysts in macrocrystic kimberlite result from the entrainment and disaggregation of partially assimilated garnet peridotite xenoliths; the range in macrocrystic kimberlite compositions can be accounted for by up to 35% addition of garnet peridotite. In contrast, aphanitic kimberlites (forming sills and late-stage dykes) have compositions inconsistent with simple olivine fractionation, and require more complex crystallization processes, including equilibration with accessory phases. The primary magma responsible for the Kimberley Group I kimberlites is inferred to have a composition of 26–27 wt % SiO₂, 26–27 wt % MgO, 2·2 wt % Al₂O₃, 12 wt % CaO, 8 wt % FeO^* and K₂O1–2 wt %, with Mg number 0·86, corresponding to the change in composition from aphanitic kimberlite to macrocrystic kimberlite.
   
3. Simple batch melting models show that average, close-to-primary, Group I Kimberley kimberlites could have formed by variable degrees (0·4–1·5%) of partial melting of a previously metasomatized garnet lherzolite source. The metasomatized source requires enrichment in highly incompatible elements (by 2–5 x primitive mantle abundances), and depletion in the HREE (0·3 x primitive mantle abundances, but dependent on amount of residual garnet). The low HREE abundances, and high Mg number and Ni contents of the inferred primary magma are interpreted to indicate that the garnet lherzolite source had experienced a depletion event prior to this metasomatic enrichment. If phlogopite or other exotic metasomatic phases are present in the initial source, they do not appear to survive the melting event.
   
4. Incompatible trace element ratios in these Group I kimberlites are consistent with derivation from sub-lithospheric mantle similar to that giving rise to some South Atlantic OIB. This suggests a two-stage process whereby fluids derived from an upwelling mantle plume initially metasomatize the overlying sub-continental lithospheric mantle (e.g. Wyllie, 1987), and thus introduce their geochemical signature into the lower reaches of the geochemically depleted subcontinental lithospheric mantle. Subsequent melting of this metasomatized source gives rise to the primary kimberlite magma carrying OIB-like incompatible trace element ratios.
   

	ACKNOWLEDGEMENTS

 
We are grateful to De Beers for providing access to their sample sheds, and particularly Dr Jock Robey for his advice on sample localities and for his hospitality in Kimberley. De Beers Consolidated Mines (Ltd), the University of Cape Town and the National Research Foundation provided financial support for this research. Reviews by Nick Rogers, Balz Kamber, Don Francis and Fred Frey contributed greatly to improving the final version of this manuscript, and are gratefully acknowledged.

	REFERENCES

TOP ABSTRACT INTRODUCTION GEOLOGICAL SETTING AND SAMPLING PETROGRAPHY ANALYTICAL METHODS BULK-ROCK GEOCHEMISTRY PETROGENESIS DISCUSSION CONCLUSIONS REFERENCES

 
Adam, J. & Green, D. H. (1994). The effects of pressure and temperature on the partitioning of Ti, Sr and REE between amphibole, clinopyroxene and basanitic melts. Chemical Geology 117, 219–233.[CrossRef][Web of Science]

Baker, M. B., Hirschmann, M. M., Ghiorso, M. S. & Stolper, E. M. (1995). Compositions of near-solidus peridotite melts from experiments and thermodynamic calculations. Nature 375, 308–311.[CrossRef]

Beard, A. D., Downes, H., Hegner, E., Sablukov, S. M., Vetrin, V. R. & Balogh, K. (1998). Mineralogy and geochemistry of Devonian ultramafic minor intrusions of the southern Kola Peninsula, Russia: implications for the petrogenesis of kimberlites and melilitites. Contributions to Mineralogy and Petrology 130, 288–303.[CrossRef][Web of Science]

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

Birch, G. F. (1981). The karbonat-bombe; a precise, rapid and cheap instrument to determine calcium carbonate in sediments and rocks. Transactions of the Geological Society of South Africa 84, 199–203.

Blundy, J. & Dalton, J. (2000). Experimental comparison of trace element partitioning between clinopyroxene and melt in carbonate and silicate systems, and implications for mantle metasomatism. Contributions to Mineralogy and Petrology 139, 356–371.[CrossRef][Web of Science]

Brey, G. P. & Ryabchikov, I. D. (1994). Carbon dioxide in strongly silica undersaturated melts and origin of kimberlite magmas. Neues Jahrbuch für Mineralogie, Monatshefte 10, 449–463.

