SULPHIDE MINERALOGY AND CHALCOPHILE/SIDEROPHILE ELEMENT GEOCHEMISTRY OF ABYSSAL PERIDOTITES FROM THE MARK AREA (20-24°N MID-ATLANTIC RIDGE, SOUTH OF THE KANE TRANSFORM)

Abyssal peridotites equilibrated in the spinel lherzolite facies and sampled along ultra-slow, slow and intermediate spreading ridges, at transform faults and on-axis, in the Atlantic and Indian oceans, are compared for available mineralogical / geochemical data, with special emphasis on modal compositions and textures. Large mantle regions (100->1000 km) below the Atlantic and Indian ridges have been highly depleted by partial melting and melt extraction, not always in relation with known nearby hotspots. In those regions, residual harzburgites locally underwent secondary crystallization of clinopyroxene, spinel and olivine as a result of late-stage melt / rock reactions involving dissolution and/or incongruent (re)melting of residual orthopyroxene. In extreme cases, these reactions lead to the formation of secondary lherzolites and/or wehrlites closely associated with dunites. Trace element compositions of the secondary clinopyroxenes, as well of residual clinopyroxenes where some are still present, suggest that reacting melts derived from primary melts generated by partial melting of already depleted upper mantle. Residual peridotites sampled along transform faults are in the whole less depleted than on-axis peridotites. However, they are also characterized by a wide compositional range that includes highly depleted harzburgites. Thus, offaxis more fertile abyssal peridotite compositions probably not reflect a colder thermal regime of the ridge near the transforms. Instead, peridotite compositional variability along the transforms likely results from episodic magmatic activity at slow spreading mid-oceanic ridges, with temporal variations in the degree of melting. Pyroxenes dissolution textures and intergranular igneous clinopyroxenes can also be observed locally in less depleted, residual harzburgites and lherzolites although in much lower amounts than in the highly depleted harzburgites. In all peridotite samples, igneous clinopyroxenes and other minerals produced by melt precipitation and last-stage melt-rock reactions crystallized at the convective – conductive thermal regime transition, such that high temperatures are responsible for near-homogeneous compositions of all textural types of minerals. At a global scale, when regional averages of a large number of samples are considered, modal compositions and mineral chemistry (except sodium and the most incompatible trace elements) are relatively well correlated, and consistent with variable degrees of partial melting (~5-25%) of a lherzolitic source. Nevertheless, at a local scale, each subaxial mantle domain has its own compositional trend indicating variable melting parameters and/or source composition. Furthermore, orthopyroxene contents for a given olivine content allow to define a subgroup of orthopyroxene- rich harzburgites which is characteristically found in the most refractory subaxial mantle regions and whose clinopyroxenes are, in average, richer in sodium and chromium than those from orthopyroxene-poor harzburgites and lherzolites. Such differences in compositions might result from partial melting occurring at higher pressures in the more depleted regions. However, even in taking in account various late magmatic / metamorphic events which could have modified the primary mineral proportions, the present variations in the olivine / orthopyroxene ratios of all abyssal peridotites cannot be explained by any partial melting processes, at a global and regional / local scales. In the eastern region of the Southwest Indian ridge, along-axis compositional variability is as high as in transform zones. The scale of this compositional heterogeneity is correlated with the roughness and irregularity of the rift floor. In contrast with transform zone peridotites, however, mineral compositions and modal compositions do not correlate with each other and classical partial melting indicators are not internally coherent. The most refractory harzburgites clearly underwent metasomatism by fertile, alkali-rich, hydrous, asthenospheric melts, before subridge partial melting and high temperature annealing, while lherzolites appear to be well equilibrated hybrid rocks resulting from refertilization of refractory peridotites by basaltic melts mostly generated in the garnet stability field. Such subaxial mantle may represent heterogeneous upper mantle that upwelled beneath the present ridge and has been preserved because of very low degrees of melting. Alternatively, it may possible that in this very slow spreading / cold environment, thick subaxial lithosphere formed during the previous magmatic episode, was thermochemically eroded during the subsequent (present-day) magmatic event.

Thirty peridotites and eleven pyroxenites collected in the southwestern part of the Ronda peridotite massif have been analysed by iodometric titration, XRF and ICPMS to investigate the effect of the regional porous flow percolation process documented by Van der Wal and Bodinier (1996) on chalcophile/siderophile elements. The peridotites come from the different petro-structural domains now recognised within the Ronda peridotite. These are garnet/spinel peridotites (fertile lherzolites to harzburgites) from the Spinel Tectonite domain, the old (presumably middle Proterozoic; Reisberg and Lorand, 1995) sub-continental lithospheric protolith, lherzolites, harzburgites and dunites from the Granular Domain that extensively recrystallized during a young (Cainozoic?) regional melt percolation and lithospheric erosion event and plagioclase lherzolites from the Plagioclase Tectonite domain, which is a solid-state recrystallization feature resulting from the crustal emplacement of the massif. Each of the three sub-domains defined by Van der Wal and Bodinier (1996) in the Granular Domain has been investigated, i.e. the Coarse-Granular domain (CGD), where melt accumulation below the Recrystallization Front generated extensive recrystallization and silicate grain growth, the Fine-Granular domain (FGD), an area of melt entrapment and cpx-producing reactions (cpx2), and the Layered Granular domain (LGD) which display elongated bodies or layered-like olivine-rich lherzolites, harzburgites and dunites, sometimes rich in chromite, cpx2 and low Ti magmatic amphibole maybe present around spinel. Such olivine-rich lithologies (Mg#=0.88-0.9) may represent channels where percolating melts were focused at high-T, producing olivine by reaction with the pyroxenes (e.g., Bodinier et al., 1999). Our analyses reveal a much greater complexity in the Cu, S and PGE patterns than previously reported in the Ronda peridotites. This diversity clearly results from the superimposition of several magmatic processes in the mantle. All the lherzolites show S (150-250 ppm), Cu (20-30 ppm) and PGE concentration ranges (0.004-1.2 x CI-chondrites) mostly inherited from the ancient melting event that affected the protolith. Like many orogenic and oceanic lherzolites, they are light PGE-enriched (RuN/IrN = 1.6-2.6; RhN/IrN = 0.38- 0.48; PdN/IrN = 1.3-2.4; N = chondrite-normalised). Rather good positive correlations are preserved between S, Cu and fertility indices (r = 0.9) in cpx2-free samples. Palladium and fertility indices behave sympathetically; negativelyslopping CN normalised PGE patterns have survived in some CGD cpx-poor lherzolites and FGD harzburgites. Meanwhile coupled Pd, Cu and S enrichments are observed in cpx2-bearing samples, The olivine-rich LGD peridotites display the greatest diversity of PGE contents and PGE relative abundances. Chrome spinel dunites are strongly impoverished in all chalcophile/siderophile elements but Ru (Ir = 0.001 x CI chondrites; S 1; SN/PdN > 1 ;N = protolith normalised). Type D Cr-rich clinopyroxenites, the latest pyroxenite type, are richer in PGE (up to 55 ppb) but much lower in S and Cu (CuN/PdN < 0.2; SN/PdN < 0.04). A peculiar Au-, Pd- and As-enrichment trend proportional to the distance from the Recrystallization Front is observed in the harzburgites from the Spinel Tectonite domain. It is ascribed to upward-migration of S-poor but volatile-rich small-melt fractions that escaped from the Granular Domain. Local concentrations of these small-melt fractions could have probably given rise to the PGE-rich Cr- As-Ni mineralizations of magmatic origin that characterise the Ronda peridotite massif (Gervilla and Leblanc, 1990). The type D pyroxenites are likely precursor of these mineralizations. However, another Au enrichment trend (up to 6 x protolithic values) uncorrelated to the other noble metals also characterise lherzolites and harzburgites from both sides of the recrystallization front. Gold abundances in these rocks would record a widespread circulation of late-magmatic (oxidising) fluids through the south-western part of the Ronda peridotites during the latest stages of the melt percolation process. Because of numerous analogies between our Ronda data and similar rocks in ophiolitic complexes, our study could provide clues to understand the behaviour of chalcophile/ siderophile elements in fossil oceanic lithospheres submitted to melt/rock reactions above subduction zone.

