| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology | Volume 43 | Number 12 | Pages 2219-2259 | 2002
© Oxford University Press 2002
Mechanisms and Sources of Mantle Metasomatism: Major and Trace Element Compositions of Peridotite Xenoliths from Spitsbergen in the Context of Numerical Modelling
1DEPARTMENT OF EARTH AND ENVIRONMENTAL SCIENCES, UNIVERSITÉ LIBRE DE BRUXELLES, CP160/02, B-1050 BRUSSELS, BELGIUM
2DEPARTMENT OF GEOLOGY, NIIGATA UNIVERSITY, 2-8050 IKARASHI-NINOCHYO, NIIGATA, 950-2181, JAPAN
3LABORATOIRE DE TECTONOPHYSIQUE (UMR 5568 CNRS), ISTEEM, UNIVERSITÉ MONTPELLIER 2, CASE 049, 34095 MONTPELLIER CEDEX 05, FRANCE
4DEPARTMENT OF GEOLOGICAL SCIENCES, UNIVERSITY OF MICHIGAN, ANN ARBOR, MI 48109-1063, USA
5CNR–ISTITUTO DI GEOSCIENCE & GEORISORSE (IGG), SEEIONE DI PAVIA, 27100 PAVIA, ITALY
Received October 2, 2000; Revised typescript accepted May 22, 2002
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONMineral and whole-rock chemical data for peridotite xenoliths in basaltic lavas on Spitsbergen are examined to reassess mechanisms of melt–fluid interaction with peridotites and their relative role versus melt composition in mantle metasomatism. The enrichment patterns in the xenoliths on primitive mantle-normalized diagrams range from Th–La–Ce ‘inflections’ in weakly metasomatized samples (normally without amphibole) to a continuous increase in abundances from Ho to Ce typical for amphibole-bearing xenoliths. Numerical modelling of interaction between depleted peridotites and enriched melts indicates that these patterns do not result from simple mixing of the two end-members but can be explained by chromatographic fractionation during reactive porous melt flow, which produces a variety of enrichment patterns in a single event. Many metasomatized xenoliths have negative high field strength element and Pb anomalies and Sr spikes relative to rare earth elements of similar compatibility, and highly fractionated Nb/Ta and Zr/Hf. Although amphibole precipitation can produce Nb–Ta anomalies, some of these features cannot be attributed to percolation-related fractionation alone and have to be a signature of the initial melt (possibly carbonate rich). In general, chemical and mineralogical fingerprints of a metasomatic medium are strongest near its source (e.g. a vein) whereas element patterns farther in the metasomatic ‘column’ are increasingly controlled by fractionation mechanisms.
KEY WORDS: Spitsbergen; lithospheric mantle; metasomatism; trace elements; theoretical modelling
|
INTRODUCTION |
|---|
Peridotite xenoliths brought to the surface by alkali basaltic magmas on the island of Spitsbergen represent continental upper mantle located at present in the vicinity of oceanic spreading centres in the North Atlantic and Arctic basins. Much of the earlier work on these xenoliths focused on carbonates and related interstitial phases believed to have formed as a result of interaction between mantle rocks and carbonate-rich melts during entrainment and transportation to the surface by the host magmas (Amundsen, 1987; Genshaft & Ilupin, 1987; Ionov et al., 1993b, 1996). The interstitial carbonates, which are rare in mantle xenoliths elsewhere, attracted particular attention as direct evidence for carbonate-rich melts in the mantle. In particular, trace element compositions of acid-leached interstitial material from several xenoliths (Ionov et al., 1993b) and in situ analyses of carbonates and associated interstitial clinopyroxene and silicate glass (Ionov et al., 1996; Ionov, 1998) were used to infer geochemical signatures of mantle metasomatism by carbonate-rich fluids.
Previous studies have also addressed the petrography of the xenoliths and provided a limited amount of data on the compositions of minerals apparently unrelated to the late-stage interstitial material (Furnes et al., 1986; Ionov et al., 1996). Analyses of coarse clinopyroxene grains by proton probe (Ionov et al., 1996) indicated metasomatic enrichments in Sr. Furthermore, laser-ablation microprobe (LAM) inductively coupled plasma mass spectrometry (ICP-MS) analyses in three samples showed that abundances and elemental ratios in the coarse clinopyroxene are distinct from those in clinopyroxene from interstitial carbonate- and glass-bearing pockets (Ionov, 1998). Some xenoliths were found to contain texturally equilibrated amphibole, apatite and phlogopite (Ionov et al., 1993b, 1996). Altogether, the earlier work has yielded unequivocal petrographic and chemical evidence for modal and cryptic metasomatism that took place before the formation of the silicate glass and carbonate-bearing material. However, the data are not sufficient to unravel the nature and evolution of the mantle beneath Spitsbergen.
In this work we establish major and trace element signatures of the peridotite mantle beneath Spitsbergen through analysis of whole-rock xenoliths and of minerals that largely control the incompatible trace element inventory in the mantle. These data, together with theoretical modelling, are used to understand the sequence of upper-mantle processes recorded in the xenolith suite. The major goal of this study is to further constrain mechanisms of mantle metasomatism and define the nature and sources of the metasomatic components responsible for enrichment of the lithospheric mantle in incompatible elements. We focus on characterizing metasomatic and other events that pre-date those related to the late Cenozoic volcanic activity that brought the xenoliths to the surface.