Class, C. & Goldstein, S. L. (1997). Plume–lithosphere interactions in the ocean basins: constraints from the source mineralogy. Earth and Planetary Science Letters 150, 245–260.[CrossRef][Web of Science]

Clement, C. R. (1982). A comparative geological study of some major kimberlite pipes in the Northern Cape and Orange Free State. Ph.D. thesis, University of Cape Town.

Clement, C. R. & Reid, A. M. (1989). The origin of kimberlite pipes: an interpretation based on a synthesis of geological features displayed by southern African occurrences. In: Ross, J., Jaques, A. L., Ferguson, J., Green, D. H., O'Reilly, S. Y., Danchin, R. V. & Janse, A. J. A. (eds) Kimberlites and Related Rocks, Vol. 14. Sydney: Geological Society of Australia, pp. 632–646.

Clement, C. R. & Skinner, M. W. (1979). Textural genetic clasification of kimberlites. Transactions of the Geological Society of South Africa 88, 403–410.

Clement, C. R., Skinner, E. M., Hawthorne, J. B., Kleinjan, L. & Allsopp, H. L. (1979). Precambrian ultramafic dykes with kimberlite affinities in the Kimberley area. In: Boyd, F. R. & Meyer, H. O. A. (eds) Kimberlites, Diatremes and Diamonds: their Geology, Petrology and Geochemistry. Proceedings of the Second International Kimberlite Conference, Vol. 1. Washington, DC: American Geophysical Union, pp. 101–110.

Dalton, J. A. & Presnall, C. (1998). The continuum of primary carbonatitic–kimberlitic melt compositions in equilibrium with lherzolite: data from the system CaO–MgO–Al₂O₃–SiO₂–CO₂ at 6 GPa. Journal of Petrology 39, 1953–1964.[CrossRef][Web of Science]

Dawson, J. B. & Hawthorne, J. B. (1973). Magmatic sedimentation and carbonatitic differentiation in the kimberlite sills at Benfontein, South Africa. Journal of the Geological Society, London 129, 61–85.[Abstract/Free Full Text]

Eggler, D. H. (1978). Stability of dolomite in a hydrous mantle, with implications for the mantle solidus. Geology 6, 397–400.[Abstract/Free Full Text]

Eggler, D. H. (1987). Discussion of recent papers on carbonated peridotite, bearing on mantle metasomatism and magmatism: an alternative. Earth and Planetary Science Letters 82, 398–400.[CrossRef][Web of Science]

Eggler, D. H. & Wendlandt, R. F. (1979). Experimental studies on the relationship between kimberlite magmas and partial melting of peridotite. In: Boyd, F. R. & Meyer, H. O. A. (eds) Kimberlites, Diatremes and Diamonds: their Geology, Petrology and Geochemistry. Proceedings of the Second International Kimberlite Conference, Vol. 1. Washington, DC: American Geophysical Union, pp. 330–338.

Erlank, A. J., Waters, F. G., Hawkesworth, C. J., Haggerty, S. E., Allsopp, H. L., Rickard, R. S. & Menzies, M. (1987). Evidence for mantle metasomatism in peridotite nodules from the Kimberley pipes, South Africa. In: Menzies, M. & Hawkesworth, C. J. (eds) Mantle Metasomatism. London: Academic Press, pp. 229–311.

Fraser, K. J., Hawkesworth, C. J., Erlank, A. J., Mitchell, R. H. & Scott-Smith, B. H. (1985–1986). Sr, Nd and Pb isotope and minor element geochemistry of lamproites and kimberlites. Earth and Planetary Science Letters 76, 57–70.[CrossRef]

Fraser, K. J. & Hawkesworth, C. J. (1992). The petrogenesis of group 2 ultrapotassic kimberlite from Finsch mine, South Africa. Lithos 28, 327–345.[CrossRef][Web of Science]

Girnis, A. V., Brey, G. P. & Ryabchikov, I. D. (1995). Origin of Group 1A kimberlites: fluid-saturated melting experiments at 45–55 kbar. Earth and Planetary Science Letters 134, 283–296.[CrossRef][Web of Science]

Greenough, J. D. (1988). Minor phases in the Earth's mantle: evidence from trace- and minor-element patterns in primitive alkaline magmas. Chemical Geology 69, 177–192.[CrossRef][Web of Science]

