Fractionation of Platinum-group Elements and Gold in the Upper Mantle: a Detailed Study in Pyrenean Orogenic Lherzolites
JEAN-PIERRE LORAND*, LAURE PATTOU AND MICHEL GROS
LABORATOIRE DE MINÉRALOGIE, MUSÉUM NATIONAL D'HISTOIRE NATURELLE, UNITÉ ESA CNRS 7058, 61 RUE BUFFON, 75005 PARIS, FRANCE
RECEIVED MAY 18, 1998; REVISED TYPESCRIPT ACCEPTED DECEMBER 14, 1998
Platinum-group elements (PGE) and gold have been analysed in 68 mantle peridotite samples (lherzolites, cpx-poor lherzolites and harzburgites) from Pyrenean orogenic massifs. These samples preserve PGE and Au distribution patterns mostly inherited from a 2-Ga-old low-degree, partial melting event. The light rare earth element (LREE)-depleted lherzolites display a remarkably homogeneous PGE and Au concentration range (0·005 to 0·012 * CI chondrites). The light PGE (Pd, Rh, Ru) are slightly enriched relative to the heavy PGE and CI chondrites [(Pd/Ir)N = 1·8 ± 0·2; (Pd/Pt)N = 0·8 ± 0·2]. The harzburgites are Au and Pd depleted [(Pd/Ir)N < 1; (Au/Ir)N < 0·2] in addition to being Cu and S depleted relative to fertile lherzolites. These coupled variations indicate that Pd and Au were removed along with a sulphide melt. The contents of Ir, Ru, Rh and sometimes Pt are constant over the whole range of peridotite compositions (0·6 < Al2O3 < 4·2 wt %), thus reflecting the compatible behaviour of these elements in low-degree mantle melting processes. Ruthenium and rhodium show similar concentrations to Pd in lherzolite-hosted Cu-Fe-Ni sulphides but their whole-rock abundances do not decrease from lherzolites to sulphide-free harzburgites. To a lesser extent, this is also true for Pt and Ir. Such behaviour suggests that, except for Pd and Au, the PGE reside in mantle sulphides as melting-resistant atomic clusters or micro alloys. Pd, Rh and Ru enrichment relative to Ir similar to that observed in Pyrenean orogenic lherzolites has now been identified in all the orogenic lherzolites analysed so far for PGE, in abyssal peridotites and a few basalt-hosted mantle xenoliths. This is inferred to be a characteristic feature of the shallow mid-ocean ridge basalt source mantle since at least the early Proterozoic.
Platinum-group elements (PGE) are classified, along with Au and Re, as highly siderophile elements. Pioneering studies on basalt-hosted mantle xenoliths (see
Jagoutz et al., 1979; Mitchell & Keays, 1981; Morgan et al., 1981) suggested that these `iron-loving elements' are over-abundant in the Earth's mantle compared with theoretical values predicted by core-mantle equilibrium partitioning models (e.g. Borisov et al., 1994). Moreover, highly siderophile elements occur in approximately chondritic relative abundances. These two features have been used to support the `late-veneer hypothesis' according to which highly siderophile elements were reintroduced into the mantle by the intense meteoritic bombardment that hit the primitive Earth after core formation (Chou, 1978; Morgan, 1986; O'Neill, 1991). Although consistent with the chondritic Os isotopic composition inferred for the primitive upper mantle (Meisel et al., 1996), the late veneer hypothesis has been contradicted by recently published PGE data. Pattou et al., (1996) identified higher than chondritic Pd/Ir and Rh/Ir ratios in fertile Pyrenean orogenic lherzolites. Likewise, Snow & Schmidt, (1998) found Ru/Ir, Rh/Ir, Pt/Ir and Pd/Ir ratios about 40% higher than those of the CI chondrites in abyssal peridotites.
Higher than chondritic Ru/Ir, Rh/Ir and Pd/Ir ratios have now been identified in all the orogenic peridotite massifs analysed for PGE, i.e. Lanzo, Italy (Lorand et al., 1993), Pyrenean massifs (Pattou et al., 1996), Baldissero and Balmuccia in the Ivrea Zone, Italy (Garuti et al., 1997), Ronda (southern Spain) and Beni Bousera (northern Morocco) in the Betic-Rif cordillera (Gueddari et al., 1996; Garuti et al., 1997), and Zabargad Island, Red Sea (Schmidt et al., 1999). Compared with abyssal peridotites, these orogenic peridotites display non-trivial variations in PGE abundances that are currently ascribed to processes occurring within the sub-continental lithospheric mantle (SCLM). Specifically, orogenic lherzolites are considered to have been part of the SCLM for a considerable time span during which they have generally experienced a long history of melt infiltration, melt-rock reaction and metamorphic recrystallization (see
Menzies & Dupuy, 1991) which may have perturbed primitive-mantle-like PGE abundances. Garuti et al., (1997) ascribed the high Pd, Rh and Pt contents of the Balmuccia lherzolites to assimilation of PGE-rich dyke materials. They documented PGE enrichments in peridotite slabs collected from the vicinity of pyroxenite dykes, both in the Ivrea-Verbano massifs and at Ronda (Spain). In a preliminary study of Pyrenean peridotites, Lorand, (1989a) recognized significant variations of PGE patterns in some peridotite wallrocks of pyroxenites.
The petrogenesis of the Pyrenean orogenic massifs is now well constrained (Bodinier et al., 1987a, 1987b, , 1988, , 1990; Fabriès et al., 1989, , 1998; Downes et al., 1991; Mukasa et al., 1991; Reisberg & Lorand, 1995a; Woodland et al., 1996; Zanetti et al., 1996). Several stages of melt infiltration, focused into vein conduits, have been documented, mainly in the eastern part of the Pyrenees. However, these massifs mostly escaped the pervasive melt-rock reactions over the kilometre scale recently documented in the Ronda massif (Van der Waal & Bodinier, 1996), and a suite of peridotites variably depleted by partial melting (lherzolites, cpx-poor lherzolites and harzburgites) is preserved. Thus, Pyrenean peridotites provide the opportunity to address the behaviour of PGE and Au during the main petrogenetic processes that commonly affect samples from the sub-continental lithospheric mantle. Systematic PGE and Au analyses have been performed on 68 peridotite samples collected throughout the Pyrenean chain. A preliminary discussion of the data based on 14 analyses of lherzolite was presented by
Pattou et al., (1996). Here we report the complete database for the lherzolites as well as a detailed investigation of peridotites with more refractory compositions and of samples that have clearly interacted with infiltrating basaltic melts.
Orogenic peridotites occur as seven groups of massifs distributed along the North Pyrenean Fault (NPF), a major transcurrent lithospheric fault which separates the Iberian plate from the European plate (Fig. 1). About 40 bodies, varying in size from a few metres to two or three square kilometres, were tectonically emplaced along with deep crustal rocks into pull-apart sedimentary basins opened during the Mid-Cretaceous counter-clockwise motion of Spain relative to Western Europe (see
Vielzeuf & Kornprobst, 1984). The emplacement of lower crust-upper mantle fragments was coeval with a high-T, low-P metamorphism of the host limestones. The area is now referred to as the North Pyrenean Metamorphic zone (NPMZ).
Figure1. Structural sketch map of the Pyrenees showing the location of the main group of orogenic peridotite massifs investigated in the present paper [redrawn after
Fabriès et al., (1991)]. Stippled area indicates Hercynian and older materials (North Pyrenean Massifs and the Palaeozoic Axial Zone). Oblique ruling indicates Mesozoic sediments of the North Pyrenean Zone and equivalents. The area shown in black is the North Pyrenean Metamorphic Zone. SPU, South Pyrenean Units; PAZ, Palaeozoic Axial Zone; NPF, North Pyrenean Fault; NPFT, North Pyrenean Frontal Thrust. 1, Turon de Técouère; 2, Moncaup-Tuc d'Esse; 3, Lherz, Fontête-Rouge, Freychinède, Sem, Porteteny, Pic Couder; 4, Bestiac, Caussou, Pic de Géral, Prades.