Petrographic and chemical studies of mantle rocks worldwide have found a variety of mineralogical associations and trace element enrichment patterns that appear to result from interaction of depleted peridotites with magmas and fluids (e.g. Frey & Green, 1974; Menzies & Hawkesworth, 1987; McDonough & Frey, 1989; Bodinier et al., 1990; Johnson et al., 1996; Mukasa & Wilshire, 1997; Yaxley et al., 1998; Ionov et al., 1999a). It has become apparent that the enrichment phenomena generally referred to as mantle metasomatism may be implemented through a large number of processes (e.g. batch crystallization, melt fractionation in conduits and chambers, melt or fluid percolation). These mechanisms can be identified and constrained based on trace element distributions in natural samples compared with results of theoretical modelling.
Another widely debated topic of mantle geochemistry is the provenance (sources) of melts and fluids responsible for metasomatism in the continental mantle. It has been suggested that the metasomatic media may originate within the lithospheric mantle, convecting asthenospheric mantle, deep-mantle plumes, or subducted oceanic lithosphere (Beccaluva et al., 2001; Churikova et al., 2001; Downes, 2001 and references therein). Furthermore, mantle metasomatism may involve a variety of media (silicate and carbonate melts, water- and CO2-rich fluids) that are derived from different sources. Particular attention has recently been given to metasomatism by carbonate-rich melts, which seem to have a strong enrichment potential and impose specific trace element signatures (e.g. Green & Wallace, 1988; Yaxley et al., 1991; Dautria et al., 1992; Hauri et al., 1993; Ionov et al., 1993b; Rudnick et al., 1993; Downes, 2001).
Overall, mantle metasomatism has been attributed to a variety of: (1) processes, (2) sources, and (3) melt–fluid compositions. Which factors dominate in creating specific metasomatic signatures in the continental lithospheric mantle remains widely debated. For example, trace element patterns earlier seen as evidence for metasomatism by carbonate-rich melts have been recently interpreted as source signatures or the result of fractionation processes (Bedini et al., 1997; Blundy & Dalton, 2000; Laurora et al., 2001). Addressing these complex issues necessitates comprehensive geochemical investigation of a representative sample series, which has been undertaken in this study, and in the accompanying paper that addresses Sr–Nd–Pb isotope composition of the same sample series (Ionov et al., 2002).
|
GEOLOGICAL SETTING |
|---|
Spitsbergen is the largest island of the Svalbard archipelago off Norway, which is the subaerially exposed corner of the Barents–Kara shelf between the North Atlantic and Arctic (Eurasian) oceanic basins (Fig. 1a). Tectonic reconstructions indicate that Svalbard was contiguous with Greenland until the northward progression of the North Atlantic opening produced Late Cretaceous–Neogene separation of Eurasia and North America (Blythe & Kleinspehn, 1998). At present, Spitsbergen is located at a distance <200 km from the North Atlantic mid-ocean ridge (MOR) (Fig. 1b).
|
The xenoliths for this study have been collected in hawaiitic to nepheline basanitic lavas at three Quaternary volcanic centres on the western side of Woodfjorden in NW Spitsbergen (Fig. 1b): Sverre, Halvdan and Sigurd. Tholeiitic lavas, 9–12 Myr old, from nearby areas do not contain mantle xenoliths. Relevant data on geology, tectonic setting and geophysical results for this region have been summarized by Amundsen et al. (1987) and Yevdokimov (2000). Additional information on xenoliths from these localities has been provided by Furnes et al. (1986), Genshaft et al. (1992) and Ionov et al. (1996).
|
SAMPLE PREPARATION AND ANALYTICAL PROCEDURES |
|---|
Samples for this study were selected from a larger collection (Yevdokimov, 2000) to represent major peridotite rock types and modal variations. We have also tried to select fresh xenoliths that are large enough to prepare representative whole-rock samples and provide material for mineral separation. The samples are listed in Table 1, along with a summary of the petrography, modal compositions, geothermometry and other information on the analyses performed.
|
The xenoliths were cut using a diamond saw; central parts containing no host basalt or saw marks were crushed in a bench-top jaw crusher, which was carefully cleaned after each sample to avoid cross-contamination. Aliquots of the crushed samples were ground in an agate mortar to produce whole-rock powders. Another aliquot was sieved, size-fractioned and magnetically separated to yield sub-fractions enriched in clinopyroxene and amphibole. Ultrapure mineral separates were handpicked from these for isotope analyses (Ionov et al., 2002). Several handpicked grains (
0·2–0·3 mm) from each sample were put on mounts for in situ analyses. The handpicked clinopyroxene grains are larger than clinopyroxene from interstitial carbonate- and glass-bearing material (Fig. 2a); the latter are therefore absent from the grain mounts. Samples of host basalts were sawn off from lava attached to the xenoliths. Basalt chips free of saw marks, weathering products and xenocryst material were leached in 6 M HCl and ground to powder in an agate mortar.
|
Major elements in bulk rocks were determined by X-ray fluorescence (XRF) spectrometry at Niigata University using low-dilution fused beads [see Takazawa et al. (2000) for analytical details] and at Michigan State University following standard protocols. Major-element compositions of minerals were determined in thin sections with a Cameca Camebax SX-50 electron microprobe (EMP) at Macquarie University, Sydney [for analytical details see Ionov et al. (1996)] and with Camebax SX-100 at Université Blaise Pascale, Clermont-Ferrand, France. Modal compositions were calculated from whole-rock and mineral major-element analyses using least-squares regression.