Gregoire, M., Bell, D. R. & le Roex, A. P. (2002). Trace element geochemistry of glimmerite and MARID mantle xenoliths: their relationship to kimberlite and to phlogopite-bearing peridotite revisited. Contributions to Mineralogy and Petrology 142, 603–625.[Web of Science]

Gregoire, M., Bell, D. R. & le Roex, A. P. (2003). Garnet lherzolites from the Kaapvaal Craton (South Africa): trace element evidence for a metasomatic history. Journal of Petrology 44, 629–657.[Abstract/Free Full Text]

Haggerty, S. E. (1983). The mineral chemistry of new titanates from the Jagersfontein kimberlite, South Africa: implications for metasomatism in the upper mantle. Geochimica et Cosmochimica Acta 47, 1833–1854.[CrossRef][Web of Science]

Hanson, G. N. & Langmuir, C. H. (1978). Modelling of major elements in mantle–melt systems using trace element approaches. Geochimica et Cosmochimica Acta 42, 725–741.[CrossRef][Web of Science]

Hart, S. R. & Davis, K. E. (1978). Nickel partitioning between olivine and silicate melt. Earth and Planetary Science Letters 40, 203–219.[CrossRef][Web of Science]

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

Herzberg, C. & O'Hara, M. J. (2002). Plume-associated ultramafic magmas of Phanerozoic age. Journal of Petrology 43, 1857–1883.[Abstract/Free Full Text]

Hill, E., Wood, B. J. & Blundy, J. D. (2000). The effect of Ca-Tschermaks component on trace element partitioning between clinopyroxene and silicate melt. Lithos 53, 203–215.[CrossRef][Web of Science]

Hirschmann, M. M. & Ghiorso, M. S. (1994). Activities of nickel, cobalt, and manganese silicates in magmatic liquids and applications to olivine/liquid and to silicate/metal partitioning. Geochimica et Cosmochimica Acta 58, 4109–4126.[CrossRef][Web of Science]

Hofmann, A. W. & Feigenson, M. D. (1983). Case studies on the origin of basalt, I. Theory and reassessment of Grenada basalts. Contributions to Mineralogy and Petrology 84, 382–389.[CrossRef][Web of Science]

Hofmann, A. W., Jochum, K. P., Seufert, M. & White, W. M. (1986). Nb and Pb in oceanic basalts: new constraints on mantle evolution. Earth and Planetary Science Letters 79, 33–45.[CrossRef][Web of Science]

James, D. E., Fouch, M. J., VanDecar, J. C., van der Lee, S. & Group, K. S. (2001). Tectospheric structure beneath southern Africa. Geophysical Research Letters 28, 2485–2488.[CrossRef][Web of Science]

Janney, P. E., le Roex, A. P., Carlson, R. W. & Viljoen, K. S. (2002). A chemical and multi-isotopic study of olivine melilitites and associated rocks from the Western Cape, South Africa: implications for the sources of group 1 kimberlites and constraints on the origin of the HIMU component in Africa. Journal of Petrology 43, 2339–2370.[Abstract/Free Full Text]

Kesson, S. E., Ringwood, A. E. & Hibberson, W. O. (1994). Kimberlite melting relations revisited. Earth and Planetary Science Letters 121, 261–262.[CrossRef][Web of Science]

le Roex, A. P. (1985). Geochemistry, mineralogy and magmatic evolution of the basaltic and trachytic lavas from Gough Island, South Atlantic. Journal of Petrology 26, 149–186.[Abstract/Free Full Text]

le Roex, A. P. (1986). Geochemical correlation between southern African kimberlites and South Atlantic hotspots. Nature 324, 243–245.[CrossRef]

le Roex, A. P., Erlank, A. J. & Needham, H. D. (1981). Geochemical and mineralogical evidence for the occurrence of at least three distinct magma types in the ‘Famous’ region. Contributions to Mineralogy and Petrology 77, 24–37.[CrossRef][Web of Science]

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

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

le Roex, A. P., Späth, A. & Zartman, R. E. (2001). Lithospheric thickness beneath the southern Kenya Rift: implications from basalt geochemistry. Contributions to Mineralogy and Petrology 142, 86–106.

le Roux, P. J., le Roex, A. P., Schilling, J.-G., Shimizu, N., Perkins, W. W. & Pearce, N. J. G. (2002). Mantle heterogeneity beneath the southern Mid-Atlantic Ridge: trace element evidence for contamination of ambient asthenospheric mantle. Earth and Planetary Science Letters 203, 479–498.[CrossRef][Web of Science]