The peridotite massifs may be brecciated and serpentinized along their tectonic contacts. Away from these contacts, the peridotites are generally <30% serpentinized except in some massifs from the western part of the Pyrenees where the peridotites may be 50-70% serpentinized (Fabriès et al., 1998). Spinel lherzolites (10-16 wt % cpx) account for >90% of the total volume of the 40 massifs. Clinopyroxene-poor lherzolites (5-10 wt % cpx) are much rarer and mainly occur in the Prades-Bestiac and Vicdessos-Lherz groups (usually referred to as the Eastern Pyrenean Massifs; see
Bodinier et al., 1988). True harzburgites are known only in the Fontête Rouge and Lherz massifs (Vicdessos-Lherz group). In the well-known Lherz massif, harzburgites form decametre-thick boudinaged bands intercalated within a lherzolite host rock. The peridotites display a metamorphic foliation which is ascribed to the uplift of the massifs to a depth of 30-40 km during the Mid-Cretaceous counter-clockwise motion of Spain relative to Western Europe (Vétil et al., 1988). This late-stage plastic deformation event [D2-R2 stage of
Conquéré & Fabriès, (1984)] variously affected the microstructures of the peridotites. Some peridotites preserve protogranular or coarse-granular microstructures inherited from re-equilibration into the spinel lherzolite stability domain [D1-R1 stage of
Conquéré & Fabriès, (1984) and
Fabriès et al., (1991)]. Others display porphyroclastic to mylonitic microstructures. Mylonitic lherzolites are common in some western Pyrenean massifs (e.g. Turon de Técouère, Table 1). Nevertheless, all the microstructure sub-types may be observed in a single massif (Conquéré & Fabriès, 1984; Fabriès et al., 1998).
Table1. Key petrographic features of the Pyrenean orogenic peridotites analysed for PGE and Au
The massifs from the central and western parts of the Pyrenees contain highly fertile lherzolites with bulk-rock compositions approaching those inferred for the convecting mid-ocean ridge basalt (MORB) source mantle [Al2O3 = 4·2 wt %; MgO/CaO = 10; mg-number = 0·89; middle REE (MREE) and heavy REE (HREE) 2 * CI chondrites; Table 1]. The lherzolites display light REE (LREE)-depleted REE patterns (0·15 < [La/Sm]N < 0·5; N indicates normalized to CI chondrites) as well as strongly positive [epsilon]Nd (+10 to +15) and negative [epsilon]Sr (-13·5 to -35) isotopic compositions characteristic ofMORB-source mantle (Downes et al., 1991; Mukasa et al., 1991). In the Eastern Pyrenean massifs, the lherzolites have mildly depleted bulk-rock compositions (2·8-3·6 wt % Al2O3). They have been interpreted as residues from a low-degree (2-7%) partial melting event (Bodinier et al., 1988), which has been dated at 2·1 Ga by the Re-Os chronometer (Reisberg & Lorand, 1995a). The degree of melting increased locally in the Eastern Pyrenean massifs to form harzburgitic bands without, however, exceeding 25% because the most refractory harzburgites still contain ~3 wt % cpx (Bodinier et al., 1988). It is currently believed that these melt extraction events recorded by the peridotites occurred during the incorporation of MORB-source mantle protoliths into the sub-continental lithospheric mantle (Reisberg & Lorand, 1995a). However, some Fontête Rouge harzburgites show distinctive U-shaped chondrite-normalized REE patterns coupled with negative [epsilon]Nd and positive [epsilon]Sr, which testify to subsequent melt infiltrations unrelated to the host peridotites (see
Bodinier et al., 1988; Downes et al., 1991).
The other rock types occurring in the Pyrenean orogenic massifs are pyroxenites. Two generations have been defined on the basis of modal compositions and structuralrelationships between the pyroxenites and the metamorphic foliation (e.g. Fabriès et al., 1991). Amphibole-free spinel and garnet pyroxenites define a layering concordant to this foliation. Their modal compositions range from spinel websterite in the thinnest layers to garnet ariegite (Cpx + Sp + Gt) and garnet clinopyroxenites in the core of the thickest layers (>30 cm thick). Layered pyroxenites have been interpreted as crystal cumulates from tholeiitic magmas intruded as dykes when the massifs were at the boundary between lithosphere and asthenosphere and later rotated toward the foliation plane (Bodinier et al., 1987b). Their pressure of crystallization has been estimated at 16 ± 2 kbar (i.e. 45-50 km; Sautter & Fabriès, 1990). A later generation of amphibole-rich veins (kaersutite-rich, spinel- and/or garnet-bearing clinopyroxenites sometimes accompanied by hornblendites) intersect the metamorphic foliation at high angles between 20 and 60° in the Lherz and Freychinède massifs. These veins were emplaced at a depth of 30-35 km during the D2 plastic deformation stage that produced the metamorphic foliation in the host peridotite (Vétil et al., 1988). The amphibole-rich veins display the mineralogical and geochemical characteristics of high-pressure cumulates from alkali lavas(Bodinier et al., 1987a). On the basis of 39Ar-40Ar radiometric age determinations on separated kaersutite (around 100 Ma), Bodinier et al., (1987a) suggested a genetic link with the alkali lavas that erupted throughout the NPMZ during the Mid-Cretaceous counter-clockwise motion of Iberia relative to the European plate. This genetic link is supported by the Sr and Nd isotopic compositions (Downes et al., 1991) and Sm-Nd model ages of separated clinopyroxenes from the veins (B. Azambre, personal communication, 1998). On the other hand, the layered pyroxenites do not relate to any known magmatic event in the Pyrenees (Downes et al., 1991). Re-Os analyses of some spinel and garnet pyroxenite layers suggest model ages of up to 2 Ga with, however, a considerable scatter as a result of subsequent perturbations of the Re-Os systematics (Reisberg & Lorand, 1995b).
Metasomatic reactions between the tholeiitic melts that generated the layered pyroxenites and the host peridotites occurred only adjacent to the thinnest (<3 cm) peridotite sheets that were isolated between two pyroxenite layers. Re-equilibrations of [Mg] down to 0·8, precipitation of clinopyroxene (up to 30%) and spinel (4-8%), and high contents of Al2O3, Sc, Ti, Hf and HREE are the three metasomatic features documented in these interbedded peridotites (Bodinier et al., 1988). In the peridotitic wallrocks of amphibole-rich veins, exchanges with alkali melts resulted in Fe enrichment of the anhydrous minerals ([Mg] down to 0·8), precipitation of ilmenite and Ti-rich amphibole (kaersutite), and increasing LREE, Hf, Sr and Ti contents relative to the peridotites collected away from the veins (Bodinier et al., 1988). Whereas it is restricted to zones a few centimetres wide in wallrocks of lherzolitic modal composition, metasomatism related to infiltration of alkali melts can extend over several decimetres in harzburgitic wallrocks (Bodinier et al., 1990; Woodland et al., 1996). In a harzburgitic wallrock adjacent to a kaersutite-rich clinopyroxenite vein, Bodinier et al., (1990) documented the existence of modal metasomatism (kaersutite, ilmenite) in the immediate vicinity of the veins and selective enrichments of the LREE vs MREE and HREE (cryptic metasomatism) extending over distances of ~70 cm. They ascribed this cryptic metasomatism to long-range (>1 m) porous flow percolation of alkali mafic melts within the harzburgites. In a small outcrop of peridotites at Caussou (Eastern Pyrenees) spinel lherzolites have been extensively replaced by kaersutite-rich peridotites (up to 18 wt % kaersutite) although no amphibole-rich veins are present (Fabriès et al., 1989). Over a few hundred square metres, the kaersutite-rich peridotite pockets coexist with spinel lherzolites with high amounts of Ti- and LREE-rich metasomatic clinopyroxene. The metasomatism at Caussou has been dated at 103 Ma and was ascribed to the entrapment of Cretaceous alkali melts within the peridotite host [see
Fabriès et al., (1989) for more details].
Outside the restricted area modified by the Cretaceous alkali magmatism in the Lherz and Caussou massifs, Pyrenean orogenic peridotites are essentially anhydrous or contain a few crystals of Ti-pargasite per polished thin section. Amphibole-free lherzolite samples are not uncommon (e.g. Fontête Rouge), and 1-2 wt % disseminated Ti-pargasite has been observed in a few massifs devoid of amphibole-rich veins (e.g. Sem; Freychinède; Tuc d'Esse). Because it is spatially unrelated to amphibole-rich veins and displays different major element compositions and incompatible trace element patterns compared with the amphibole from the veins, the disseminated Ti-pargasite has been ascribed to volatile-rich, small volume melt fractions (1%) that have percolated through the Pyrenean orogenic massifs long before the intrusion of the Cretaceous alkali basalts (Fabriès et al., 1991; Zanetti et al., 1996).