Trace elements in whole-rock peridotites were determined by solution ICP-MS at Niigata University; basalts were analysed at CODES (Hobart). At both laboratories, the instrument was an HP4500; calibration was performed against BHVO-1 using values of Eggins et al. (1997). As, Rh, In, Tm and Bi were used as internal standards. Several peridotites (including duplicates and acid-leached rocks) were analysed on a VG PlasmaQuad instrument in Montpellier following the method of Ionov et al. (1992). Reference samples BIR-1 and JP-1 (Makishima & Nakamura, 1997) were analysed as unknowns for quality control. Trace elements in minerals were determined by LAM-ICP-MS in grain mounts and polished rock sections of 200 µm thickness in Sydney and Pavia. The instrument at Macquarie University is a Perkin–Elmer Sciex ELAN 5100 coupled with a UV (266 nm) laser [see Norman et al. (1996) and Ionov (1998) for details of operating conditions]. The instrument at the CNR–IGG of Pavia is a double focusing sector field analyser (‘Element’ Finnigan Mat) coupled with a Q-switched Nd:YAG laser source (Quantel Brilliant), whose fundamental emission in the near-IR region (1064 nm) is converted to 266 nm by two harmonic generators (Bottazzi et al., 1999). Helium was used as carrier gas and mixed with Ar downstream of the ablation cell. A BCR2-g reference sample was used as an external standard, with 44Ca as an internal standard for clinopyroxene and amphibole and 29Si for phlogopite. Precision and accuracy (<10% and <5%, respectively) were assessed from repeated analyses of SRN NIST 612 reference sample. Trace elements were also determined in profiles across clinopyroxene and amphibole grains by SIMS (secondary ion mass spectrometry) on Cameca 4f instruments in Pavia and Montpellier following procedures reported by Bottazzi et al. (1994).
|
PETROGRAPHY AND MAJOR ELEMENT COMPOSITIONS |
|---|
Petrography
Most of the xenoliths from Halvdan and Sverre are medium- to coarse-grained spinel lherzolites with low to moderate modal clinopyroxene (Table 1). Samples 26a and 318 have <3% clinopyroxene and are therefore classified as harzburgites (Streckeisen, 1976). Many of the Halvdan xenoliths contain amphibole (Table 1), whose modal abundance ranges from rare individual grains to >4%. Amphibole in samples 318 and 315-6 has textural position, grain size and shape similar to those of neoblast olivine and pyroxenes. It does not appear to replace clinopyroxene or spinel because, in particular, spinel abundances in amphibole-rich peridotites are not noticeably lower than in amphibole-poor samples (Fig. 2b). A single sample from Sigurd (4-90-1) is a composite xenolith of amphibole wehrlite (spinel-free) grading into a coarse-grained olivine-dominated rock that contains clusters of amphibole and less abundant phlogopite in an olivine matrix (Fig. 2c). Two Halvdan xenoliths (too small for mineral separation) contain apatite in addition to amphibole (Fig. 2a). Fe–Ni sulphides are common accessory minerals. The petrography of these rocks has been presented in more detail by Ionov et al. (1996).
Some of the peridotites contain interstitial veins and pockets of fine-grained material made of silicate glass, carbonate and second-generation clinopyroxene, olivine and Cr-spinel, enclosing resorbed relict grains of minerals of the host peridotite (Fig. 2a). This late-stage fine-grained material is not considered in this work, and readers are referred to Ionov (1998) for more information. Sample 311-9 has pockets of silicate glass with micro-phenocrysts and vugs associated with spinel (Fig. 2d); no amphibole is present in the thin section available. The shapes and textural position of the glassy pockets resemble those of amphibole mantling spinel in some other Spitsbergen xenoliths; similar pockets in samples 21-5 and 21-6 enclose resorbed amphibole relics (Fig. 2a) (Ionov et al., 1996). It is most likely that xenolith 311-9 originally contained some amphibole, which broke down shortly before or during the eruption of the host basalts, as has also been inferred for certain types of glass-bearing pockets in mantle xenoliths worldwide (e.g. Yaxley et al., 1997). We therefore classify sample 311-9 as amphibole-bearing to properly characterize its modal composition before the late-stage melting events.
Whole-rock compositions
Whole-rock major element compositions are given in Table 2 and illustrated in Fig. 3. The xenolith suite can generally be characterized as moderately depleted in basaltic components (Al, Ca, Na, Ti). Most of the samples plot in the middle of compositional fields of basalt-borne, off-cratonic peridotite xenoliths and massif peridotites worldwide on covariation plots vs MgO and Mg# [molar Mg/(Mg + Fe) ratio] (Fig. 3; McDonough, 1990). Two samples (including an amphibole-rich peridotite) have high MgO and low CaO and Al2O3 similar to some xenoliths from kimberlites; however, they differ from typical cratonic mantle peridotites by higher FeO and lower Mg# (Fig. 3a and b). A few samples plot above the MgO–CaO trend on Fig. 3c, consistent with the presence of <1% of Ca-rich interstitial carbonate indicated by earlier studies (Ionov et al., 1996). Carbonate-poor sample 63-90-30 shows a very high molar Ca/Al ratio of 1·65 compared with a range of 1·0–1·2 typical for the spinel lherzolites (Table 2). The composition of Fe-rich sample 39-86-1 is distinct from that of a duplicate sample (39-86; Figs 3a and b, and 4a) prepared from the same xenolith (Ionov et al., 1996) indicating rock heterogeneity.