Lenoir, X., Garrido, C. J., Bodinier, J.-L., Dautria, J.-M. & Gervilla, F. (2002). The recrystallization front of the Ronda peridotite: evidence for melting and thermal erosion of subcontinental lithospheric mantle beneath the Alboran Basin. Journal of Petrology 42, 141–158.[CrossRef][Web of Science]

Marsh, J. S., Hooper, P. R., Rehacek, J., Duncan, R. A. & Duncan, A. R. (1997). Stratigraphy and age of Karoo basalts of Lesotho and implications for correlations within the Karoo igneous province. In: Mahoney, J. J. & Coffin, M. F. (eds) Large Igneous Provinces: Continental, Oceanic and Flood Volcanism. Washington, DC: American Geophysical Union, pp. 247–272.

Mitchell, R. H. (1984). Mineralogy and origin of carbonate-rich segregations in a composite kimberlite sill. Neues Jahrbuch für Mineralogie, Abhandlungen 150, 185–197.[Web of Science]

Mitchell, R. H. (1986). Kimberlites: Mineralogy, Geochemistry and Petrology. New York: Plenum, 442 pp.

Mitchell, R. H. (1995). Kimberlites, Orangites and Related Rocks. New York: Plenum, 410 pp.

Moore, R. O. & Gurney, J. J. (1985). Pyroxene solid solution in garnets included in diamond. Nature 318, 553–555.[CrossRef]

Muramatsu, Y. (1983). Geochemical investigations of kimberlites from the Kimberley area, South Africa. Geochemical Journal 17, 71–86.[Web of Science]

Muramatsu, Y. & Wedepohl, K. H. (1985). REE and selected trace elements in kimberlites from the Kimberley area (South Africa). Chemical Geology 51, 289–301.[CrossRef][Web of Science]

Nelson, D. R. (1989). Isotopic characteristics and petrogenesis of the lamproites and kimberlites of central West Greenland. Lithos 22, 265–274.[CrossRef][Web of Science]

Price, S. E., Russell, J. K. & Kopylova, M. G. (2000). Primitive magma from the Jericho Pipe, N.W.T., Canada: constraints on primary kimberlite melt chemistry. Journal of Petrology 41, 789–808.[Abstract/Free Full Text]

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

Robinson, D. N. (1975). Magnetite–serpentine–calcite dykes at Premier Mine and aspects of their relationship to kimberlite and carbonatite of alkali carbonatite complexes. Physics and Chemistry of the Earth 9, 61–70.[CrossRef]

Rogers, N. W., Hawkesworth, C. J. & Palacz, Z. A. (1992). Phlogopite in the generation of olivine-melilitites from Namaqualand, South Africa, and its implications for element fractionation processes in the upper mantle. Lithos 28, 347–365.[CrossRef][Web of Science]

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

Rudnick, R. L., McDonough, W. F. & Chappell, B. W. (1993). Carbonatite metasomatism in the northern Tanzanian mantle: petrographic and geochemical characteristics. Earth and Planetary Science Letters 114, 463–475.[CrossRef][Web of Science]

Shee, S. R. (1985). The petrogenesis of the Wesselton Mine Kimberlite, Kimberley, Cape Province, R.S.A. Ph.D. thesis, University of Cape Town.

Skinner, E. M. & Clement, C. R. (1979). Mineralogical classification of South African kimberlites. In: Boyd, F. R. & Meyer, H. O. A. (eds) Proceedings of 2nd International Kimberlite Conference. Washington, DC: American Geophysical Union, pp. 129–139.

Smith, C. B. (1983). Pb, Sr and Nd isotopic evidence for sources of southern African Cretaceous kimberlites. Nature 304, 51–54.[CrossRef]

Smith, C. B., Gurney, J. J., Skinner, E. M. W., Clement, C. R. & Ebrahim, N. (1985). Geochemical character of the southern African kimberlites: a new approach based on isotopic constraints. Transactions of the Geological Society of South Africa 88, 267–280.

Späth, A., le Roex, A. P. & Duncan, R. A. (1996). The geochemistry of lavas from the Comores Archipelago, western Indian Ocean: petrogenesis and mantle source region characteristics. Journal of Petrology 37, 961–991.[Abstract/Free Full Text]

Späth, A., le Roex, A. P. & Opiyo-Akech, N. (2001). Plume–lithosphere interaction and the origin of continental rift-related alkaline volcanism—the Chyulu Hills Volcanic Province, southern Kenya. Journal of Petrology 42, 765–787.[Abstract/Free Full Text]

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

Sweeney, R. J. & Winter, F. (1999). Kimberlite as high-pressure melts; the determination of segregation depths from major element chemistry. In: Gurney, J. J., Gurney, J. L., Pascoe, M. D. & Richardson, S. H. (eds) Proceedings of the VIIth International Kimberlite Conference. Cape Town: Red Roof Design, pp. 846–851.