Highly refractory peridotites are defined here as peridotites more depleted in melt components than abyssal peridotite, i.e. peridotites with spinel showing Cr# (Cr/(Cr+Al) atomic ratio) higher than 0.6. These correspond to the type III peridotite defined by Dick and Bullen (1984) and are reported from forearc regions, such as Izu-Ogasawara, Mariana and Tonga trenches (e.g. Bloomer and Hawkins, 1983; Bloomer and Fisher, 1987; Ishii et al., 1992) and from the Kamuikotan belt, the central axial zone of Hokkaido, northern Japan (e.g. Katoh and Nakagawa, 1986; Makita and Arai, 1997). These peridotites are, however, not common in the upper mantle and their genesis has not been explained thoroughly. We discuss the genesis of highly refractory peridotite from three complexes, the Takadomari, Iwanaidake and Nukabira, in the Kamuikotan belt, Japan, and its bearing on the process of high-Mg andesite magma genesis. The Kamuikotan belt is a tectonic melange zone composed of metamorphic rocks, ultramafic rocks, greenstones and sedimentary rocks. The metamorphic rocks are of typical high-pressure/low-temperature type and ophiolitic rocks which recorded low-pressure ocean-floor metamorphism are distributed in the Horokanai area, northern part of the belt (Ishizuka et al., 1983). Ultramafic rocks form the basal member of the ophiolite in the northern part and are exposed as peridotite complexes of various size in the southern part of the Kamuikotan belt. They are generally in fault contact with surrounding Kamuikotan metamorphic rocks, sediments and mafic volcanic rocks (e.g. Niida and Katoh, 1978). The peridotite complexes mainly consist of harzburgite and dunite which suffered serpentinization to various extent. The Takadomari complex from the northern part, and the Iwanai-dake complex from the southern part of the belt consist of harzburgite, dunite and small amounts of orthopyroxenite. The Nukabira complex (southern part of the belt) consists of lherzolite and harzburgite with dunite and pyroxenites. Podiform chromitite deposits sometimes accompany dunite in the latter two complexes. Primary hydrous minerals are sometimes included in chromian spinel of the former two complexes. The Cr# of chromian spinel and Fo (forsterite) content of olivine in peridotites from the Takadomari complex are high (Cr#=0.64-0.92, Fo=91.9- 94.0), and are higher in dunite than in harzburgite on average. The Cr# and Fo are also high (Cr#=0.43-0.87, Fo=90.8-93.5) in the Iwanai-dake complex and have wider ranges (Cr#=0.18-0.86, Fo=89.0-93.2) in the Nukabira complex. The Cr# of spinel and Fo of olivine in dunite complexes are sometimes similar to or even lower than the values in harzburgite from Iwanai-dake and Nukabira complexes. The mineral chemistry of dunites in the Iwanai-dake and Nukabira complexes depends on their thickness within harzburgite or lherzolite bodies. Thicker dunite layers tend to show higher Cr# of chromian spinel and Fo content of olivine. In addition to this, there are clear systematic variations in lithology across dunite layers and surrounding peridotites. In the case of thick dunite layers, the Cr# of chromian spinel and the Fo content of olivine gradually increase from the surrounding peridotite to the central part of the dunite layer (Cr#>0.6 and Fo>91.5 in dunite). In contrast to this fact, these values in thin dunite layers are similar to the surrounding peridotite, lherzolite or harzburgite (Cr#=0.4- 0.6 and Fo=89-91). These values are generally lower in thinner dunite layers than in the thick ones compared at the center. The harzburgites and dunites in the Takadomari complex are interpreted to be a series of refractory residue after extraction of high-Mg andesite magma; partial melting of the fertile peridotite possibly occurred under hydrous conditions. Harzburgite and lherzolite are mostly simple residue in the Iwanai-dake and Nukabira complexes but their dunites are not simple residue because the Cr# and Fo content are not systematically higher in the dunite than in the harzburgite (e.g. Arai, 1987; 1994). The dunite may be produced by a reaction between wall peridotite and a melt formed at higher pressures (e.g. Quick, 1981; Fisk, 1986; Kelemen, 1990). The differences in thickness and mineral chemistry of dunite layers are due to the difference of the melt/wall peridotite volume ratio. Supplying a large amount of melt caused formation of thick dunite layers and generated high-Mg andesitic magma by the reaction process.