|
|
, amphibole-bearing peridotites; •, ‘anhydrous’ peridotites. Shown for comparison as grey symbols are xenoliths in basalts from Siberia and Mongolia (small open circles), xenoliths in Yakutian kimberlites (crosses) and peridotites from Horoman and Ronda massifs (small rhombs). Continuous grey line outlines typical compositions of xenoliths in basalts; dashed lines contour typical compositions of xenoliths in kimberlites and massif peridotites. Shaded fields in (c) and (d) show chemical trends calculated for residues from fractional partial melting of fertile spinel lherzolites (
|
Mineral compositions and temperature estimates
Major element compositions of minerals are given in Table 3. Values of Mg# in olivine (Mg#Ol) are tightly correlated with Mg# in whole rocks (Fig. 4a) and show a general positive trend with the Cr/(Cr + Al) ratios of spinel (Cr#Spl) (Fig. 4b). Most of the Spitsbergen samples on the latter diagram plot between the tight field of fertile (Mg# <0·90) xenoliths and highly depleted xenoliths from East Greenland and abyssal peridotites, consistent with partial melting relationships (Bernstein et al., 1998; Hellebrand et al., 2001). Sample 26a, which has the highest Cr#Spl, plots with off-cratonic xenoliths that show a wide range of Cr#Spl at moderately high Mg#Ol (0·90–0·915; Fig. 4b). In some xenoliths from Spitsbergen (and Samoa), the Cr#Spl values are higher, at a given Cr2O3 content in clinopyroxene, than in other mantle peridotites (Fig. 4c).
|
The abundances of Ti, Al and Na in clinopyroxene are shown in Fig. 4d–f vs Mg#Ol. Olivine hosts 72–90% of Mg and Fe in the xenoliths and therefore its Mg# established during partial melting is less affected by posterior temperature-dependent sub-solidus intermineral Mg–Fe partitioning (Brey & Köhler, 1990) than the Mg# of pyroxenes. The abundances of Ti and Al in clinopyroxene in the majority of the samples are negatively correlated with the Mg#Ol (Fig. 4d and e). By contrast, Na2O contents are nearly the same (1·0–1·5%) and are higher than for abyssal peridotites and xenoliths from Samoa at similar Mg#Ol (Fig. 4f). Clinopyroxene in 26a has much lower abundances of Na, Al and Ti than in the rest of the xenoliths. Amphibole wehrlite 4-90-1 plots away from spinel peridotites in Fig. 4 because of its low Mg# and high TiO2 in clinopyroxene.
The Mg# values of amphibole show a near-linear correlation with the Mg#Ol. The Cr abundances in amphibole are positively correlated with the Cr#Spl (Fig. 5a and b). These and other results indicate that amphibole in all of the samples is chemically equilibrated with the other minerals. In comparison, the abundances of K and Ti, minor elements strongly concentrated in the amphibole, are poorly correlated (Fig. 5d). Most amphiboles have low K2O (<1%), particularly 315-6 (0·02%).
|
Equilibration temperature estimates calculated from the electron microprobe data using methods of Wells (1977) and Brey & Köhler (1990) are listed in Table 1. The values for the Halvdan xenoliths seem to be generally higher than those for the Sverre xenoliths, but this difference may not be meaningful considering the small number of samples analysed. Our T estimates (840–1030°C) are much lower than the range of 940–1170°C reported for Spitsbergen spinel lherzolites by Amundsen et al. (1987). We failed to obtain T values >1080°C applying the same thermometer (Sachtleben & Seck, 1981) to samples in this study or published data (Furnes et al., 1986). Projection of our T estimates onto the geotherm for northwestern Spitsbergen after Amundsen et al. (1987) yields unrealistically low pressures of 7–11 kbar. We conclude that the above geotherm overestimates temperatures in the uppermost mantle, apparently because of the use of inappropriate thermometers. We have roughly estimated equilibration pressures for the lherzolites in this study (11–15 kbar) by projecting their Ca-opx T range (890–1025°C) onto a P–T trend calculated by applying the same thermometer (Brey & Köhler, 1990) and relevant barometers to published electron microprobe analyses of garnet-bearing pyroxenites from Spitsbergen (Amundsen et al., 1987).
Minerals in most samples show no, or only limited, chemical zoning. In several xenoliths, pyroxene rims have lower Al and Cr than the cores. Harzburgite 26a has particularly strong Al–Cr zoning and low Al in pyroxenes (coupled with Cr-rich spinel, Fig. 4b) and a very high Mg# in clinopyroxene relative to that in olivine and orthopyroxene. Because Fe–Mg partitioning between olivine and pyroxenes is temperature dependent, the high Mg# in clinopyroxene may be related to the very low equilibration temperatures estimated for that sample (740–800°C). The low T values indicate a relatively shallow depth of origin, compared with the other xenoliths, possibly near the crust–mantle boundary. The cores of the largest orthopyroxene grains from sample 26a have exsolution lamellae and very high Ca abundances, indicating incomplete equilibration after cooling from T >1000°C.
|
TRACE ELEMENT COMPOSITIONS OF MINERALS |
|---|
Average trace element abundances for the cores of clinopyroxene, amphibole and mica grains determined by LAM-ICP-MS are given in Table 4. Differences between individual analyses of the same mineral were normally within analytical precision, except for highly variable Zr (28–97 ppm), Nb (0·4–1·1 ppm), Hf (1–2·5 ppm) and Ta (<0·02–0·18 ppm) in clinopyroxene 39-86-1, and Rb (5·6–10·5 ppm) and Ba (760–1350 ppm) in amphibole 318-1. Sample 39-86-1 is obviously heterogeneous and is not considered below. We also use LAM-ICP-MS analyses of minerals in four xenoliths from Ionov (1998). Several analyses of apatite were published earlier (Ionov et al., 1996).
|
The abundances of moderately incompatible elements, such as the heavy rare earth elements (HREE) and Y, are lower in the Spitsbergen clinopyroxenes than for fertile lherzolites worldwide and decrease systematically with increasing Mg#Ol (Fig. 6a). In contrast, the abundances of highly incompatible elements [Sr, light REE (LREE), Nb, Th, U] do not define coherent trends with Mg# variations (Fig. 6b and c) and for nearly all the samples are much higher than those expected for residues after partial melting (e.g. Johnson et al., 1990). Similar observations were made earlier for many other suites of mantle peridotites and attributed to metasomatic enrichment in incompatible elements of residual rocks formed by earlier partial melting events (e.g. Frey & Green, 1974; Kempton, 1987; McDonough & Frey, 1989). Minerals from composite xenolith 4-90-1 and harzburgite 26a commonly plot off trends defined by the other xenoliths.