The 68 peridotite samples selected for the present study were collected in 13 of the most important orogenic massifs of the Pyrenean chain (Fig. 1). Information on their microstructure, modal composition and some key major elements is given in Table 1. Further information on mineral chemistry and major and trace element geochemistry has been given by
Conquéré & Fabriès, (1984), Bodinier et al., (1988) and
Fabriès et al., (1989, , 1991, , 1998). Sulphide petrology and S and Cu geochemistry were investigated by Lorand, (1989a, 1989b, , 1991), and
Lorand et al., (1989) reported Au data for 25 Eastern Pyrenean peridotites.
All the studied peridotites are <25% serpentinized. Serpentinized peridotites will be considered in a subsequent paper (Lorand, in preparation). Fifty-two samples are peridotites collected at distances >1 m from pyroxenite layers or amphibole-rich veins. These peridotites have sometimes been referred to as `massive peridotites' in previous works (e.g. Bodinier et al., 1988; Lorand, 1989a). However, to avoid confusion, they will be termed `unmetasomatized peridotites' in the present paper to describe the fact that they largely escaped contamination from the melts that circulated in hydrofracturing dykes. The unmetasomatized peridotites were selected to encompass the largest range of microstructures and fertility indices reported for Pyrenean lherzolites (0·3 < CaO < 3·7 wt %). Four well-exposed massifs (Turon, Tuc d'Esse, Fontête Rouge, Lherz) have been studied in more detail based on a larger number of samples. The 17 Turon de Técouère samples were collected over the total thickness of the massif (500 m) to test the homogeneity of PGE abundances at a scale of a single massif.
To investigate the effects of melt-rock interactions on PGE compositions, both peridotite sheets interbedded within pyroxenite layers and wallrock peridotites adjacent to amphibole-rich veins have been analysed. The interbedded peridotites come from three massifs from the Eastern Pyrenees (Lherz, Freychinède and Prades) and show either lherzolitic or harzburgitic modal compositions (Table 1). The peridotite wallrocks are 5 cm thick lherzolite slabs collected in the Lherz and Freychinède massifs and a 20 cm thick harzburgite similar to sample LH investigated by
Bodinier et al., (1990). In this harzburgitic wallrock, three slabs (LH A, B and C), collected at 5, 10 and 15 cm from the vein, have been analysed. In the Caussou massif, nine samples representing the various metasomatic stages defined by
Fabriès et al., (1989), i.e. spinel peridotites, cpx-rich lherzolites, cpx-rich amphibole peridotites and amphibole-rich peridotites, have been analysed. This massif allows us to document the behaviour of the PGE in a mantle peridotite pervasively percolated by alkali basalts.
Iridium, Ru, Rh, Pt, Pd and Au were determined by inductively coupled plasma mass spectrometry (ICP-MS) after NiS fire-assay and tellurium co-precipitation according to the methods described by
Jackson et al., (1990) and
Sun et al., (1993). The complete data sets are given in Tables 3-
5 along with the S and Cu contents of the analysed samples. Information about accuracy, precision and detection limits has already been reported by
Pattou et al., (1996). Further procedural blank data and analyses of the WGB1 rock standard are given in Table 2. As shown by replicate analyses of WGB1, our data are 90-95% accurate. Precision, expressed as relative standard deviation (RSD = standard deviation/mean), varies in the range 17% for Ir, 20% for Ru, 7% for Rh, 10% for Pt, 7·5% for Pd and 11% for Au. Pattou et al., (1996) reported precision ranging between 3 and 8% from replicate analyses of FON B, 71-324, 71-335 and 71-336, four lherzolites from Eastern Pyrenean massifs. It is well known that the reproducibility of PGE analyses at the ppb level is governed mainly by the degree to which PGE carriers are homogeneously distributed in the rock to be analysed (Hall & Pelchat, 1994). Thus, the lower reproducibility of the PGE analyses of WGB1 may be explained by a less homogeneous distribution of the PGE phases. For Au, Pattou et al., (1996) reported larger RSD values (30%) than for PGE and high blank values (0·7 ppb). The difficulty of obtaining good Au analyses using ICP-MS and NiS collection has often been addressed (see
Sun et al., 1993). This stems from the fact that the NiS collection for Au is not always as quantitative as it is for PGE. Gold may also adhere to the nickel sampling cones, generating strong memory effects during ICP-MS analysis. For this reason, gold was reanalysed in 24 samples using an epithermal neutron activation method (see
Lorand et al., 1989).
Table2. Blank values, accuracy and precision of PGE and Au analyses
Table3. Sulphur, copper and platinum-group element contents of unmetasomatized peridotites
Table4. Sulphur, copper and platinum-group element contents of metasomatized peridotites
Table5. PGE and Au contents of Caussou peridotites
Further tests of accuracy and precision have been performed for cpx-poor lherzolites and harzburgites. Precision obtained from four replicate analyses of samples 71-322 and 71-325 is 12-17% for Ir, 16% for Ru, 9-11% for Rh, 16-21% for Pt and 18-22% for Pd. The analyses of samples 72-432 and 82-4 were checked by ICP-MS in a commercial laboratory (Analab Limited, Australia) using the lead collection procedure for recovering PGE and Au. Samples 73-104 and 71-325 were reanalysed by a radiochemical neutron activation procedure (University of Melbourne, Australia) after separation of the PGE and Au in a small (2 g) NiS bead. The results generally fall within the precision range of our analyses (see sample 71-325). However, the agreement varies largely between samples and between elements. It is generally better for Ir (90-97%) than for Ru (80-90%). Platinum and palladium show the lowest reproducibility, with discrepancies reaching 25% for Pt and 33% for Pd between two replicate analyses of the same sample.
The new PGE and Au analyses of lherzolites presented here corroborate the main results of the
Pattou et al., (1996) paper. Considered as a whole, the 37 lherzolites analysed show a narrow PGE and Au concentration range (2·0-4·3 ppb Ir, 4·4-7·3 ppb Ru, 0·75-1·4 ppb Rh, 4·3-7·5 ppb Pt and 4·0-6·0 ppb Pd). This range is even narrower in the 17 lherzolites analysed in the Turon de Técouère massif; for instance, Ir contents range between 2·6 and 3·3 ppb. No systematic variation of PGE abundances has been observed between the different massifs of the Pyrenean chain, nor between plastically deformed (porphyroclastic, mylonitic) and undeformed samples (Table 3). Likewise, the Sem, Tuc d'Esse or Moncaup samples, which may contain up to 1-2% Ti-pargasite, show the same PGE and Au contents as amphibole-free lherzolites from Fontête Rouge.
PGE characterized by widely different crystal chemical properties and geochemical behaviour (e.g. Pd and Ir; Rh and Ir; Pt and Ir) positively correlate (Fig. 2). Such correlations suggest a strong partitioning of all the PGE into the same discrete phases, probably Cu-Fe-Ni sulphides. Pentlandite, pyrrhotite, chalcopyrite, pyrite are common accessory phases in Pyrenean orogenic lherzolites (Lorand, 1989b), and
Pattou et al., (1996) found 1207 ppb Ir, 3114 ppb Ru, 625 ppb Rh, 2335 ppb Pt, 3239 ppb Pd and 605 ppb Au in a magnetically separated sulphidefraction from a lherzolite. The average Au content calculated from the 37 ICP-MS analyses of lherzolites is 1·0 ± 0·3 ppb, whereas that calculated from instrumental neutron activation analysis (INAA) is 1·16 ± 0·2 ppb. Both values are consistent with the previous averages reported for Pyrenean lherzolites by
Lorand et al., (1989, , 1993). However, discrepancies of up to 50% may be observed in a single sample between analyses by ICP-MS and epithermal INAA, mainly because of the lower reproducibility of the ICP-MS Au analyses.
Figure2. Variations of Pd, Rh and Pt vs Ir and Pd in LREE-depleted unmetasomatized lherzolites. CI indicates CI chondritic ratios (after
Palme & Beer, 1993). The chondritic Pt/Ir ratios and systematic deviation of Pd/Ir ratios from the CI chondrite range should be noted. Error bars represent analytical imprecision as determined from
Pattou et al., (1996) replicate analyses.