Numerous small ultramafic bodies are exposed in Mesozoic cover units of the central Eastern Alps. The major occurrences are restricted to the tectonic windows of the Penninic zone and their surroundings. In the Lower Engadin window, three different mantle peridotite groups have been investigated. The Idalp ophiolite, situated at the northern rim of the Engadin window, is believed to be of south Penninic origin (Trumpy, 1972), whereas the Ramosch ophiolite at the southwestern margin is assigned to the north Penninic area (Vuichard, 1984). The tectonic position of the ultramafics in the vicinity of Nauders, at the southern margin, is unknown. Ophiolite remnants of south Penninic affinity are also present in the Glockner nappe of the Tauern Window. At the eastern end of the Alpine orogen similar peridotites appear in the small Rechnitz window. Numerous, usually small and highly serpentinized bodies of ultramafic rocks can be traced in the Matrei zone along the southern margin of the Tauern window; the Matrei Zone comprises both, Penninic and Lower Austroalpine elements. The Reckner ophiolite complex, near the NW corner of the Tauern window, is part of the tectonically higher units of the Lower Austroalpine nappe. Geochemical and petrological investigations reveal considerable differences between these mantle slices. The different ultramafic bodies are influenced by a pervasive regional metamorphism locally reaching amphibolite facies. Primary mineral assemblages of the peridotites have been mostly replaced by metamorphic parageneses. The products of metamorphism are dominantly serpentine minerals accompanied by various combinations of diopside, tremolite, chlorite, hydrogarnet and magnetite. In contrast to all other occurences, the samples from Nauders are characterized by a well preserved primary assemblage. Relic clinopyroxene and spinel can be found in peridotites of the Reckner and in the Matrei zone. Olivine and orthopyroxene pseudomorphed by serpentine minerals are present in less deformed areas only. Preserved clinopyroxenes of these mantle peridotites have up to 7 wt% Al2O3 and 1.5-2.0 wt% Na2O, which corresponds to a jadeite component of more than 10 %. The Cr2O3 content is generally high (up to 1 wt%). Clinopyroxene is frequently zoned with progressive depletion of Al, Na, Ti and Cr towards pure diopside. In samples from Nauders the Fo-content of olivine is about 90% and the NiO concentration varies between 0.4-0.5 wt%. The orthopyroxenes range in XMg from 91-92 and Al2O3 reaches up to 5 wt%. Titanian pargasite (XMg 0.88, about 3.5 wt% TiO2) is a secondary phase in Nauders peridotites only. Spinels from Nauders are different in composition to those of the Reckner complex. They are typically poor in Cr2O3 (about 7 wt%), high in Al2O3 (about 60 wt%) and develop rims which usually have lower Cr contents. Relics of spinel with significantly higher Cr values (Cr# 40-50, Cr2O3 35-40 wt%) are preserved in samples of the Reckner complex. The dominant rock types of all Mesozoic ultramafic bodies are residual lherzolites and harzburgites (XMg 87-94). The serpentinized ultramafics of the southern Penninic realm (Idalpe, Tauern and Rechnitz windows) are generally harzburgites with low Al2O3 (<2 wt%) and CaO contents (Fig. 1). The HREE concentrations (0.3-0.5 times chondrite) display patterns of moderately depleted restitic mantle (Fig. 2). Lherzolites and cumulate rocks (XMg = 0.80) are restricted to a few occurrences within the main mass of harzburgites. In contrast, more fertile lherzolites and harzburgites form ultramafic bodies of the Matrei zone, the Reckner complex, the Ramosch ophiolite and Nauders.The high abundances of Al2O3 (up to 4.6 wt%) and TiO2 (up to 0.22 wt%) are consistent with the estimated composition of a primitive upper mantle. All samples display higher HREE concentrations (up to 2.5 times chondrite) and a significant LREE depletion trend (Figs. 1 and 2). The Cr-Yb projection (Pearce and Parkinson, 1993) has been used to investigate the degree of mantle depletion. This diagram provides further evidence that the mantle material of the south Penninic region represents a more depleted source (about 15% melt depletion) than those of the Matrei zone and the Lower Austroalpine region (up to 5% melt depletion). These data support the model of a slow spreading oceanic environment (Hock and Koller, 1992) for the ophiolites of the Idalpe, Tauern and Rechnitz window, whereas the ultramafics of the Reckner, Matrei Zone and Nauders were probably generated in a pre-oceanic stage.

Spinel-peridotite xenoliths entrained by Plio-Pleistocene alkaline basic lavas from Sardinia (Italy) indicate a complex petrological history of the uppermost lithospheric mantle. They mostly show protogranular textures and are characterised by a four-phase equilibrated assemblage, ranging in composition from lherzolites (up to 18% of Cpx) to harzburgites, suggesting that the Sardinian subcontinental mantle underwent partial melting episodes with extraction of basic magmas. Trace element analyses (LAM-ICP-MS) and Sr- Nd isotope data, carried out on clinopyroxene (Cpx) separates, indicate a multistage history of depletion and enrichment processes. Clinopyroxene from Cpx-rich lherzolites are characterised by a LREE depletion, with low 87Sr/86Sr (0.70262-0.70391) and high 143Nd/144Nd (0.51323 - 0.51286) values, while clinopyroxene from less fertile lherzolites and harzburgites show LREE enrichments, higher 87Sr/86Sr (0.70410-0.70461) and lower 143Nd/144Nd (0.51288-0.51251). Nd model ages (relative to CHUR) of the most LREE-depleted samples, with 87Sr/86Sr<0.703, suggest that partial melting events occurred during Pre-Palaeozoic times. Modelling of the HREE distribution in clinopyroxene indicates that the Cpx-rich lherzolites could be interpreted as a residue after low (< 5%) melting degrees of an inferred fertile source, while higher melting degrees (up to 20-25%) are necessary to fit the Cpx composition of the most refractory harzburgites. Subsequent metasomatic processes are testified by the isotopic/ LREE enrichments, mainly recorded in the Cpx-poor peridotites. This fact implies that the most refractory domains of the mantle are more easily percolated by fluids, while Cpx-rich domains are less permeable to metasomatic agents, as indicated by experiments on melt connectivity in peridotite materials. Geochemical modelling suggests that the above mentioned enriched compositions can be obtained by metasomatising previously depleted mantle peridotite with a small amount (< 3%) of a strongly alkaline silicate melt. Neither the inferred metasomatic agents nor the Plio- Pleistocene Sardinian lavas show the HIMU geochemical imprint which, in addition to enriched mantle EM components, is recognised in Cenozoic anorogenic magmas throughout Central Europe (Wilson and Downes, 1991). The available data therefore indicate that the lithospheric mantle beneath Sardinia is heterogeneously enriched mainly by EM components, which reflect the complex multistage evolution occurring over the last 500 Ma. Similar geochemical features are observed in other samples of the European lithospheric mantle, such as the peridotite xenoliths entrained in alkaline lavas from the Massif Central (Zangana et al., 1997), and Tallante (Southern Spain; unpublished data). Analogous metasomatic enrichments can also be recognised in the most residual peridotites from the Pyrenean and Lanzo massifs (Bodinier et al., 1991; Downes et al., 1991). This suggests that the observed geochemical features were probably acquired during pre-Middle Mesozoic times, due to the repeated percolation of uprising EM metasomatic fluids in the European lithosphere. It should be emphasised that this metasomatic signature has not generally been observed for the lithospheric mantle of the African plate, where Cenozoic anorogenic magmas and associated mantle xenoliths are characterised by a prevalent HIMU metasomatic component (Beccaluva et al., 1998 and reference therein).