|
Two major types of clinopyroxene can be identified in the Spitsbergen xenoliths on the basis of their incompatible element abundances and ratios (Fig. 6b–f). Type-1 clinopyroxene is characterized by lower concentrations of the middle REE (MREE) and Sr (and therefore lower MREE/HREE ratios, e.g. Nd/Yb) relative to Type-2 clinopyroxene, and by high La/Ce ratios (Fig. 6e). We emphasize that the terms Type-1 and Type-2 are used here simply to distinguish between the two most common trace element patterns in this xenolith suite. They are not related to the classification of Frey & Green (1974), which would identify all those samples as Type I (based on their high Mg#), or with types 1a (LREE depleted) and 1b (LREE enriched) of Kempton (1987).
The differences between the two types are further highlighted on primitive mantle (PM)-normalized trace element distribution diagrams (Fig. 7). Type-1 clinopyroxenes have nearly flat HREE–MREE patterns (at 
4 times the PM level) followed by moderate depletion from Eu to Nd (Pr) and a steep La–Ce–(Pr) inflection (Fig. 7a). Type-2 clinopyroxenes have somewhat lower HREE concentrations (3 x PM) followed by a continuous increase in normalized REE concentrations from Ho to Ce (Fig. 7c). Both types of clinopyroxene have strong negative Nb anomalies and small to moderate positive Sr anomalies (Fig. 7b and d). Type-2 clinopyroxenes have similar or only slightly lower concentrations of Ti, Zr and Hf compared with Type-1 clinopyroxenes, but their high MREE levels result in pronounced negative anomalies of those elements (Fig. 7d) and low Ti/Eu ratios. Both clinopyroxene types also have a similar range of Pb concentrations, but high Nd and Ce in the Type-2 bring about conspicuous negative Pb anomalies (Fig. 7d). A Ce/Pb vs Ti/Eu diagram clearly distinguishes between the two types (Fig. 6f).
|
|
Clinopyroxenes 26a and 63-90-30 cannot be grouped with either Type-1 or Type-2: 26a has very low abundances of all lithophile trace elements and a nearly flat REE pattern; 63-90-30 has a convex-upwards REE pattern (Fig. 7e and f). Vein clinopyroxene 4-90-1 has an REE pattern resembling those of Type-2 clinopyroxenes but differs from them largely by its much higher HFSE (high field strength element) abundances and lower HREE (Fig. 7d–f).
Only one Type-1 sample contains amphibole (315-6), whereas all Type-2 xenoliths are amphibole bearing. However, it is the elemental abundances and ratios rather than presence of ‘hydrous’ phases that define and characterize the two geochemical xenolith types. The REE pattern of Type-1 amphibole 315-6 is clearly distinct from that of Type-2 amphiboles (Fig. 8a); it also has unusually low abundances of K (Fig. 5d), Nb and Ta (Fig. 8b). REE patterns in amphiboles are very similar to those of coexisting clinopyroxenes; these two minerals also have very similar abundances of Th, U, Zr and Hf (Figs 8d and 9b). However, the amphiboles have much higher concentrations of the alkalis, Ba, Nb, Ta and Ti, and moderately higher concentrations of Pb and Sr (Fig. 8c), consistent with amph–cpx relationships observed in other mantle rocks (e.g. Ionov & Hofmann, 1995; Kempton et al., 1999; Tiepolo et al., 2001). Vein amphibole has high Nb and Ta abundances, no negative anomalies for Ti and Zr, and a negative Pb anomaly (Fig. 8b). All the amphiboles are relatively low in Rb and Cs, with Rb/Ba and Rb/Sr ratios below those of the primitive mantle.
|
Ion probe analyses in profiles across mineral grains found slightly lower Ti and V in rims compared with cores of clinopyroxene (but not in the single amphibole grain analysed). Other elements show no significant differences in abundances between the cores and the rims (Table 5, Fig. 8d). The ion probe and LAM-ICP-MS data for the same samples agree well (Fig. 8d).
|
|
TRACE ELEMENT COMPOSITIONS OF BULK ROCKS |
|---|
Trace element abundances in whole-rock samples are given in Table 6. Ionov et al. (1993b) and Ionov (1998) found high incompatible trace element concentrations in interstitial material within some Spitsbergen xenoliths, in particular high Sr, Ba and U in the carbonate. Because this study is concerned with the composition of Spitsbergen peridotites before the late-stage formation of the interstitial materials, it is clearly important to establish their relative role in the whole-rock budget. Table 6 and Fig. 9a give results of a leaching experiment on sample 25, similar to those reported by Ionov et al. (1993b). Xenolith 25 has abundant interstitial material and higher modal carbonate than most other samples studied in this work (Table 1) and represents a ‘worst case’ example of their combined effects on whole-rock compositions. The leachate (in 10% HNO3) from the crushed rock has very low abundances of REE, HFSE and Th but is strongly enriched in Sr, Pb, U, Ba and Rb. It is obvious that the spikes for those elements in the whole-rock pattern are mainly related to the interstitial material (Fig. 9a). On the other hand, the residue after leaching has small but conspicuous positive anomalies for the same elements indicating qualitative similarities between trace element signatures of the late-stage interstitial material and metasomatic patterns in the rest of the rock, as discussed by Ionov et al. (1993b).