Tainton, K. M. & McKenzie, D. (1994). The generation of kimberlites, lamproites, and their source rocks. Journal of Petrology 35, 787–817.[Abstract/Free Full Text]

Taylor, W. R., Tompkins, L. A. & Haggerty, S. E. (1994). Comparative geochemistry of West African kimberlites: evidence for a micaceous kimberlite endmember of sublithospheric origin. Geochimica et Cosmochimica Acta 58, 4017–4037.[CrossRef][Web of Science]

Ulmer, P. & Sweeney, R. J. (2002). Generation and differentiation of group II kimberlites: constraints from a high-pressure experimental study to 10 GPa. Geochimica et Cosmochimica Acta 66, 2139–2153.[CrossRef][Web of Science]

Wagner, P. A. (1914). The Diamond Fields of Southern Africa. Cape Town: Struik, 355 pp.

Wass, S. Y. & Rogers, N. W. (1980). Mantle metasomatism—precursor to continental alkaline volcanism. Geochimica et Cosmochimica Acta 44, 1811–1823.[CrossRef][Web of Science]

Waters, F. G. & Erlank, A. J. (1988). Assessment of the vertical extent and distribution of mantle metasomatism below Kimberley, South Africa. Journal of Petrology 29, 185–204.

Wendlandt, R. F. & Eggler, D. H. (1980). The origins of potassic magmas: 2, stability of phlogopite in natural spinel lherzolite and in the system KAlSiO₄–MgO–SiO₂–H₂O–CO₂ at high pressures and high temperatures. American Journal of Science 280, 421–458.[Abstract/Free Full Text]

Wilson, M., Rosenbaum, J. M. & Dunworth, E. A. (1995). Melilitites: partial melts of the thermal boundary layer? Contributions to Mineralogy and Petrology 119, 181–196.[Web of Science]

Wyllie, P. J. (1979). Magmas and volatile components. American Mineralogist 64, 469–500.[Abstract]

Wyllie, P. J. (1980). The origin of kimberlite. Journal of Geophysical Research 85, 6902–6910.

Wyllie, P. J. (1987). Discussion of recent papers on carbonated peridotite, bearing on mantle metasomatism and magmatism. Earth and Planetary Science Letters 82, 391–397.[CrossRef][Web of Science]

Wyllie, P. J. & Lee, W.-J. (1999). Kimberlites, carbonatites, peridotites and silicate–carbonate liquid immiscibility explained in parts of the system CaO–(Na₂O + K₂O)–(MgO + FeO)– (SiO₂ + Al₂O₃)–CO₂. In: Gurney, J. J., Gurney, J. L., Pascoe, M. D. & Richardson, S. H. (eds) Proceedings of the VIIth International Kimberlite Conference. Cape Town: Red Roof Design, pp. 923–932.

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

This article has been cited by other articles:

				T. C. Birkett FIRST-ROW TRANSITION ELEMENTS, Y AND Ga IN KIMBERLITE AND LAMPROITE: APPLICATIONS TO DIAMOND PROSPECTIVITY AND PETROGENESIS Can Mineral, October 1, 2008; 46(5): 1269 - 1282. [Abstract] [Full Text] [PDF]

				T. F.D. Nielsen and K. K. Sand THE MAJUAGAA KIMBERLITE DIKE, MANIITSOQ REGION, WEST GREENLAND: CONSTRAINTS ON AN Mg-RICH SILICOCARBONATITIC MELT COMPOSITION FROM GROUNDMASS MINERALOGY AND BULK COMPOSITIONS Can Mineral, August 1, 2008; 46(4): 1043 - 1061. [Abstract] [Full Text] [PDF]

				V. S. Kamenetsky, M. B. Kamenetsky, A. V. Sobolev, A. V. Golovin, S. Demouchy, K. Faure, V. V. Sharygin, and D. V. Kuzmin Olivine in the Udachnaya-East Kimberlite (Yakutia, Russia): Types, Compositions and Origins J. Petrology, April 1, 2008; 49(4): 823 - 839. [Abstract] [Full Text] [PDF]