CI-normalized PGE and Au abundances have been plotted in order of decreasing melting point of the pure metals as recommended by
Naldrett & Duke, (1980; Fig. 3). All the lherzolites show the PGE signature characteristic of upper-mantle peridotites, i.e. absolute concentrations in the range 0·005 to 0·011 * CI chondrites and almost flat chondrite-normalized patterns [see
O'Neill, (1991) and references therein]. However, the slight fractionation between light PGE (Ru, Rh, Pd) and heavy PGE (Ir and Pt) discussed by
Pattou et al., (1996) is observed in the 37 lherzolites analysed. Ru/Ir and Rh/Ir ratios are ~10-15% higher than CI chondritic (1·8 ± 0·2 vs 1·56 and 0·32 ± 0·05 vs 0·29; Palme & Beer, 1993). The largest discrepancy relative to CI chondrites is shown by Pd/Ir and Pd/Pt, which both are 35-50% higher than chondritic values. Palladium is also slightly enriched compared with both Ru and Rh (Pd/Ru = 0·96 and Pd/Rh = 5·14). In contrast, the mean Ru/Rh (0·18) and Pt/Ir (2·0) ratios fit the CI chondritic values fairly well (Fig. 2). These slightly fractionated PGE patterns are observed whether the lherzolites are undeformed, deformed or contain disseminated Ti-pargasite (Fig. 3). Deviations around the CI chondritic value are also observed for Pd/Au, especially in the peridotites showing coarse-granular textures. However, these deviations do not exceed 20-30%.
Figure3. CI chondrite normalized PGE (+Au) abundances in LREE-depleted unmetasomatized lherzolites; normalizing values after
Palme & Beer, (1993).
Cpx-poor lherzolites and harzburgites show similar Ir, Ru and Rh concentration ranges to the lherzolites (2·5-4·6 ppb Ir; 4·5-7 ppb Ru and 0·7-1·4 ppb Rh). Cpx-poor lherzolites can be distinguished from lherzolites by a larger Pt concentration range (4·2-8·2 ppb) and slightly lower Pd and Au contents (3·7-5·6 and 0·7-0·9 ppb, respectively). The Lherz harzburgites are Pt undepleted relative to the lherzolites (6-6·6 ppb vs 5-7·5 ppb) whereas the two Fontête Rouge harzburgites analysed are Pt depleted (Pt < 3·7 ppb). In both massifs, harzburgites are Pd depleted (down to 0·8 ppb) and their Au content is below the detection limit of our ICP-MS analytical procedure (0·7 ppb). Radiochemical neutron activation analysis (RNAA) of samples 73-104 and 71-325 suggests that Au in these rocks is similarly depleted to Pd (Table 3).
Cpx-poor lherzolites exhibit CI-normalized PGE patterns akin to those of the lherzolites, except for a larger variation of Pd/Pt ratios (Fig. 4a and b). The CI-normalized PGE patterns of the harzburgites are typical of refractory mantle peridotites (Fig. 4c). The Fontête Rouge samples display negatively slopping patterns [(Pd/Pt)N = 0·4-0·6; (Pt/Ir)N = 0·5]. By contrast, Pd is distinctly depleted relative to the other PGE in the Lherz harzburgites [(Pd/Pt)N = 0·2-0·4].
Figure4. CI-chondrite normalized PGE abundances in cpx-poor lherzolites (a and b) and harzburgites (c). Normalizing values after
Palme & Beer, (1993).
Platinum-group element abundances have been plotted against modal mineralogy to see whether PGE reside in a specific silicate or in spinel (Fig. 5). There is no correlation between Rh and cpx or spinel, the two minerals that may potentially host this platinum-group element according to experiments by
Capobianco & Drake, (1990). Rhodium concentrations even tend to increase between the lherzolites and the harzburgites in spite of a continuous decrease of cpx and spinel modal proportions. It has sometimes been suggested that olivine may accommodate Ir (e.g. Brügmann et al., 1987). There is a tendency for the olivine-rich rocks to be slightly richer in Ir. However, the scatter in Fig. 5c is very large. Lherzolites and harzburgites with widely different olivine modal proportions may show the same iridium contents. Such is the case in the Lherz massif (compare 71-321 and 71-325; Table 3) as well as at Fontête Rouge (compare 73-151 and FON A). Conversely, some fertile lherzolites with similar olivine modal proportions (e.g. 71-321 and 84-1 from Lherz; Table 1) show strongly different Ir contents (2-3·6 ppb; Table 3). Finally, the Ir content decreases from 4·2 to 2·3 ppb as olivine increases in the Tuc d'Esse peridotites analysed. All these features preclude any definitive conclusion about incorporation of Ir within olivine in Pyrenean peridotites.
Figure5. Relationships between PGE and calculated modal mineralogy of unmetasomatized peridotites (see Table 1 for more details). Open stars, lherzolites; *, cpx-poor lherzolites; [closed triangles], harzburgites.
Iridium, Ru and Rh do not correlate with fertility (represented by Al2O3 in Fig. 6). The same is also true for Pt except for the two Fontête Rouge samples that are distinctly lower in Pt compared with the other Pyrenean peridotites. The Pd vs Al2O3 trend is not linear because Pd concentrations remain almost constant for a large range of Al2O3 contents and then drop rapidly for Al2O3 < 1·5 wt %. This evolution can be observed in the massifs where lherzolites, cpx-poor lherzolites and harzburgites coexist (e.g. Lherz, Fontête Rouge; Table 3). For Au, only neutron activation data give a positive covariation trend with Al2O3 (Fig. 6). This trend is determined mainly by the decreasing Au contents (0·9-0·7 ppb) of the cpx-poor lherzolites and the low Au contents of the two harzburgites analysed by RNAA.
Figure6. Variations of PGE and Au as a function of wt % Al2O3 in the bulk rock [data from
Bodinier et al., (1988), Reisberg & Lorand (1995) and
Fabriès et al., (1998)]. Error bars for the harzburgites represent analytical imprecision as determined from replicate analyses. For Au, open and closed symbols correspond to ICP-MS and neutron activation analyses, respectively; it should be noted that both methods yield roughly similar absolute abundances although deviations of up to 50% may be observed in some samples. Symbols as in Fig. 5.
When plotted against Cu or S, PGE abundances follow trends akin to those documented with Al2O3 (Figs 7 and 8). This stems from the fact that S, Cu and Al2O3 are all moderately incompatible elements in the mantle and thus behave coherently in mantle melting processes (see
Garuti et al., 1984; Lorand, 1989a, , 1991). The low Pd contents of the harzburgites are correlated with the lowest S and Cu contents measured in Pyrenean peridotites (<50 and <10 ppm, respectively). This correlation suggests that the removal of sulphides by the melting event that created the harzburgites affected Pd but not the other PGE, except perhaps Pt in the Fontête Rouge massif. On the other hand, no such relationship between Pd and S is found in the lherzolites between pargasite-poor and pargasite-bearing samples. Although having S contents in excess of 250 ppm, samples SEB, 71-335 and PCOU1 do not show systematic change in their Pd content compared with Ti-pargasite-free lherzolites (Fig. 7).
Figure7. Plot of PGE contents vs S [S data from Lorand, (1989a, , 1991)]. Stippled vertical area delineates the S concentration range inferred for the mantle source of Pyrenean orogenic lherzolites (Lorand, 1991); it should be noted that the PGE vs S diagrams reproduce those involving Al2O3 fairly well. The few lherzolites that display S contents in excess of 250 ppm generally contain more than ~1 wt % disseminated pargasite (see Table 1). Symbols as in Fig. 5.
Figure8. Variations of PGE vs Cu [Cu data from Lorand, (1989a, , 1991)]. Symbols as in Fig. 5.
No systematic variations in PGE and Au abundances have been detected between lherzolites collected in the immediate vicinity of pyroxenites and those collected away from pyroxenites (Fig. 9a). Both the thin peridotite sheets interbedded within pyroxenite layered series and the wallrocks of amphibole-rich veins display the PGE and Au concentration range and the CI-normalized PGE patterns of unmetasomatized lherzolites. Likewise, their Au contents (0·8-1·5 ppb) span the same range. Lorand, (1989a) reported strong Rh, Pt and Pd enrichments (7 ppb Rh; 260-430 ppb Pt and 37 ppb Pd) in some peridotite sheets interbedded within websteritic layers and a systematic negative Pt anomaly (Pt < 1 ppb) in wallrock peridotites adjacent to amphibole-rich veins. Our analyses do not reproduce these anomalous concentrations, although our study includes some of the Lorand samples (e.g. 86 V2) or samples with very similar petrographic characteristics and provenance (e.g. 71-310). For Pt, the disagreement is most probably due to a lack of sensitivity in the Lorand, (1989a) analytical method that generated high blank values. Contaminations could explain the Rh, Pt and Pd anomalies, although some nugget effects inherent to PGE analyses operating from restricted powder splits (<1 g) are not unlikely.