Cr- and Al-spinels from ultramafic rocks from the Iti and Kallidromon ophiolites were studied. Subhedral to anhedral Al-spinel, with lobate boundaries, occurs in lherzolite, usually surrounded and veined by a rim of magnetite. Cr-spinel in harzburgite is commonly subhedral to euhedral and shows two compositions, with Cr-rich areas irregularly distributed over Cr-poor ones. Ferritchromit and magnetite are developed along rims and fractures (Fig. 1a). Subhedral to euhedral Cr-spinel contained in dunite, commonly forms a core-to-rim zonation of Cr-rich to Cr-poor areas, with frequent ferritchromit rims (Fig. 1b). The two compositions of the harzburgitic Cr-spinels are slightly poorer in Cr, compared to the dunitic ones. The Cr# in the studied spinels displays a wide variability. The lherzolitic and harzburgitic Cr-poor spinels display low Cr# ( 60, characterize arcrelated ophiolitic sequences (Dick and Bullen 1984). The studied Al-spinels (with Cr# = 13.80 - 32.31) are analogous to those from abyssal peridotites (Fig. 2). The dunitic Crrich analyses (with Cr# = 83.06-87.05) are restricted at high Cr# values, while the dunitic Cr-poor ones (with Cr# = 60.33 - 75.53) show linear covariation with Mg# (Fig. 2). The harzburgitic Cr-poor analyses (with Cr# = 43.82 - 55.80) show linear covariation with Mg#, while the Cr-rich analyses have Cr# = 80.38 - 83.00. The dunitic and harzburgitic Cr-rich spinel analyses and the dunitic Cr-poor ones plot at high Cr#, analogous to those from both Alpine type and arc-related peridotites, but possess slightly lower Mg# values. The harzburgitic Cr-poor analyses are also similar to spinels from Alpine-type peridotites (Fig. 2). The studied spinels resemble those occurring in ultramafic rocks from Othrys, Pindos, Lesvos and Eastern Chalkidiki ophiolites. The lherzolitic spinels reveal a sympathetic increment of Cr# with Mg# indicating cogenetic relationship, while the dunitic and harzburgitic ones deviate and trend towards higher Cr#. The chemistry of the spinels from both Iti and Kallidromon suites, display a tectonomagmatic evolution from MORB-type affinities towards island-arc signatures. The lherzolitic, harzburgitic, and most of the dunitic spinel-olivine pairs plot within the OSMA (olivine-spinel mantle array; Fig. 3). The lherzolitic pairs plot in the abyssal peridotite field (although not diagnostic due to overlap with passive margin, oceanic arc and marginal basin peridotites) and close to the fertile mantle composition. The dunitic and harzburgitic Cr-poor spinel - olivine pairs are consistent with a SSZ origin. The dunitic Cr-rich spinel – olivine pairs lie within or near the OSMA and follow the trend defined by the fractionation line of boninites (Fig. 3). Textural evidence and linear evolution of Cr# and Mg# in the studied spinels reflects various degrees of partial melting, at least in the initial stages of the evolution of the host peridotites. However, the genesis of the Cr-rich dunitic and harzburgitic spinels may involve more complicated processes. According to Zhou et al. (2005), such high Cr spinels are related to crystallization from percolating SSZ boninitic melts, compatibly with the plots of the dunitic olivine – spinel pairs along the trend defined by the fractionation line of boninites (Fig. 3). The observed core-to-rim Cr2O3 depletion in the dunitic spinels is a common feature of Cr-spinel crystallised from melt (Leblanc and Ceuleneer 1992). Thus, differentiation of magma resulted in the formation of the Cr-poor compositional areas in the harzburgitic and dunitic spinels. The development of subhedral to euhedral grains in the harzburgites and dunites, in contrast to the anhedral lherzolitic Al-spinels, is also consistent with such an episode. The above evidence suggests a metasomatic origin for the Cr-spinels at the expense of the lherzolitic Al-spinels via boninitic melt-lherzolite interaction, in a SSZ regime. This is consistent with petrographic features and geochemical data from the peridotites, given elsewhere (Karipi et al., 2005). During this process, assimilation of clinopyroxene and orthopyroxene in the host lherzolite caused increment in the SiO2 content, in the percolating melt leading to the precipitation of olivine and Cr-spinel.

Abundant upper mantle (peridotite and pyroxenite) xenoliths occur in Miocene basaltic diatreme pipes and Quaternary lava flows on the Hyblean Plateau (Sicily, Southern Italy). Peridotites are spinel-facies protogranular-textured harzburgites and lherzolites (Fo90, En89, Cr-diopside, Cr-rich spinel), that equilibrated at temperatures between 950 and 1050°C. Even the freshest Hyblean peridotites exhibit some serpentine content. Major-element distribution and abundance reflect depletion (high MgO, low Al2O3) related to one or more melt extraction events, but evidence of modal and cryptic metasomatism does exist. For instance, phlogopite of metasomatic origin occurs rarely in these peridotites. Geochemical evidences of metasomatism includes Light Rare Earth Element (LREE) (e.g. LaN/YbN = 9-19). Sr-Nd and He isotopic signatures are also consistent with the refertilization of the lithospheric peridotite matrix (full descriptions in Sapienza and Scribano, 2000, Sapienza et al., 2005 and references therein). To better define the metasomatic reservoir, we performed in situ Os isotope analyses on sulphides from mantle peridotite collected from Valle Guffari Miocene diatreme. Eight peridotite xenoliths were cut in blocks, stuck on thin section glasses and polished on one side. Sulfides were imaged using a Cameca SX-100 electron microprobe. Then, Os isotope analyses were performed using a Merchantek LUV266 nm laser ablation microprobe (LAM) attached to a Nu Plasma multi-collector ICP (MC-ICP). The laser spot size was ~60 mm. Only 4 peridotite samples contain measurable sulfides, resulting in 11 analyses. Analytical details are in Pearson et al. (2002). Sulfides in the Hyblean peridotites are Ni-rich, often Ferich, and enclose a Cu-rich phase (Fig. 1). Their shape is variable, from spheroidal to irregularly shaped, usually displaying curvilinear margins (Fig. 1). The grainsize ranges from tens to ~250 mm. Sulfides are mainly related to the serpentine network. A few sulfides are olivine-enclosed but are too small to be analysed. 187O/188Os ratios range from 0.110408 to 0.124398, which correspond to sub-chondritic values. 187Re/188Os ratios range from 0.0296463 to 1.41346, i.e. supra- to subchondritic values (Fig. 2). Five out of 6 sulfides from one sample (GE12) show similar 187Os/188Os but different 187Re/188Os, while the remaining sulfide and the other samples all align along an oblique array showing a negative correlation between 187Os/188Os and 187Re/188Os. Assuming the 187Re decay constant after Smoliar et al. (1996), 187Re/188Os- CHUR and 187Os/188OsCHUR after Walker and Morgan. (1989), we calculate Re-depletion ages (TRD) and model age (TMA). TRD represents the minimum melt-depletion age of the lithospheric mantle, since the calculation minimizes the effect of metasomatism-driven Re addition, while TMA assumes that the considered mantle portion remains in closedsystem conditions through time (Walker et al., 1989). Low- 187Os/188Os sulfides from peridotite GE12 show Paleoprotezoic to Archean TRD ages, while the other sulfides are Neoand Mesoproterozoic. TMA yields meaningless age (future or even older than the Earth’s formation), reflecting recent disturbance of the Re/Os ratio. GE12 sulfides show the same unradiogenic 187Os/188Os composition (except for one sulfide) but 187Re/188Os ranging from 0.8-1.4 (Fig. 2). This indicates that Os isotope composition does not depend on time-integrated in situ 187Re, but rather suggests the metasomatic effects with Re addition probably occurring shortly prior the eruption. The negative correlation between 187Os/188Os and 187Re/188Os in the other samples also supports the metasomatic origin of Re, and rather suggests mixing between two reservoirs with different Re-Os signatures (Wang et al., 2003; Fig. 2). Thus, the Hyblean peridotites we have studied experienced melt extraction event(s) – which would remove Re from the system - and underwent later metasomatic Re-enriching event(s). Although these preliminary Re-Os data - and the large associated uncertainties - on peridotite sulfides do not allow the calculation of a very detailed ages, TRD ages are realistic and require a Proterozoic-Archean minimum age for these portions of the Hyblean lithospheric mantle. Most sulfides in peridotite GE12 testify to recent Re addition, while the other peridotite sulfides and one GE12 sulfide seem to be the results of mixing between two metasomatic end-members. The low-187Os/188Os sulfides in sample GE12 may be representative of the low-187Os/188Os endmember. The occurrence of sulfides with different Os isotope signatures in sample GE12 confirms the importance of using in situ technique in such petrological study.