				C. Bonadiman, M. Coltorti, S. Duggen, L. Paludetti, F. Siena, M. F. Thirlwall, and B. G. J. Upton Palaeozoic subduction-related and kimberlite or carbonatite metasomatism in the Scottish lithospheric mantle Geological Society, London, Special Publications, January 1, 2008; 293(1): 303 - 333. [Abstract] [Full Text] [PDF]

				M. Merry and A. le Roex Megacryst suites from the Lekkerfontein and Uintjiesberg kimberlites, southern Africa: evidence for a non-cognate origin South African Journal of Geology, December 1, 2007; 110(4): 597 - 610. [Abstract] [Full Text] [PDF]

				M. Becker, A. P. le Roex, and C. Class Geochemistry and petrogenesis of South African transitional kimberlites located on and off the Kaapvaal Craton South African Journal of Geology, December 1, 2007; 110(4): 631 - 646. [Abstract] [Full Text] [PDF]

				M. de Wit The Kalahari Epeirogeny and climate change: differentiating cause and effect from core to space South African Journal of Geology, September 1, 2007; 110(2-3): 367 - 392. [Abstract] [Full Text] [PDF]

				T. Rehfeldt, D. E. Jacob, R. W. Carlson, and S. F. Foley Fe-rich Dunite Xenoliths from South African Kimberlites: Cumulates from Karoo Flood Basalts J. Petrology, July 1, 2007; 48(7): 1387 - 1409. [Abstract] [Full Text] [PDF]

				A. Bellis and D. Canil Ferric Iron in CaTiO3 Perovskite as an Oxygen Barometer for Kimberlitic Magmas I: Experimental Calibration J. Petrology, February 1, 2007; 48(2): 219 - 230. [Abstract] [Full Text] [PDF]

				D. Canil and A. J. Bellis Ferric Iron in CaTiO3 Perovskite as an Oxygen Barometer for Kimberlite Magmas II: Applications J. Petrology, February 1, 2007; 48(2): 231 - 252. [Abstract] [Full Text] [PDF]

				S. TAPPE, S. F. FOLEY, G. A. JENNER, L. M. HEAMAN, B. A. KJARSGAARD, R. L. ROMER, A. STRACKE, N. JOYCE, and J. HOEFS Genesis of Ultramafic Lamprophyres and Carbonatites at Aillik Bay, Labrador: a Consequence of Incipient Lithospheric Thinning beneath the North Atlantic Craton J. Petrology, July 1, 2006; 47(7): 1261 - 1315. [Abstract] [Full Text] [PDF]

				M. BECKER and A. P. L. ROEX Geochemistry of South African On- and Off-craton, Group I and Group II Kimberlites: Petrogenesis and Source Region Evolution J. Petrology, April 1, 2006; 47(4): 673 - 703. [Abstract] [Full Text] [PDF]

				C. BONADIMAN, L. BECCALUVA, M. COLTORTI, and F. SIENA Kimberlite-like Metasomatism and 'Garnet Signature' in Spinel-peridotite Xenoliths from Sal, Cape Verde Archipelago: Relics of a Subcontinental Mantle Domain within the Atlantic Oceanic Lithosphere? J. Petrology, December 1, 2005; 46(12): 2465 - 2493. [Abstract] [Full Text] [PDF]

				R. Maas, M. B. Kamenetsky, A. V. Sobolev, V. S. Kamenetsky, and N. V. Sobolev Sr, Nd, and Pb isotope evidence for a mantle origin of alkali chlorides and carbonates in the Udachnaya kimberlite, Siberia Geology, July 1, 2005; 33(7): 549 - 552. [Abstract] [Full Text] [PDF]

				D. R. Bell and R. O. Moore Deep chemical structure of the southern African mantle from kimberlite megacrysts South African Journal of Geology, June 1, 2004; 107(1-2): 59 - 80. [Abstract] [Full Text] [PDF]

This Article Abstract FREE Full Text (PDF) Supplementary data Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (47) Request Permissions Google Scholar Articles by LE ROEX, A. P. Articles by DAVIS, P. Search for Related Content GeoRef GeoRef Citation Social Bookmarking          What's this?

Online ISSN 1460-2415 - Print ISSN 0022-3530

Copyright © 2009 Oxford University Press

Oxford Journals Oxford University Press

• Site Map
• Privacy Policy
• Frequently Asked Questions

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