Figure9. CI-chondrite normalized PGE patterns for peridotites collected at the vicinity of pyroxenite layers and amphibole-rich veins (a and b) and modally metasomatized peridotites from Caussou (c). Dotted areas indicate the PGE composition of unmetasomatized lherzolites (a and c) and harzburgites (c). Normalizing values after
Palme & Beer, (1993).
The metasomatized harzburgites display PGE concentrations akin to those of harzburgites collected away from pyroxenites (Fig. 9b). They are Pd depleted (1-3·7 ppb) and show negatively sloping CI-normalized PGE patterns [(Pd/Ir)N = 0·3-0·85]. Gold was always below the detection limits of our analytical method. However, sample 71-310, a thin harzburgite sheet interbedded within thick pyroxenite layers, displays CI-normalized Pd/Pt = 1. In the harzburgitic wallrock LH, sample LHA collected at ~5 cm from the amphibole-rich vein is slightly enriched in Rh, Pt and Pd compared with harzburgites collected away from pyroxenites (Fig. 9b). As shown by sample LHC, all these anomalies disappear at ~15 cm from the amphibole-rich vein.
No systematic correlation between PGE abundances and the different silicate assemblages produced by the infiltration of Cretaceous alkali basalts through the peridotites has been found in the Caussou massif (Table 5). Whether clinopyroxene rich or amphibole rich, the metasomatized lherzolites preserve similar PGE abundances and CI-normalized PGE patterns to the unmetasomatized lherzolites (Fig. 9c). The Caussou peridotites are distinguished only by more scattered Pt and Pd contents, especially in the amphibole-rich lherzolites (Fig. 10). However, although more scattered, Pd and Pt do not vary sympathetically with the strong fractionation of the MREE that characterize the cpx- and amphibole-rich peridotites [(Sm/Yb)N = 0·4-4·4]. Concerning Au, our ICP-MS analyses reproduce the strong enrichment (up to 21·4 ppb Au) that was detected in almost all the Caussou peridotites by
Lorand et al., (1989). The Au enrichment is not proportional to the amount of metasomatic clinopyroxene or amphibole. Moreover, gold does not correlate with Pd although it is usually considered to be moderately incompatible in magmatic processes (e.g. Barnes et al., 1985; Morgan, 1986; Brügmann et al., 1987).
Figure10. Variations of Pd vs Sm/Yb; noteworthy features are the decoupling between Pd and LREE and the scatter of Pd in Caussou metasomatized lherzolites compared with LREE-depleted lherzolites [REE data come from
Bodinier et al., (1988), Fabriès et al., (1989) and
Downes et al., (1991)].
Lorand, (1989a, 1991) documented systematic variations of Cu and S contents in the peridotites collected at the vicinity of pyroxenites compared with samples of similar fertility collected away from pyroxenites. The thin peridotites interbedded within layered pyroxenites are often enriched in S and Cu and display Cu/S ratios >0·1 whereas lherzolitic wallrocks of amphibole-rich veins show the opposite trend, i.e. lower Cu/S ratios ( <= 0·05). The first trend reflects precipitation of exotic sulphides from the tholeiitic melts that percolated through the interbedded peridotites whereas, in the case of wallrocks, peridotitic sulphides partially re-equilibrated with Cu-poor alkali basaltic melt (Lorand, 1991). Figure 11 indicates that even Pd does not behave sympathetically with respect to S and Cu in the metasomatized peridotites.Most of the lherzolitic wallrocks and Caussou lherzolites plot on the Pd vs S and Pd vs Cu trends defined by the unmetasomatized peridotites. Where sulphide enrichments are observed (e.g. 72-364, 71-310, harzburgitic wall-rock LH), they operate at constant Pd contents (Fig. 11 a). For example, samples LH B and LH C preserve similar Pd contents to the unmetasomatized harzburgites in spite of a five-fold increase in their S content (up to 275 ppm; Table 4). A similar decoupling between Pd and Cu is suggested by Fig. 11b. Metasomatized lherzolites preserve a narrow Pd concentration range in spite of large (a factor three) variations in their Cu contents. Such is also the case for the metasomatized harzburgites, which show a considerable spread in their Cu contents compared with harzburgites collected away from pyroxenites (0·9-26 ppm vs 6-12 ppm, respectively).
Figure11. Pd vs S and Pd vs Cu for peridotites adjacent to pyroxenite layers or amphibole-rich veins and Caussou metasomatized peridotites. [squ], interbedded lherzolites; open star, interbedded harzburgite; , harzburgitic wallrocks to amphibole-rich vein; other symbols as in Fig. 10. Stippled area delineate the Pd vs S and Pd vs Cu trends of unmetasomatized peridotites (see Figs 7 and 8). The S enrichment trend and the large scatter of Cu in some interbedded lherzolites and the harzburgitic wallrock LH should be noted. (See text for further details.)
From the foregoing discussion, it is clear that the PGE were much less sensitive to the different melt-rock reaction stages documented in the Pyrenean orogenic massifs than previously suggested (e.g. Lorand, 1989a). In general, the PGE patterns of the peridotites are not altered beyond 10 cm from the vein walls and these alterations involve only the most incompatible PGE. There are several reasons that may explain this lack of sensitivity. At first, both pyroxenite generations recognized in the Eastern Pyrenean orogenic peridotites are inferred to have crystallized from PGE-poor melts (Lorand, in preparation). Compared with the host lherzolites, anhydrous spinel pyroxenite layers are depleted in PGE by a factor of 4-10 whereas the depletion factor is 10-100 for amphibole-rich pyroxenites and hornblendites. The mean PGE content calculated from 10 analyses of layered pyroxenites is Ir 0·31 ± 0·33 ppb, Ru 1·27 ± 0·8 ppb, Rh 0·23 ± 0·13 ppb, Pt 2·03 ± 1·05 ppb and Pd 2·16 ± 1·46 ppb, whereas seven amphibole-rich veins yield the mean composition Ir 0·14 ± 0·14 ppb, Ru 0·375 ± 0·26 ppb, Rh 0·06 ± 0·05 ppb, Pt 0·575 ± 0·43 ppb and Pd 0·35 ± 0·19 ppb. Such PGE compositions are characteristic of melts having segregated Cu-Fe-Ni sulphides in their source region (e.g. Hamlyn et al., 1985); the high sulphide modal proportions of some veins (up to 0·6 wt % and 2000 ppm S; Lorand, 1991) indicate that parental magmas were S saturated when they started to crystallize in the SCLM.
Exchanges between melt vein-conduits and host peridotites are usually interpreted in term of sulphide addition from the vein to the wallrock (e.g. Garuti et al., 1997). It is clear from the lack of PGE enrichment in the interbedded peridotites that the tholeiitic melt introduced PGE-poor exotic sulphides. Concerning the peridotite wallrocks of amphibole-rich veins, they should display a general decrease of PGE contents if they had re-equilibrated with the PGE-poor alkali basaltic melts. The absence of such re-equilibration in the wallrocks of lherzolitic modal compositions is consistent with the low permeability of pyroxene-rich peridotites (see
Toramaru & Fujii, 1986), which severely restricted the amount of infiltrated melts (Bodinier et al., 1988). If we accept the hypothesis that PGE in the peridotites are hosted mainly in Cu-Fe-Ni sulphides, then this lack of re-equilibration is likely to be due to the low silicate melt/sulphide mass ratio. To fully equilibrate its PGE contents, a sulphide droplet must see large amounts of silicate melt [R factor of
Campbell & Naldrett, (1979)] to overcome the very high sulphide-silicate Nernst partition coefficients of the PGE (103-105; Bezmen et al., 1994; Peach et al., 1994; Fleet et al., 1996). In the Pyrenees, the Caussou peridotites clearly show that it is only where large amounts [several percent according to
Fabriès et al., (1989)] of alkali basaltic melt pervasively reacted with the lherzolite that Pt and Pd abundances were disturbed. However, these elements were simply remobilized on a local scale, without systematic increase or decrease in their contents. The fact that PGE patterns are more disturbed in the wallrocks of harzburgitic modal compositions is consistent with the higher permeability of olivine-rich peridotites; however, even in that case, Pt and Pd abundances are disturbed over a distance of <10 cm whereas the percolation process of the alkali basaltic melt strongly fractionated the LREE from the MREE and HREE by chromatographic effects over a metric scale (Bodinier et al., 1990). This decoupling reflects large differences of compatibility between PGE and REE. Barnes & Picard, (1993) estimated bulk solid-liquid partition coefficients (D) of 0·2 for Pd, 0·6 for Pt and 2 for Rh, whereas the D values for the REE range from 0·1 to 0·001 (Sun & McDonough, 1989). Chromatographic fractionation processes do not operate on elements with bulk solid-liquid partition coefficients exceeding 0·2 (e.g. Van der Waal & Bodinier, 1996).