The Pindos ophiolite represents non-metamorphic oceanic lithosphere obducted during the Jurassic. The mantle part was hydrothermally altered to different degrees during oceanization, but fresh peridotites are still preserved. The Pindos mantle mainly consists of highly depleted spinel harzburgites, but some plagioclase lherzolites are present in the northwestern part of the body. Textures in thin sections clearly demonstrate a direct relation between plagioclase formation and melt infiltration into the oceanic mantle. Two main textures were observed: a “porous flow” texture where plagioclase is homogeneously distributed in the rock and a channelized texture where plagioclase crystallized in microveins. Melt infiltration also led to crystallization of orthopyroxene and clinopyroxene associated with plagioclase. Similar textures were observed in the Othris ophiolite (Greece), southeast of Pindos, by Dijkstra et al. (2001) and in other Tethyan ophiolites (Piccardo et al., 2004). A study on light element contents of minerals from Pindos harzburgite and serpentinite was conducted in order to evaluate the role of the oceanic mantle as repository for these elements. Results obtained on spinel harzburgites and plagioclase lherzolites by SIMS (Secondary Ion Mass Spectrometry) show no significant difference between Li, Be and B contents of primary mantle minerals and melt impregnation phases. Li contents of olivine (0.68-1.1 ppm) and orthopyroxene (0.2-1.5 ppm) seem to be consistent with values for “normal” mantle minerals (Eggins et al., 1998; Woodland et al., 2004). The Li contents of clinopyroxene (0.2-3.7 ppm) are within the upper range of values published for unmetasomatised mantle clinopyroxene (only xenolith data available; Seitz & Woodland, 2000; Woodland et al., 2004). The Li characteristics of spinel and plagioclase peridotites differ from published data for non-metasomatised mantle by higher values for clinopyroxene than for olivine. This inverse Li partitioning could be explained by a reaction with a mafic silicate melt (Seitz and Woodland, 2000). This would be in agreement with the textures observed in the plagioclase lherzolites. Be abundances are below detection limit and B is low in all minerals from below the detection limit to 0.4 ppm. Li, Be and B contents in plagioclase are extremely low, often below the detection limit. In conclusion, the percolating melt did not add considerable amounts of B, Li and Be, probably because it was depleted in these elements. Melt impregnation is also supported by major element chemistry. Clinopyroxenes (first generation cpx and cpx recrystallized from the melt) of plagioclase lherzolites are enriched in Na, Ti and their XMg is higher compared to those of the spinel harzburgites. The same observation was made on the Othris peridotites by Dijkstra et al. (2001 and 2003). In plagioclase lherzolites, serpentinization textures confirm that hydrothermalism occured after melt infiltration. In highly hydrothermalized peridotites, light element contents of primary mantle minerals are low and similar to those of fresh samples. B is mainly concentrated in serpentine (0.4- 24 ppm) and Li in chlorite (2-3.5 ppm). Values for Be were always below the detection limit. Low contents of Li and B in primary minerals such as olivine, orthopyroxene and clinopyroxene after hydrothermalism can be explained by HT serpentinization.

The Tinaquillo orogenic spinel peridotite massif in north central Venezuela is a sub-horizontal, 3-km thick sheet overlying and in thrust contact with phyllites of the Cordillera de la Costa belt in the north and underlying and in extensional fault contact with the gabbroic to felsic Tinaco complex to the south (Ostos, 1990). While early workers interpreted the complex as a magmatic or crystal-mush intrusion (e.g., MacKenzie, 1960), Seyler and Mattson (1989, 1993) and Seyler et al. (1998) interpreted it as a fragment of the upper mantle possibly having ascended in the form of a diapir, but having been mylonitized during transit through the lower crust. The ultramafic rocks comprise ca. 75% harzburgite locally grading into lherzolite, 20% dunite, and 5% serpentinite. Pyroxenite and hornblendite veins intruding the complex are volumetrically small, but quite conspicuous where present. Minerals in the peridotite have a bimodal grain-size distribution owing to survival of remnant porphyroclasts of olivine, orthopyroxene, and minor clinopyroxene, and the late development of a mylonitic fabric with considerable grain-size reduction probably related to emplacement. The massif is thought to be a fragment of lithospheric mantle emplaced into the Caribbean belt during the late Cretaceous, but not exhumed until the late Eocene to middle Miocene by northwestward-directed thrusting along the Manrique Fault (Pindell et al., 1988; Pindell and Barrett, 1990). Seven peridoitites and three amphibole-rich veins have been investigated for major and trace element concentrations, and Sr-Nd-Pb-Hf isotopic compositions. The peridotites have geochemical characteristics of residues after moderate degrees of partial melting (estimated to be 8 to 16%) assuming a homogeneous primitive mantle source. In addition, they have strong light rare earth element (REE)- depleted to spoon-shaped REE patterns, but flat heavy REE (La/Yb at 0.04-0.55 times primitive mantle). The (La/Yb)N ratios increase with the refractoriness of the peridotites, reflecting secondary metasomatism, which is a common feature of the sub-continental lithospheric mantle xenoliths world-wide. A primitive mantle-normalized element distribution diagram for the peridotites reveals an overall depletion in the highly incompatible elements relative to less incompatible ones, but moderate enrichments in the strongly incompatible elements (Ba, Th, U), without depletion in Nb and Ta relative to La. The peridotites have isotopic compositions of 87Sr/86Sr = 0.70277-0.71044, 143Nd/144Nd = 0.513095-0.514000, 206Pb/204Pb = 19.00-19.15, 208Pb/204Pb = 38.18-38.75, and 176Hf/177Hf = 0.282692-0.284703. The Nd, Pb and Hf isotopic signatures show ranges that overlap with present-day mid-ocean ridge basalts (MORB) but also extending to more depleted compositions. The unusually radiogenic Sr for some samples appears to be the result of pervasive infiltration by seawater at some stage in the massif’s history. The Sr, Nd, Pb and Hf isotopic signatures of the veins (87Sr/86Sr = 0.70231-0.70256, 143Nd/144Nd = 0.513052-0.513289, 206Pb/204Pb = 18.30-18.52, 208Pb/204Pb = 37.66-37.87, 176Hf/177Hf = 0.282988-0.283303) are similar to the Pacific and Atlantic MORB; lack of the highly radiogenic Sr component in the veins indicates its addition to the peridotites prior to vein emplacement. The Lu-Hf data yield an errorchron suggesting a Phanerozoic age for stabilization of this piece of the lithosphere. We attribute the origin of the amphibole-rich veins in the peridotites to melting during subduction of the proto-Caribbean Plate in a back-arc tectonic setting, fairly recently in the massif’s history, possibly during the late Jurassic.