|
|
We further address the role of interstitial ‘contaminants’ by plotting abundances of incompatible trace elements in the bulk rocks versus those in clinopyroxene and amphibole (Fig. 10). The majority of the data points plot along well-defined trends consistent with enrichments in LREE, Sr, Pb, Th and U mainly hosted by clinopyroxene and amphibole. Anomalously high Sr in two whole-rock samples is clearly related to their high modal calcite (Table 1), as for sample 25 (Fig. 9a), also reflected in elevated CaO on the CaO–MgO plot (Fig. 3c). Overall, the cases of anomalous whole-rock enrichments as a result of eruption-related metasomatism or post-eruption alteration are rare and taken into account below. Whole-rock abundances calculated from mineral and modal compositions commonly are close to those directly measured by solution ICP-MS (Fig. 9c), except for elements that tend to reside in interstitial micro-phases (Eggins et al., 1998; Bedini & Bodinier, 1999; Garrido et al., 2000; Kalfoun et al., 2002).
The REE patterns of whole-rock xenoliths are similar to those of their clinopyroxenes (compare Figs 11a and 7a, and 11c and 7c). The Type-1 and Type-2 patterns can be easily identified from LREE–MREE relationships for most of the xenoliths, including those from the earlier work (Ionov et al., 1993b). For some elements, in particular HFSE, their whole-rock abundances relative to those of adjacent REE are distinct from those for clinopyroxene, whose composition is significantly affected by partitioning into coexisting amphibole (Fig. 8c) and/or orthopyroxene (Rampone et al., 1991; Ionov et al., 1995; Eggins et al., 1998; Garrido et al., 2000; Tiepolo et al., 2001). Whole-rock Type-1 xenoliths, unlike their clinopyroxenes, have no or only minor negative Ti–Zr anomalies (compare Figs 7b and 11b). The low Ti/Zr values, common in Type-2 clinopyroxenes (Fig. 7d) as a result of strong Ti partitioning into coexisting amphibole, are not seen in the bulk rocks. Similarly, whole-rock Type-2 xenoliths do not have negative Nb anomalies, and some yield high Nb/La and Nb/Th values. However, all Type-2 rocks have strong negative Ta anomalies, and therefore, high Nb/Ta ratios (Fig. 11d) consistent with data for amphiboles 318 (Fig. 9b) and 318-1, and clinopyroxene 311-9 (Table 4).
|
The abundances of moderately incompatible V, Sc, Y and HREE are lower than in fertile lherzolites worldwide and decrease with increasing MgO (Fig. 12). There are no differences in HREE and Y levels at similar MgO (or Mg#) between amphibole-bearing and amphibole-free peridotites, unlike for clinopyroxene (Fig. 6a). Similar to Ti, Zr abundances are much lower than in the primitive mantle (Fig. 13) and do not seem to be related to MgO (Figs 3d and 12). Abundances of Sr, LREE, Nb, Th, U, Ba and Rb vary widely and do not define coherent trends with MgO variations (Fig. 12). If samples with Sr-rich interstitial material are discarded, Type-2 rocks show higher Sr (as well as Pr, Nd, Sm and Eu) than Type-1 rocks (Figs 12 and 13). The abundances of highly incompatible elements (La, Ce, Pb, Th and Ba) overlap in Types 1 and 2, but the two types are clearly discriminated on the La/Ce vs Nd/Yb plot; this demonstrates the strong LREE fractionation in Type-1 xenoliths. The two types are also distinguished on a plot of La vs Th (or U) because of higher Th at similar La in highly LREE-enriched Type-1 xenoliths (Fig. 13).
|
, other amphibole-free; 
, Type-1 amphibole-bearing;
|
The distribution of Nb is distinct from that of Ti and Zr (Fig. 13). Ti and Zr are uniformly low in both rock types and do not seem to be affected by metasomatism. Nb abundances are usually much higher in Type-2 xenoliths, and show no correlation either with Zr or La and Th, elements that have relative peridotite–melt compatibility similar to that of Nb. The high Nb cannot be attributed to the presence of amphibole in the Type-2 xenoliths alone because amphibole-bearing xenolith 315-6 is as low in Nb as other Type-1 rocks. Zr/Hf and Nb/Ta range from subchondritic in some Type-1 rocks to superchondritic in the rest of the xenoliths and are particularly high in Type-2 rocks (Fig. 13). Zr/Hf values are positively correlated with some incompatible elements, e.g. Nd, but vary broadly at similar Zr. By contrast, Nb/Ta increases with Nb abundances both in Type-1 and Type-2 rocks (Fig. 13). Ce/Pb is commonly below the PM value in both Type-1 and Type-2 rocks, which is also seen as negative Pb anomalies in Fig. 11d (when interstitial Pb contamination is taken into account).
|
CHEMICAL RECORD OF DEPLETION AND METASOMATISM |
|---|
Evidence for partial melting and mineral–melt equilibria in chemical compositions
Abundances of major oxides and moderately incompatible elements in whole rocks and minerals vary systematically with variations in MgO contents and Mg# in most of the xenoliths (Figs 3, 4, 6a and 12). Such trends are commonly interpreted as indicating that the peridotites formed as residues from variable degrees of partial melting and melt extraction from a fertile lherzolite source. This assertion is supported by abundant experimental and modelling results and data on natural peridotites [e.g. a recent review by Takazawa et al. (2000)].