The peridotites collected in the immediate vicinity of pyroxenites also demonstrate that PGE were much less sensitive to melt-rock reactions than the moderately chalcophile-siderophile element Cu. The fact that Cu re-equilibrated more easily than PGE reflects its lower sulphide melt-silicate melt partition coefficient; in other words, the R factor (silicate melt/sulphide mass ratio) necessary for Cu to fully equilibrate between sulphide melt droplets and silicate melt need not to be as high as for the PGE (e.g. Hamlyn et al., 1985). The case of Au also deserves specific comments as it may be uncorrelated with PGE or Cu and S. The Au enrichment of the Caussou peridotites was previously ascribed to percolation of H2O-CO2-rich fluids associated with the alkali basalts that reacted with this massif (Lorand et al., 1989). However, radiogenic isotope data (Sr, Nd) suggest a contamination by country rocks, mainly metamorphic Mesozoic limestones (Downes et al., 1991). Such a crustal origin explains both the decoupling between Au and the different metasomatic stages recognized at Caussou and the lack of similar Au enrichments in the peridotite wallrocks of amphibole-rich veins. A slight remobilization of Au in the unmetasomatized lherzolites is suggested by the variability of CI-normalized Au/Pd ratios. However, there is no strong positive anomaly in the CI-normalized patterns, which would be observed if Au had been added by the volatile-rich small melt fractions that precipitated the disseminated Ti-pargasite. Clearly, Cl-rich supercritical fluids were not involved at that stage; otherwise, strong disturbance of Pt, Pd and Au would be observed in the pargasite-bearing lherzolites (see
Fleet & Wu, 1996).
Unmetasomatized peridotites display PGE distribution patterns typical of mantle peridotites having experienced low to moderate degree of partial melting. The homogeneous concentration ranges of Ir, Ru and Rh reflect the compatible behaviour of these elements in mantle melting processes (e.g. Barnes et al., 1985). Osmium, which is usually considered to be as compatible as Ir (e.g. Brügmann et al., 1987; Barnes & Picard, 1993), behaves similarly (Reisberg & Lorand, 1995a). The drop of Pd in the harzburgites is consistent with the moderately incompatible behaviour of this PGE. Regarding Pt, the fact that it is either affected (e.g. Fontête Rouge) or unaffected by the partial melting event (e.g. Lherz) reflects a complex geochemical behaviour in the mantle which will be addressed below.
The compatibility of Os, Ir and Ru for mantle melting residues is usually ascribed to their ability to form refractory alloys whereas Pt, Pd and Au would be present in low-melting sulphides (see
Mitchell & Keays, 1981; Barnes et al., 1985). However, whether compatible (Ir-Ru-Rh) or incompatible (e.g. Pd), PGE occur at ppm levels in the sulphides analysed by
Pattou et al., (1996). Jagoutz et al., (1979) found 5 ppm Ir and
Hart & Ravizza, (1996) up to 10 ppm Os in sulphides separated from basalt-hosted mantle xenoliths. This strong preference of the PGE for Cu-Fe-Ni sulphides corroborates experimental results, which reveal the similar affinities of Ir and Pd for sulphide melts and sulphide liquid-silicate liquid Nernst partition coefficients in excess of 104 for both elements (Bezmen et al., 1994; Peach et al., 1994; Fleet et al., 1996). It is difficult to establish accurate mass-balance estimates of the PGE residing in the sulphides without detailed analyses of each silicate and spinel. However, the fraction of Pd residing within the sulphides can be roughly estimated by comparing the most fertile lherzolites and harzburgite 71-325 from Lherz (see Table 3). Palladium contents decrease from 5·0-6·0 ppb down to 1·0 ± 0·1 ppb. As sample 71-325 is totally devoid of sulphides, we may assume that at least 80% of the Pd was originally present as sulphides in the mantle protolith of Pyrenean lherzolites and was subsequently removed along with the sulphides by the 2 Ga melting event. Our estimate agrees well with the
Mitchell & Keays, (1981) mass-balance calculations performed for basalt-hosted spinel lherzolite xenoliths.
Lorand, (1989a, 1989b) provided arguments for a residual origin of the sulphides occurring in unmetasomatized peridotites, i.e. crystallization from a cognate (not exotic) sulphide melt, which persisted trapped after extraction of a S-saturated partial melt. These are: (1) a positive correlation between S (or sulphide modal abundances) and fertility indices (CaO, Al2O3); (2) sulphide grains preferentially occurring with opx-cpx-sp clusters; (3) the predominance of Ni-rich sulphides (pentlandite) over Fe sulphides. Residual sulphides are generally thought to exert a strong buffering effect on PGE in mantle melting residues (Peach et al., 1990; Lorand et al., 1993; Keays, 1995; Snow & Schmidt, 1998). To investigate this buffering effect, PGE concentrations in the unmetasomatized peridotites have been modelled as a function of Al2O3, assuming a priori that PGE are 100% hosted into Cu-Fe-Ni sulphides and that equilibrium melting models are applicable. The PGE concentrations in the residue are then determined by the amount of sulphide retained in the residue and their PGE content, as determined by the appropriate sulphide melt-silicate melt partition coefficients.
The amount of S trapped in the mantle residue (Cr) is determined by the following mass-balance equation:
FCl + (1 - F)Cr = C0
where C0 is the S content in the source, Cl is the S dissolved in the melt and F is the amount of melt removed. F is deduced from Al2O3 contents using the inverse method of
Bodinier, (1988) and C0 = 250 ± 20 ppm (Lorand, 1991). Cl, the solubility of S in the partial melt, is a complex function of T, P, oxygen and sulphur fugacities and iron activity, all factors that can vary during melting processes [see
Naldrett, (1989) and references therein]. For T = 1400°C, P = 25 kbar, appropriate to the physical conditions that governed partial melting of Pyrenean peridotites (Fabriès et al., 1991), a tholeiitic partial melt may dissolve about 1000-1500 ppm S at moderate fO2 [WM (wüstite-magnetite)-QFM (quartz-fayalite-magnetite); see
Poulson & Ohmoto, 1990]. None the less, no sulphide would have survived in the harzburgites if the partial melt had dissolved 1500 ppm S; the fact that sulphides are observed in almost all the harzburgites but 71-325 makes 1000 ppm S more likely. The different Cr values are then converted into sulphide weight fractions (Msulph) by assuming 38 wt % S in 100% sulphides (Lorand, 1989b).
The next step is to calculate the PGE content in Msulph. The mass-balance equation for any PGE (i) between melt (melt) and residue (res.)
MmeltCimelt + MresCires = C0
can be rearranged into
(MmeltCisulph/D) + (MresMsulphCisulph) = C0
and then
Cisulph = C0/[(Mmelt/D) + (MresMsulph)]
where Cisulph is the concentration of PGE in sulphides, Mmelt the weight fraction of melt and D the sulphide liquid-silicate liquid Nernst partition coefficient for element i. The concentration of any element i in the melting residue (Cires) is easily deduced from Cisulph and Msulph.