The Nikanbetsu complex, which is the neighbor complex of the Horoman peridotite complex, is located at the southern end of the Hidaka belt, which is characterized by a low- P and high-T type metamorphism in the Paleogene to early Miocene ages (e.g., Saeki et al., 1991; Arita et al., 1993). The complex is approximately 2 x 1 km in size and about 1300 m in thickness and is in fault contact with the surrounding metamorphic rocks. This complex is composed of mainly of plagioclase-bearing lherzolite with subordinate amounts of plagioclase-free lherzolite, harzburgite and two pyroxenes and olivine gabbro, minor amounts of clinopyroxenite and websterite (Takahashi, 1997). Harzburgite and lherzolite layers alternate with gabbro, and the layering is nearly parallel to the foliation. Plagioclase-rich veins are observed in plagioclase lherzolite and always have more than a few cm thick zone in which plagioclase is absent on the both sides. These veins are always parallel to the lineation and oblique to the foliation with 23~34° angle. The size of each vein is several tens centimeters to a few meter in length and a few centimeter in thickness. Several zones of abundant plagioclase-rich veins, that are several to a few hundred meters in thickness, occur in different lithology of gabbro layer-rich zones. Plagioclase lherzolite from the southwestern part of the complex only contains a round symplectitic aggregate, which is composed of spinel and two pyroxenes without plagioclase and olivine. Plagioclase (6 to 9 % in modal composition) in plagioclase lherzolite from the Nikanbetsu complex is classified into three type (Takahashi, 1997). These are clinopyroxene-bearing seam type, isolated type and vein type by its mode of occurrence. Plagioclase of the seam type occurs as one of minerals forming a seam-like aggregate. It is partly a product of the decompression reaction among two pyroxenes and spinel which were derived from the reaction between garnet and olivine as observed in the Horoman complex (e.g., Takahashi, 1992), and partly a product from interstitial melt. Plagioclase of the isolated type occurs outside plagioclase-rich seams as an interstitial isolated grain in both of pyroxenes-rich part and olivine-rich part. The modal amount of the isolate type which occur in olivine grains, is slightly increases from the southwestern to the central part of the complex, and significantly increases from the central to northeastern part of the complex. Plagioclase- rich veins are composed mainly of plagioclase and coarse orthopyroxene. Plagioclase from the Nikanbetsu complex shows also Na-Ca zoning, which has An-poor core (An 61~66) and Anrich rim (up to An 90). Large isolated plagioclase grains show a W-shaped Na-Ca zoning pattern; in the core region the An content slightly decreases from 70 to 65 followed by an increase up to An 85 at the rim. And they often have oscillatory zoning at their margin to rim. All plagioclase grains from plagioclase-rich veins have also compositional zoning similar to the other types. Olivine occurring in the vicinity of such the plagioclase-rich vein has Fo-poor (89~89.5) and NiO-poor (0.33~0.35 wt%) characters as compared with olivine far away from the vein (90~91 in Fo content, 0.36~0.38 in NiO wt%). Some olivine from the area where abundant plagioclase-rich veins and isolated plagioclase occur, shows also NiO-poor (0.34~0.35 wt%) character. The existence of two pyroxene and spinel symplectite from the southwestern part of the complex indicates that the Nikanbetsu complex started to ascend from the garnet stability field as advocated from the Horoman complex (Ozawa and Takahashi, 1995). Calculated temperature for corecore pairs of coarse pyroxenes according to Wells (1977) indicates more than 1100°C at the southewestern part of the complex, to 1700°C at the northeastern part of the complex, which is similar to the highest value for the Upper Zone of the Horoman complex. The texture of isolated type of plagioclase interstitial to pyroxenes grains suggests that it was crystallized from a very small amount of partial melt formed in the vicinity of clinopyroxene and orthopyroxene grains. The W-shaped Na-Ca zoning pattern of large isolated plagioclase grains suggests that a rapid decompression, which caused the marginal Ca enrichment under above or below solidus conditions, took place after crystallized of a trapped partial melt. Mode of occurrence of oblique plagioclase-rich veins with plagioclase-free vein wall indicate that a crack suck partial melt from the surrounding partially molten peridotite (e.g., Nicolas, 1989). However, the Fe-rich and NiOpoor nature of the surrounding olivine indicates that the melt was not simply extracted from the host peridotite, because the simple melt extract will cause Mg and Ni enrichment of the surrounding olivine. The sharply decreasing trend for Ni content of the olivine in the surrounding zone of the vein can be explained by crystallization from a melt in equilibrium with the plagioclase lherzolite host away from the veins. Olivine is though to have been crystallized on the vein wall from the segregated melt. There may have been a modification by a reaction with an evolved melt derived by olivine fractionation from the segregated melt. This mechanism, however, requires some extent of olivine fractionation to produce less magnesian melt, however, requires olivine crystallization on the vein wall. The variation of the core composition of plagioclase within plagioclase-rich veins featuring Ca enrichment toward the vein wall indicates that plagioclase was crystallized on the vein wall at first, followed by filling up the center of the vein, because plagioclase at the center of the vein is more sodic. The existence of Na-Ca oscillatory zoning pattern of large isolated type of plagioclase from plagioclase lherzolite strongly indicate that the incipient melt was formed within several limited parts of the lherzolite, and each melt may have migrated independently.

Refractory spinel peridotites were drilled during Leg 125 from two diapiric serpentinite seamounts: Conical Seamount in the Mariana forearc (Sites 778-780) and Torishima Forearc Seamount (Sites 783-784) in the Izu-Ogasawara forearc. Harzburgite is the predominant rock type in the recovered samples, with subordinate dunite; no lherzolite was found.The harzburgite is diopside-free to sparsely diopside-bearing, with modal percentages of diopside that range from 0% to 2%. Spinels in the harzburgites are chrome-rich (Cr/[Cr + Al] = 0.38-0.83; Fe3+/[Fe3+ + Cr + Al] = 0.01-0.07). Olivine and orthopyroxene are magnesian (Mg# = 0.92). Discrete diopsides reveal extreme depletion of light rare earth elements. Primary hornblende is rare. The bulk major-element chemistry shows low average values of TiO2 (trace), Al2O3 (0.55%) and CaO (0.60%), but high Mg# (0.90).These rocks are more depleted than the abyssal peridotites from the mid-oceanic ridge. They are interpreted as residues of extensive partial melting (= 30%), of which the last episode was in the mantle wedge, probably associated with the generation of incipient island-arc magma, including boninite and/or arc-tholeiite. These depleted peridotites probably represent the residues of melting within mantle diapirs that developed within the mantle wedge.