Most of the Spitsbergen xenoliths plot close to calculated evolution lines for residual spinel peridotite at 1–2·5 GPa (Niu, 1997) on major oxide covariation diagrams (e.g. MgO vs CaO; Fig. 3c). Relatively high CaO in several samples is due to late-stage precipitation of interstitial calcite. On Mg# vs Al2O3 and MgO vs FeO plots (Fig. 3a and b), the Spitsbergen peridotites fall within the compositional field of the Horoman peridotites inferred to have been derived by polybaric melting at 2·5–0·4 GPa. Takazawa et al. (2000) concluded from data on Mg–Fe partitioning that the Horoman peridotites with olivine Mg# and whole-rock MgO similar to those in the most depleted Spitsbergen xenoliths (26a and 318) have been produced by 19–25% of partial melting. Importantly, amphibole-bearing Spitsbergen peridotites plot within the compositional fields defined by the majority of the xenoliths and follow major oxide variation trends related to partial melting, indicating that amphibole precipitation was not accompanied by significant additions of ‘basaltic’ components (FeO, CaO, Al2O3, TiO2). An apparent exception is sample 318 (with highest modal amphibole, >4%), which has Al, Ca and Ti abundances too high relative to MgO to be an unaltered melting residue; yet the very low FeO in sample 318 (Fig. 3b) rules out amphibole formation from an evolved, Fe-rich basaltic melt.
We calculated trace element abundances in residues after incremental (1% steps) partial melting of primitive spinel lherzolite using an algorithm and partition coefficients from Takazawa et al. (2000). Compositions of clinopyroxene and whole-rock samples from this study are compared with the calculation results in Figs 7e and 11e. The whole-rock abundances of HREE and MREE and Zr (elements least affected by metasomatism) are consistent with 3–11% of partial melting for Type-1 xenoliths and 5–12% for Type-2 xenoliths. The very low HREE values in harzburgite 26a cannot be reproduced by batch melting and require 20–25% of incremental partial melting (see also Fig. 12), consistent with its high whole-rock MgO, low whole-rock Ca, Al and Ti, low modal clinopyroxene (2·9%), and low Al, Ti and Na in the clinopyroxene (Fig. 4). Sample 318 has higher MgO and Mg# than 26a (Table 2). However, the degree of incremental melting estimated for amphibole-rich xenolith 318 from HREE abundances (12%) is much lower than for 26a, possibly as a result of REE enrichment by metasomatism.
Many Spitsbergen xenoliths plot off calculated partial melting trends on covariation diagrams of moderately incompatible elements vs MgO (Fig. 12). Similar features in other mantle peridotites were previously attributed to melting in the garnet stability field, variable amounts of trapped melt, changes in melting regime, etc. (e.g. Takazawa et al., 2000). Other processes, such as solid–melt mixing (Elthon, 1992) and solid–melt reaction (Kelemen et al., 1992; Van der Wal & Bodinier, 1996) may have accompanied or followed the partial melting and melt extraction. It is hardly possible, however, to better constrain melting conditions for the Spitsbergen peridotites because of widespread overprinting of the initial melting signatures by later metasomatism.
The convex-upward trace element patterns of whole-rock xenolith 63-90-30 (Fig. 11) and its clinopyroxene (Fig. 7) cannot be produced by partial melting and indicate equilibration with a melt that has lower Nb and LREE, and higher HREE than the host basalt (based on published cpx/meltD; Fig. 14). Several other features also set that sample apart from the rest of the xenoliths: high whole-rock Ca/Al, high Ti in the bulk rock (Fig. 3d) and clinopyroxene (Fig. 4d), and high Cr# in clinopyroxene relative to spinel (Fig. 4c). However, the high Mg# in whole rock (0·909) and olivine (0·911) appear to rule out an origin from or equilibration with an evolved, Fe-rich basaltic melt.
|
Mineral compositions in composite xenolith 4-90-1 are distinct from those in the other xenoliths, in particular low Mg# and high Ti (Figs 4–7). That sample appears to be a fragment of a magmatic conduit system in the peridotite mantle. The modal gradient from dunite to wehrlite (Fig. 2c) may reflect different degrees of magma–host interaction in the cross-section of a magmatic channel (Kelemen, 1990), whereas abundant amphibole and phlogopite may be due to significant water and alkali contents of the magma and/or in situ crystallization of trapped fluid. Estimates of the trace element composition of a basaltic liquid in equilibrium with clinopyroxene 4-90-1 show strong enrichments in highly incompatible elements, with Th, U, Nb and LREE several times higher than in the basalts hosting the xenoliths (Fig. 14). The REE pattern of that hypothetical liquid is similar to calculated liquids for clinopyroxene from several Type-2 xenoliths (e.g. 311-9, Fig. 14a), except for somewhat higher LREE and MREE and lower HREE for cpx 4-90-1. The minor differences between the patterns calculated for clinopyroxene from the vein and from the Type-2 xenoliths could reflect local variations in metasomatic melt compositions. Alternatively, they can be explained assuming that the vein clinopyroxene equilibrated with an initial LREE-enriched melt (flowing in a conduit), whereas clinopyroxenes in the peridotite xenoliths record equilibration (either partial or complete) with a melt modified during percolation from a conduit into the host peridotites. We conclude that a melt with REE abundances in equilibrium with vein cpx 4-90-1 could be an appropriate metasomatic agent for the Spitsbergen xenolith suite.