In spite of numerous assumptions, the above melting model reproduces the variation patterns of Pd and Au satisfactorily, using the PGE compositions of the most fertile lherzolites TUR 7 and DES 7 to define C0 and 104 and 103 for DPd and DAu, respectively (Fig. 12). These D values are in the range of experimentally determined values (Bezmen et al., 1994; Peach et al., 1994; Fleet et al., 1996). The good agreement between modelled and observed Pd contents corroborates our conclusion that Pd resides mostly within sulphides. However, to reproduce the contrasting behaviour of Ir and Pd in the harzburgites, we must assume DIr and DPd differing by one order of magnitude. Small differences have effectively been reported in some experiments (Stone et al., 1990; Fleet et al., 1991; Bezmen et al., 1994) but they were not so drastic. In addition, we also have to assume that DRu, DRh and DPt are higher than DPd to account for the compatible behaviour of these elements relative to Pd in the Lherz massive peridotites. Although reliable data are still scarce, the partition coefficients reported for Ru, Rh and Pt are generally lower than for Ir and Pd [(8·5-9·3) * 103 for Pt
(Stone et al., 1990; Fleet et al., 1991); 1·4 * 103 and 2·7 * 104 for Ru and Rh, respectively
(Bezmen et al., 1994)]. This obviously questions the exact amount of PGE held in the sulphides as well as their speciation.
Figure12. Modelling of Ir, Pd and Au abundances (continuous curves) in unmetasomatized peridotites by an equilibrium partial melting model using experimentally determined sulphide liquid-silicate liquid partition coefficients. Observed and predicted PGE contents are in good agreement for DIr = 4 * 104, DPd = 104 and DAu = 103. This agreement suggests that the PGE concentration range of the unmetasomatized peridotites was not significantly perturbed since the 2·1 Ga magmatic extraction event, which was recorded by the Re-Os systematics (Reisberg & Lorand, 1995a) (for further detail on modelling see text). Symbols as in Fig. 5.
For Ru and Rh, it is worth noting that the
Pattou et al., (1996) separated sulphide fraction displays a nearly flat CI-normalized PGE pattern that reproduces the pattern of the whole rocks fairly well [see fig. 1 of
Pattou et al., (1996)]. This suggests that Ru and Rh reside in similar proportions to Pd in the sulphides. Nevertheless, their whole-rock contents do not decrease in sample 71-325, in which all of the sulphide was consumed by partial melting (Table 3). The only possibility to reconcile these two observations is that Ru and Rh occur in the sulphide as mechanically collected refractory discrete minerals that are not liberated by partial melting. Ballhaus, (1995) suggested that PGE are already present as atomic clusters or metal-sulphide complexes (MS) in monosulphide liquid solutions (MLS), the high-temperature precursor of Cu-Fe-Ni sulphides. The PGE would be liberated by incongruent melting of MLS according to the following reaction:
MSMLS + 2e- = M0alloy + S2-melt
where M0 refers to zero valence metal, and S2- togaseous sulphur dissolved into the silicate melt, respectively. The behaviour of the metallic phase M0 is governed by its stability with respect to T and redox conditions, and the preference of the corresponding element for silicate melts. Those PGE like Pd, which has very weak affinity for PGE alloys(e.g. Fleet & Stone, 1991) and dissolves in silicate melts as oxides (e.g. Borissov et al., 1994) or S-Se-Te-bearing complexes, would be extracted along with the partial melt. Those like Os-Ir-Ru, which are poorly soluble in silicate melts and form refractory alloys stable over a large temperature and fO2 range, would not be extracted.
The same explanation as for Ru and Rh must pertain to Pt and Ir to reconcile their concentrations in the
Pattou et al., (1996) separated sulphide and the lack of correlation with S in Fig. 7. However, this sulphide shows more fractionated Pd/Ir and Pd/Pt ratios compared with whole-rock patterns [(Pd/Ir)]N = 2·68 vs 1·8; (Pd/Pt)N = 1·36 vs 0·8] so that Ir and Pt seem to be less concentrated than Ru, Rh and Pd in the sulphides. In the lack of conclusive evidence for significant incorporation of Ir into olivine, this Ir and Pt coupled deficit must be ascribed to Pt-Ir-rich alloys. Such alloys are acid resistant and may not have been quantitatively extracted in the
Pattou et al., (1996) procedure used to separate and to dissolve sulphides. Nevertheless, they may also occur as discrete minerals unrelated to the sulphides. Whether collected into sulphides or not, Pt-Ir-rich alloys may account for the good Pt vs Ir positive correlation in Fig. 2. For specific compositions, they may accommodate Rh (see
Fleet et al., 1991) and changes in alloy compositions depending on fO2 may explain the contrasting behaviour of Pt in the Lherz and Fontête Rouge harzburgites. Attempts to locate Pt-Ir alloys or Ru- and Rh-bearing minerals by optical microscopy and scanning electron microscopy (SEM) have as yet been unsuccessful. Only a single, Pt-rich discrete grain too small to provide quantitative microprobe data has been found enclosed in a pentlandite in more than 68 peridotites analysed for PGE. This deficiency may be explained by the very small size of PGE phases in mantle peridotites. Recent Laserprobe-ICP-MS data suggest that PGE atomic clusters or metal-sulphide complexes occur at the atomic scale in the sulphides (Ballhaus, 1997).
On the basis of the PGE pattern of Pyrenean lherzolites, Pattou et al., (1996) concluded that the Pd and Rh contents vary in the terrestrial mantle on spatial scales of ~100 km. In fact, the PGE database recently published for mantle peridotites reveals heterogeneities operating on a much larger scale in the upper mantle. Approximately two groups of mantle samples emerge from Fig. 13. All the orogenic lherzolites display positive Ru, Rh and Pd anomalies relative to chondritic patterns, even though intrinsic variations may be important in a single massif. The mean composition calculated for Pyrenean lherzolites is, within the range of uncertainties, akin to those for Ronda (Gueddari et al., 1996) and the LREE-depleted Zabargad lherzolites analysed by
Hamlyn et al., (1985). This group of orogenic lherzolites exhibits similar Ru/Ir (2·0 ± 0·5), Rh/Ir (0·4 ± 0·05) and Pd/Ir ratios (2·0 ± 0·3) to the abyssal peridotites analysed by
Snow & Schmidt, (1998). Closely similar PGE abundances have also been reported for fertile spinel lherzolite xenoliths from the Kilbourne Hole maar, USA, by
Morgan, (1986), and from Montferrier, a volcanic vent from southern Massif Central (Lorand & Pattou, 1996). Other orogenic lherzolites display more fractionated PGE ratios. Such is the case for the Lanzo lherzolites analysed by
Lorand et al., (1993), the Beni Bousera (Northern Morocco) and Balmuccia-Baldissero lherzolites (Ivrea Zone massifs) analysed by
Garuti et al., (1997) and some Zabargad lherzolites (Schmidt et al., 1999). Lorand et al., (1993) ascribed the strong positive Pd and Pt anomalies characteristic of Lanzo to incomplete extraction of sulphides during the adiabatic uplift and partial melting of the massif. Garuti et al., (1997) postulated a recycling of sulphide-bearing dyke material in the protolith of Balmuccia lherzolites and limited contamination of Baldissero lherzolites by recycled oceanic crust.
Figure13. Comparison of PGE abundances in Pyrenean lherzolites and rocks of similar origin from the oceanic and sub-continental lithospheric mantle; error bars correspond to one standard deviation from the mean. Only LREE-depleted, unmetasomatized lherzolites were selected for this comparison; analytical data less than 90% accurate and with precision exceeding 20% were excluded. A, abyssal peridotites (Snow & Schmidt, 1998); [closed circles], orogenic lherzolites; [open squares], basalt-hosted mantle xenoliths except Montferrier, southern Massif Central, France (m; Lorand & Pattou, 1996) and
Morgan, (1986) Kilbourne Hole data ([closed squares]). P, Pyrenees (this study); Z, Zabargad (Hamlyn et al., 1985; Schmidt et al., 1999); B, Baldissero (Ivrea Zone; Garuti et al., 1997); b, Balmuccia (Ivrea Zone; Garuti et al., 1997); R, Ronda (R1, Gueddari et al., 1996; R2, Garuti et al., 1997); BB, Beni Bousera (BB1, Gueddari et al., 1996; BB2, Garuti et al., 1997); L, Lanzo (Lorand et al., 1993): D, Dish Hill (Wilson et al., 1996); KH, Kilbourne Hole (Mitchell & Keays, 1981); C, Cameroon Line (Rehkämper et al., 1997); M, Massif Central, France (Lorand & Pattou, 1996); V, Victoria, Australia (Mitchell & Keays, 1981); CI-chondrite PGE compositions after
Palme & Beer, (1993), Pd/Ir ratios for other chondrite compositions [carbonaceous (Co, Cv), ordinary (H-CI) and enstatite chondrites (Eh) are from
Wasson & Kallemeyn, (1988)].