Sulfides are the main hosts for highly siderophile elements (HSE); therefore the mantle S content is a crucial datum for geochemical studies relevant to platinum-group element abundances or Re-Os systematics. The estimated S content of the primitive mantle is about 250±50 ppm. Although S contents of fertile peridotite from orogenic massifs support this estimate, peridotites hosted in alkali basalt commonly contain only a few tens of ppm S, whatever their fertility (Lorand, 1990). Some xenoliths from southern France, i.e. Montferrier volcano (Herault) have very high sulfide abundances (up to 0.2 wt.%) with intergranular sulfides predominating over enclosed sulfides (Lorand and Conquere, 1983). The process responsible for such high sulfide modal abundances and its possible effect on HSE abundances are not clear. To shed some light on this process we have carried out an extensive study of 21 xenoliths from this locality, involving major elements and lithophile trace element data, chalcogenide (S, Se) geochemistry, platinum-group element analyses and Re-Os systematics. Montferrier xenoliths are either breccia-hosted or lavahosted. The former have a yellowish color indicating weathering, while the lava-hosted have a fresh green color and do not show obvious alteration features. Some xenoliths display disseminated Ti-pargasite (up to 6%) regardless of their host. Montferrier peridotites are characterized by a strongly deformed porphyroclastic microstructure giving them a mylonitic appearance. It has been shown that this deformation took place at low temperature (900-750°C), probably during an uplift on the nearby Cevennes lithospheric shear zone (e.g., Fabries et al., 1987). They also display very fertile compositions. Most of the studied samples contain between 3.0 and 4.5% Al2O3 and the least fertile still contains 2.0 wt.% Al2O3. The S concentration range (7 - 592 ppm) is the highest reported for basalt-hosted peridotite xenoliths. S contents are uncorrelated with fertility indices (e.g., Al2O3%). Breccia- hosted xenoliths have the lowest S contents ( 10). Ba and Rb enrichments are determined by amphibole. The second end-member is strongly enriched in LREE (La/SmN > 1). This LREE enrichment, is accompanied by large ion lithophile elements (Sr, U, Th...), but without concomitant enrichment of the high field strength elements (Nb, Ta, Zr, Hf). In contrast to the depleted type, the U/Th ratios are chondritic. Such enriched patterns are commonly ascribed to mantle metasomatism by carbonated and/or volatile-rich small volume melts. The positive correlation of figure 1B indicates that large amounts of S have been added by this metasomatism. In thin sections, the xenoliths containing more than c.a. 200 ppm S show numerous large intergranular sulfide blebs (up to 700 µm in maximum dimension) much richer in Fe and Cu than those from the depleted samples. Contrary to S, Se contents increase only moderately from 17-48 ppb in depleted samples to 50- 63 ppb in enriched xenoliths. The most S-rich samples may show markedly higher than chondritic S/Se ratios (104) which are inconsistent with a simple addition of immiscible sulphides during metasomatism. This observation suggests that S was a dissolved species in the volatile-rich metasomatic fluid whose effects are indicated by the trace element studies. Apart from the most refractory lherzolite which is unusually poor in PGE (total PGE = 5.2 ppb), the 8 xenoliths analyzed for the PGEs show a narrow PGE concentration range (20-28.2 ppb) independant of their fertility indices and S contents. Ru/Ir and Rh/Ir ratios are remarkably constant and slightly higher than chondritic. Pd/Ir ratios (1.37-2.26) are suprachondritic, as reported for Pyrenean orogenic lherzolites (Pattou et al., 1996) and some abyssal peridotites (Snow and Schmidt, 1998, Luguet et al., this volume). Pd/Ir ratios are correlated with S contents, mostly because the enriched samples have low Ir and Ru contents. Whole-rock fractionated Pd/Ir ratios are reflected in the pattern of the sulphides (LA-ICP-MS data). Os/Ir varies sympathetically with Pd/Ir, increasing from nearly chondritic values (1.1) in the depleted sample to suprachondritic (1.99) in the most S-rich xenolith. 187Os/188Os shows a similar increase from DMM-like values (0.128) in the depleted samples to highly radiogenic values (0.175) in the most S-rich sample. Our study provides further evidence that metasomatic process(es) involving volatile-rich small melt fractions may significantly alter the HSE relative abundances and Os isotopic composition of the lithospheric mantle (cf Gutierez Narbonna et al., this volume). Such parallel variations of Os, Pd and S are consistent with the experimentally-demonstrated affinity of these elements for S-bearing vapor phases at magmatic temperatures.

MELT DIFFUSE REFERTILIZATION AND FOCUSED MIGRATION IN LITHOSPHERIC MANTLE OF A FOSSILE RIFTED MARGIN: THE MT. NERO PERIDOTITE (EXTERNAL LIGURIDES, ITALY)

The Uralian chain (Russia) is a 2500 km long linear mid- Paleozoic orogenic zone representing the closure of an ocean basin-island arc system between the European plate to the west and the Siberian plate to the east. As a result, sections of oceanic lithosphere were obducted onto the European plate producing more than 150 ophiolite occurrences with a size range from 200 km to minor smaller occurrences (Savelieva and Nesbitt, 1996). The Nurali massif is a mafic-ultramafic complex located in the Southern Urals and belonging to the belt of small dismembered ophiolites occurring along the Main Uralian Fault, the suture zone of the Ural orogenic belt. The massif crops out over an area of 45 km2 and extends for about 20 km. From west to east the Nurali massif includes a lherzolitic mantle section (lherzolite, harzburgites, dunites, rare chromitites), an ultramafic cumulate sequence (dunites, wehrlites, pyroxenites, chromitite lenses), and an amphibole gabbro cut eastward by a melange zone (andesites, pillow lava, marbles, skarns, mafic and ultramafic rocks within a serpentinite matrix). The geodynamic evolution of the massif has been long debated. While some Authors consider it as an ophiolite related to aborted slow-spreading zone (rift) within continental margin, others consider it as an orogenic lherzolite massif. The chromitite lenses belonging to the Nurali massif show a wide range of textural and chemical features strictly related to the their host rock. The massive chromitite within the mantle portion occurs as a small metric lens, settled in a broadly serpentinized lherzolitic unit whose spinel is rimmed by and partially reacted with plagioclase. The compositions of Cr-spinels from different mantle rocks (lherzolite, harzburgite and dunite) and from the chromitite lens, draw two different parallel trends with negative correlation between xCr and xMg. The first one, marked by the lherzolitic and harzburgitic rocks, shows Mg# and Cr# between 0.40 - 0.60 and 0.35 - 0.50 respectively. The second one, related to dunite and chromitite, shows values of Mg# and Cr# between 0.30 - 0.60 and 0.60 - 0.75 respectively. The PGE analysis of the chromitite shows a general enrichment in Os (65 ppb), Ir (47.5) Ru (50 ppb) and Rh (9.7 ppb).The REE content of clinopyroxenes within the chromitite lens shows a linear pattern with chondrite- normalized values ranging from 1.55 (LaN) to 4.38 (YbN). The ultramafic cumulate sequence consists of a banded dunite-wehrlite-pyroxenite unit cropping out along the eastern flank of the Nurali massif, in a zone about 20 km long and 0.2-1.3 km wide. The sequence was interpreted as formed by multiple injections of melt derived from a depleted mantle source in a spreading geotectonic setting (Pertsev et al., 1997). Five chromitite lenses were found within the cumulate sequence: the first one located 0.3 km to the north of Miass river (North Miass), the second one about 5 km to the south of Miass river (Jalchigulovo), and the last three ones, in a floor to the top section of the cumulate sequence about 0.5 km to the south of Miass river, (South Miass; levels 1 to 3 of Grieco et al., 2004). Chromitite bodies’ size, shape and composition change within the different levels. Level 1 consists of centimetric irregular yet flattened and massive lenses, hosted in a clinopyroxenite; this level has chromites with the highest Cr# (between 0.65 and 0.71) and the lowest Mg# (between 0.48 and 0.50) and with peculiar high Fe3+ content. Level 1 shows intermediate contents of PGE (1449 ppb) and is enriched in Pt and Pd. Level 2 is made up of a 3 _ 1.5 m deformed massive chromitite lens and of a another smaller and disseminate body, both with a orthopyroxene-amphibole-rich matrix and hosted within a spinel-bearing amphibole-rich olivineclinopyroxenite rock layer. This level has the lowest and the most variable Cr#, ranging between 0.33 and 0.54, and a high Mg# (between 0.60 and 0.80). This level shows a strong PGE enrichment (up to 26.754 ppb), mostly in Os, Ir and Ru. Level 3 consists of a decimetric irregular and very massive lens hosted in olivine-enstatite rocks, and it has been correlated to North Miass and Jalchigulovo chromitite bodies by both textural-petrological evidences and geochemical analyses. Level 3 has high Cr# (between 0.62 and 0.72) and the highest Cr2O3 content, ranging between 47 and 52 wt%; this level is depleted in all PGEs compared to the other levels. Within the serpentinite melange four main chromitite localities were investigated, two of which were mined for chromium during the World War II. Here chromitite occurs mainly as metric to decametric massive lenses within serpentinite whose protholith has been completely obliterated. In some cases late veins of quartz and aragonite cut chromitite. Mineralization often has cataclastic textures with rare undeformed crystals. The composition of chromites from Nurali serpentinite melange shows very high Cr# (0.75- 0.85) and highly variable Mg# (0.11-0.71) and chromite grains are usually rimmed by magnetite. TiO2 contents are lower than 0.25 wt%. Low Mg# chromites are associated to green hornblende and Cr-chlorite (kammererite). PGE contents are very low (PGEtot <150ppb), Ru is the highest, (up to 64 ppb), Os and Ir have a intermediate value (33 and 38 ppb) and Pt and Pd are below detection limits ( < 5 ppb). Only two primary Os-Ir alloys ,2-3 µm in size,have been found within chromites.