Mineral zoning and the timing of metasomatism
Strong compositional gradients between the depleted cores of minerals and their rims (in direct contact with an enriched melt) may develop during initial stages of metasomatism of melting residues. The lack of significant trace element zoning—even in large, porphyroblastic Type-1 clinopyroxenes (revealed by ion probe analyses, Table 5)—indicates that the mineral grains were internally homogenized by diffusion during or after metasomatism. The time required for homogenization is primarily dependent on diffusion coefficients and temperature as well as on the distance. We use here experimental data of Sneeringer et al. (1984) on Sr diffusion in natural diopside, which are consistent with Sr diffusion coefficients obtained on clinopyroxene (and amphibole) in other studies. Possible temperatures may range from 1200°C for metasomatism by a silicate melt to 
1100°C for metasomatism by a carbonate-rich melt or fluid. The lower T values may be more appropriate in view of the experimentally determined amphibole stability field in pyrolite compositions (1050°C) (Niida & Green, 1999) if it is applicable to less fertile Spitsbergen peridotites. The solution of Fick’s equation given by Crank (1975) for Sr homogenization (from 45 to 181 ppm, to 95% of the final equilibrium level) to occur in a grain with a radius of 1–2 mm yields about 106 years for T in the range 1050–1100°C and 104 years at 1200°C. These results further indicate that the metasomatism was not a very recent event related to formation of interstitial silicate glass and carbonate.
Trace element evidence for mechanisms and sources of metasomatism and its interpretation
Nearly all the xenoliths studied in this work show evidence for modal and/or cryptic metasomatism, which must have been widespread in the source regions of the xenoliths. In the discussion below we will focus on the nature of the metasomatism, leaving aside the few samples that do not bear well-defined enrichment signatures (26a, 63-90-30).
Geochemical evidence outlines two major types of the metasomatized rocks. One of them (Type-1) shows smaller and highly variable degrees of metasomatism, and only one xenolith from that group contains amphibole (with an unusually low K concentration). In contrast, all Type-2 xenoliths contain amphibole (some also have apatite) and show strong enrichments in some incompatible elements, with very similar enrichment patterns (Figs 6–8 and 11). Although trace element patterns and covariation plots (e.g. Fig. 13) clearly distinguish between the two groups, one cannot rule out that intermediate compositions also exist in the mantle beneath Spitsbergen, but are not represented in this sample series. Because both groups have similar equilibration temperatures, they appear to come from the same depth range in the uppermost mantle, but we have no information on specific spatial relationships between the two rock types, e.g. with relation to the sources of metasomatic media.
This study has to address two principal questions: (1) whether the two rock types were formed by two different mechanisms of metasomatism or, alternatively, reflect two distinct stages of the same mechanism (event); (2) whether distinct compositions (and sources) of metasomatic media were involved in the formation of the two rock types. The second question can be reworded as how to distinguish between metasomatic features produced by fractionation processes versus those that should be attributed to distinct metasomatic melt–fluid compositions.
These questions are addressed below using theoretical modelling of metasomatism. We first establish that simple mixing of depleted peridotites with melts enriched in incompatible trace elements cannot produce strong LREE fractionations and the diversity of enrichment patterns found in the Spitsbergen xenoliths. We then use numerical modelling to demonstrate that those features can be explained by chromatographic effects of melt percolation in peridotite matrix. Two modelling techniques are employed: (1) one-dimensional (1-D) melt percolation with fixed rock/melt ratios; (2) ‘plate’ models of reactive porous flow (percolative fractional crystallization). The results are presented in order of increasing model complexity, beginning with the REE distribution in spinel peridotites during melt percolation at constant rock/melt ratios. More complex models explore the effects of trapped melt, variable rock/melt ratios, distinct modal compositions of the peridotite, and the role of amphibole in element fractionation. These models are first run for REE to constrain their abundances in the initial metasomatic melt and outline the most likely physical parameters of melt percolation. We then introduce a broader range of elements (HFSE, Th, U and Pb) assuming that their ratios to adjacent REE in the initial melt are close to those in the primitive mantle. Finally, alternative melt compositions (with anomalies on PM-normalized diagrams for certain elements) are considered to provide the best fit between the models and the Spitsbergen xenoliths. For simplicity, a single set of mineral–melt partition coefficients (mineral/meltDElement) is used (Table 7), except for Nb and Ta in amphibole.
|
The origin of strong LREE enrichment and fractionation in the Spitsbergen xenoliths
The strongly fractionated LREE patterns of several Type-1 xenoliths with extremely high La/Ce, La/Pr and La/Nd ratios provide key evidence for their mode of origin. These extreme fractionations between individual REE do not appear to have analogues among any type of terrestrial magma. Therefore, such REE patterns in xenoliths cannot be modelled by simple mixing between LREE-depleted residues of partial melting and a magmatic liquid. Figure 15 demonstrates that mixtures of a model LREE-depleted peridotite with either host basalt or a hypothetical highly LREE-enriched liquid (in equilibrium with vein 4-90-1, Fig. 14) cannot reproduce high La/Ce or Sr/Nd ratios in typical Type-1 rocks. Furthermore, the ratios of highly incompatible elements in the mixing models are nearly uniform for moderately to highly enriched compositions (resulting in subparallel LREE–MREE patterns) because they are dominated by those in the liquid. By contrast, the LREE patterns and La/Ce, La/Nd and other ratios vary broadly among individual Type-1 xenoliths and seem to require a specific liquid composition for each sample in the mixing models. Finally, Navon & Stolper (1987) have demonstrated that plots of LREE vs MREE for peridotite suites with LREE enrichment patterns similar to those in this work do not form linear arrays expected in the case of simple mixing.
|
Similarly, it is unlikely that an exotic metasomatic fluid with extreme LREE fractionations and very high LREE/HREE (implied in mixing models) could form in equilibrium with any common mantle rock type. It follows that possible origins for the trace element patterns of Type-1 xenoliths may be related to enrichment processes rather than to unusual metasomatic source compositions. The key idea is that some metasomatic mechanisms may generate a range of fractionated REE patterns in the course of an enrichment event from a single initial LREE-enriched liquid.
The REE patterns of Type-1 xenoliths are very similar to those produced by modelling chromatographic effects during melt percolation (Navon & Stolper, 1987; Bodinier et al., 1990; Vasseur et al., 1991;