In the first two parts of this discussion, it was demonstrated that the PGE patterns of Pyrenean lherzolites were not significantly modified by the multiple melt infiltration stages that have been documented in these massifs. In the Pyrenean massifs, evidence of melt infiltration events pre-dating the 2 Ga partial melting event is lacking. Fertile lherzolites would not have preserved chondritic Os isotopic compositions if they had interacted with basaltic melts before separation from the convecting mantle (Reisberg & Lorand, 1995a). More likely, Fig. 13 suggests that an Ru-, Rh- and Pd-enriched reservoir has existed in the mantle since at least the early Proterozoic, the age of lithospheric accretion of the Pyrenean lherzolites, and persisted to the present in the MORB-source mantle, as demonstrated by abyssal peridotites. The case of Pt is more debatable, as Pt/Ir varies considerably around the chondritic ratio irrespective of the other PGE. Where several sets of analytical data are available (e.g. Ronda; Fig. 13c), analytical problems resulting from the occurrence of Pt as Pt-rich discrete alloys cannot be ruled out. However, this does not exclude the hypothesis of heterogeneous distribution of Pt on a regional scale.
The fact that the Ru-, Rh- and Pd-enriched mantle reservoir has now been identified on a world-wide basis can been explained only by processes having operated on the scale of the whole mantle, such as late-accreting materials with PGE distributions different from primitive (chondritic) meteorites or episodic core-mantle exchanges (see
Pattou et al., 1996; Snow & Schmidt, 1998; Schmidt et al., 1999). Enstatite chondrites, a reduced group of chondritic meteorites, were considered to be a potential candidate for explaining the Pd excess of Pyrenean orogenic lherzolite (Pattou et al., 1996). It is, however, clear from Fig. 13d that no primitive meteorite displays as fractionated Pd/Ir ratios as orogenic lherzolites and abyssal peridotites. Pattou et al., (1996) pointed out that enstatite chondrites would have delivered too much Au to the terrestrial mantle. For instance, their Au/Ir ratio is 0·6 (Wasson & Kallemeyn, 1988) vs only 0·4 in the 37 lherzolites analysed in the present paper. Finally, this class of chondrites cannot account for the Os/Ir ratio of abyssal peridotites (Snow & Schmidt, 1998). On the other hand, Snow & Schmidt, (1998) observed strong similarities of PGE patterns between abyssal peridotites and evolved magmatic iron meteorites, which represent asteroidal cores. They deduced a model that explains the particular PGE pattern of the MORB-source mantle by mixing of differentiated outer-core material entrained back into the mantle after core separation. As the core is assumed to contain ~2 wt % S (Dreibus & Palme, 1996), the
Snow & Schmidt, (1998) model might also account for the high S contents of the MORB-source mantle (200 ± 50 ppm S), as estimated from modern MORBs and orogenic lherzolites (Garuti et al., 1984; Lorand, 1989a, , 1991, , 1993; O'Neill, 1991; Hartmann & Wedepohl, 1993). Addition of only 1 wt % outer-core material to the MORB-source mantle would have established the Pd and S contents of abyssal peridotites and Pyrenean lherzolites, if, as estimated by
Snow & Schmidt, (1998), the outer core contained 572 ppb Pd.
Orogenic lherzolites such as those disseminated in the Pyrenean chain and basalt-hosted mantle xenoliths are generally considered to provide complementary sampling of the SCLM beneath post-Archaean continental areas (e.g. Menzies & Dupuy, 1991). However, apart from the Kilbourne Hole and Montferrier samples discussed above, the few xenolith suites recently analysed for PGE (i.e. Dish Hill, California; USA; Wilson et al., 1996; Cameroon Line, Africa
; Rehkämper et al., 1997; French Massif Central; Lorand & Pattou, 1996) define a second group of PGE compositions in Fig. 13. They differ from orogenic lherzolites by chondritic Pd/Ir and Pt/Ir ratios (Fig. 13c and d). Good quality Rh analyses are still scarce; nevertheless, the two available data sets (Massif Central and Dish Hill) have Rh/Ir ratios in the range of carbonaceous or ordinary chondrites (Fig. 13b). Only Ru/Ir deviates slightly from the chondritic pattern but the error bars are large and still encompass the chondritic range (Fig. 13a). Numerous workers (e.g. Barnes et al., 1988; McDonough & Sun, 1995) ascribed this chondritic PGE signature to the primitive mantle of the Earth. Accepting that idea, basalt-hosted mantle xenoliths would provide the only straightforward evidence of a late-accreting chondritic component in the Earth. In fact, the iridium concentrations compiled in Fig. 13 spread over similar ranges (1-5 ppb) in spinel lherzolite xenoliths, orogenic lherzolites and abyssal peridotites. This observation is corroborated by older Ir analyses on basalt-hosted mantle xenoliths (see
Jagoutz et al., 1979; Spettel et al., 1991). Thus, the variation of PGE ratios between orogenic lherzolites and mantle xenoliths is mainly due to variations in the content of the less refractory PGE, especially Pd. Such variations may reflect additional fractionation processes superimposed on the mantle in within-plate areas of the sub-continental lithosphere. Indeed, except for the Montferrier and Kilbourne Hole samples, basalt-hosted mantle xenoliths are depleted in S and Se by a factor of 5-10 compared with the MORB-source mantle and they show evidence for late-stage remobilization of their Cu-Fe-Ni sulphides (Lorand, 1990, , 1993; O'Neill, 1991; Lorand et al., 1998). This latter hypothesis, however, does not offer explanation for the fact that chondritic PGE ratios are preserved in xenoliths from different localities. Once this issue is solved, PGE may provide additional constraints for modelling the sub-continental lithospheric mantle from different mantle reservoirs characterized by different PGE signatures.
Pyrenean peridotites preserve PGE and Au distribution patterns typical of mantle peridotites having experienced low to moderate degrees of partial melting. Compared with other orogenic peridotites (e.g. Balmuccia or Beni Bousera), they escaped significant contamination from the various generations of basaltic melts that cross-cut the Pyrenean massifs in the sub-continental lithospheric mantle. This lack of contamination results from both the PGE-poor nature of the infiltrated melts and the low melt/sulphide ratios in the peridotites that display metasomatic features.
The harzburgites provide further evidence that low-degree partial melting fractionates Au and Pd from the other PGE (including Ir) even if operating in S-saturated conditions. Our results support recent theories assuming that, apart from Pd, PGE reside in mantle sulphides as melting-resistant atomic clusters or refractory micro alloys. The latter concept may be a better starting point to explain (1) the systematic fractionation of Pd from Ir in mantle peridotites and mantle-derived melts resulting from moderate degrees of partial melting, which is not predicted by sulphide melt-silicate melt partition coefficients (e.g. Peach et al., 1994; Keays, 1995); (2) the compatible behaviour of some PGE (e.g. Pt) usually classified as moderately incompatible (e.g. Barnes et al., 1985).
It is likely that an Ru-, Rh- and Pd-enriched reservoir similar to that sampled by Pyrenean orogenic lherzolites has existed in the MORB-source mantle since at least the early Proterozoic. Modelling this light PGE overabundance by differentiated outer-core material entrained back into the mantle after core separation would also account for the high S content of orogenic lherzolites. Whether it is a Primitive Mantle-derived feature or not now depends on how well PGE abundances in basalt-hosted mantle xenoliths will be constrained.
Liliane Savoyant (University of Montpellier) is gratefully acknowledged for operating the inductively coupled plasma mass spectrometer in Montpellier. Thanks are extended to R. R. Keays and S. Reeves for the radiochemical neutron activation analyses. This paper benefited from helpful discussions with participants at the EAG Workshop `The Origin and Fractionation of Highly Siderophile Elements in the Earth's Mantle', Mainz, 14-16 May (1997). Editorial comments of M. Wilson and reviews by H. Palme, G. Garuti and a third anonymous reviewer greatly improved the final version. Financial support was provided by an EEC grant (SC1*CT91-0660; `Ophiolites mineralized in platinum-group elements') and a grant from the French Ministry of Education `Action spécifique métaux nobles'.