The Nurali massif is one of the several mafic-ultramafic bodies located along the Main Uralian Fault (MUF), the suture zone of the Ural orogenic belt, thrusted westward over the East European continental margin. Nurali consists of a spinel±plagioclase lherzolite/ harzburgite/ dunite mantle section, bordered to the east by a layered dunite/olivineclinopyroxenite/ wehrlite cumulate sequence, an amphibolerich gabbro-diorite and the serpentinite melange of MUF. The cumulate sequence was interpreted as the one formed by multiple injections of melt derived from a depleted mantle source in a spreading geotectonic setting (Pertsev et al., 1997). South of Miass river, three chromitite-hosting levels at different stratigraphic heights were found in a full section of the cumulate sequence (Grieco et al., 2004). The present work deals with deeper insight of PGE-rich (up to 26.754 ppb) second level chromitite and with correlations between the section of Grieco et al. (2004) and two more chromitite outcrops respectively to the north (North Miass) and to the south (Jalchigulovo, Colpani, 2003) of the South Miass section, and with their relationships with the host rocks. Level 2 comprises a 3 _ 1.5 m deformed massive chromitite lens and a smaller chromitite body next to it, both with orthopyroxene-amphibole rich matrix and orthopyroxenitic halo, hosted within a spinel-bearing amphibole-rich olivineclinopyroxenite/ olivine-orthopyroxenite rock layer. The North Miass chromitite body is a decimetric highly massive lens hosted within a clinopyroxenite rock. Textural and petrological analyses highlight the crustal nature of the harzburgite-type cumulate sequence, where the cumulus texture is often evident: amphibole is a very common phase, ranging in composition within the fields of magnesio- hornblende, tschermakitic, edenitic and pargasitic hornblende. It forms at least two generations of amphiboles: the intercumulus phase, which testifies the hydrous condition of crystallisation of the cumulate sequence, and a late hydrate phase, which reacts with spinels, olivines and pyroxenes in all the samples of the cumulate sequence and testifies the role of late fluids that tend to obliterate the original texture of the rock. The chromitite orthopyroxene-amphibole-rich matrix and the orhopyroxenitic halo show a non-cumulate texture where orthopyroxene crystals are arranged in polygons, suggesting a strong genetic relationship between the chromitite lenses and their halo. The main metallic-mineral paragenesis of both level 2 and North Miass chromitites are chromite, scattered sulphides, such as pentlandite, heazlewoodite and laurite, and platinum group alloys. The data concerning mineral chemistry of chromite show very low contents of Cr2O3, ranging between 32 wt % and 44 wt % for the level 2 whereas it is up to 58 wt % for North Miass chromitites. According to xCr-xMg, Cr2O3-Al2O3 and MgO-FeO diagrams, spinels from the orthopyroxenitic halo are comparable with the level 2 chromites, together with which they form patterns separated from those belonging to the scattered spinels of the cumulate rocks. The level 2 chromitite has the highest xMg of all the South Miass section in contrast to the cumulate sequence genetic model of Pertsev et al. (1997), in which the cumulate sequence is formed by several multistage magma batches, with an increasing Mg content from the bottom to the top. All these diagrams also emphasize the correlation between North Miass chromitite body, the third level of Grieco et al. (2004) and Jalchigulovo chromitite body (Colpani, 2003). Mineral chemistry of amphiboles confirms the presence of more than two generations of such mineral, characterized by different TiO2 contents: amphiboles of the level 2 chromitite matrix, intercumulate amphiboles, amphiboles from impregnations and amphiboles of late amphibolic veins. REE patterns show the strong enrichment in all REE in amphiboles of the level 2 chromitite, where amphiboles of the level 2 chromitite matrix have flat patterns and are up to 50 times enriched compared to chondritic abundance, and amphiboles from both impregnations and late amphibolic veins are 10 times enriched compared to coexisting clinopyroxenes; orthopyroxene patterns have a typical LREE depletion compared to HREE; the high content of all REE with respect to chondritic abundance suggests the role played by melts/fluids which would have favoured the remobilization of both compatible and incompatible elements towards the chromitite body. PGE pattern shows an enrichment in Os, Ir and Ru of the level 2 chromitite body, and confirm the geochemical correlation between North Miass, Jalchigulovo and the South Miass third-level chromitite: this suggests that chemical conditions of formation and relative PGE distribution of these chromitite bodies are not due to a local process but are probably the result of processes that involved the whole extension of the cumulate sequence. As a summary, textural, mineralogical and geochemical evidences suggest that the level 2 chromitite body and its orthopyroxenitic halo cannot be related to the same cumulus processes which formed the cumulate sequence, confirming the important role played by metasomatic/hydrothermal processes in their formation. We propose a metallogenic model in which metasomatic fluids can separate from residual intercumulus melts; once PH2O of the fluid reaches 2 Kbar, the first phase to be stabilized is spinel followed by orthopyroxene (Nicholson and Mathez, 1991). The effect of evolved fluids interacting with cumulate host rocks is shown in the Ol-Chr-Q system (Zhou et al., 1994), where chromite can precipitate owing to the interaction between fluids and host rocks or to a change in fluid composition, which moves into the chromite stability field. The fractionation of chromite drives the fluid composition towards chr-opx cotectic, with the consequent co-precipitation of chromite and orthopyroxene and the formation of an orthopyroxenitic halo.