- © 2016 Mineralogical Society of America
The highly siderophile elements (HSE) consist of the Platinum Group Elements (PGE: Ru, Rh, Pd, Os, Ir, Pt) along with rhenium and gold. These transition elements show relative chemical inertness and high market values, which respectively earned them the additional names of noble metals and precious metals.
The HSE show a very pronounced affinity for iron metal, which translates into metal/silicate partition coefficients similar to or higher than 10,000 over large ranges of both pressure and temperature (e.g., O’Neill et al. 1995; Borisov and Palme 2000; Ertel et al. 1999, 2001, 2006, 2008; Fortenfant et al. 2003, 2006; Brenan et al. 2005; Cottrell and Walker 2006; Brenan and McDonough 2009; Laurenz et al. 2010; Mann et al. 2012; see Brenan et al. 2016, this volume for detailed review). Consequently, the HSE are thought to have been efficiently sequestered within the metallic core of our planet during the metal–silicate differentiation of Earth, leaving the silicate counterpart almost HSE-barren. Investigations of mantle peridotites since the 1970s revealed ng.g−1 level abundances as well as close-to-chondritic proportions of the HSE (Chou 1978; Jagoutz et al. 1979; Mitchell and Keays 1981; McDonough and Sun 1995; Becker et al. 2006; Fischer-Gödde et al. 2011). Such abundances and inter-HSE fractionations are not predicted for the silicate Earth left after separation of the metallic core for low- or high-pressure core–mantle differentiation (see Brenan et al. 2016, this volume). The close agreement between the osmium isotopic compositions of fertile mantle peridotites and those of chondritic meteorites (Walker et al. 2002a), which requires nearly identical Re/Os ratios in these two reservoirs, provides particularly convincing evidence that the mantle’s HSE content cannot simply represent the residue left after core formation. These observations have led to suggestions that the HSE systematics in the Earth’s mantle could in fact reflect 1) inefficient core formation (Arculus and Delano 1981; Jones and Drake 1986), 2) repeated equilibrium-fractionation events, each involving only part of the mantle (Azbel et al. 1993), 3) core–mantle differentiation followed by core–mantle interaction (Snow and Schmidt 1998), and/or 4) a late accretionary event, also known as the “late veneer” hypothesis (e.g., Kimura et al. 1974; Chou 1978; Jagoutz et al. 1979; Morgan et al. 1981, 2001; Pattou et al. 1996). Among these, the last scenario, consisting of global-scale metal–silicate differentiation followed by late addition of an extraterrestrial component 4.2–3.8 Ga ago, after core formation has ceased, is currently the most favoured hypothesis, although some issues remain unresolved (Walker 2009; see review of Yokoyama and Walker 2016, this volume). It has also been proposed that a late heavy bombardment event of this type established the HSE abundances and 187Os signatures in the silicate Moon (Walker et al. 2004; Day et al. 2007; see detailed review in Day et al. 2016, this volume). If such a 2-fold scenario (core–mantle differentiation followed by late reintroduction of HSE) accounts for the HSE signatures in the Earth’s mantle, the relative HSE abundances within rocks from the terrestrial mantle (as well as in the lunar mantle) should hold the key to the nature of the extraterrestrial component, which “refertilized” the silicate Earth–Moon system in HSE. This late accreted material has also been postulated to have (re)-introduced volatiles and the organic molecules necessary for the emergence of life on our planet (Cooper et al. 2001; Kring and Cohen 2002). Therefore, understanding the origin of the HSE signatures may help us to understand when and how our planet acquired a favorable environment for the development of life. Identifying the nature of this late meteoritic bombardment would furthermore provide firm constraints on the origin of the impactors and thus on models of the formation and early evolution of the inner solar system.
CONSTRAINING THE HSE AND 187Os/188Os ISOTOPIC COMPOSITION OF THE PRIMITIVE BULK SILICATE EARTH
The primitive Bulk Silicate Earth that was left after core–mantle differentiation and the late bombardment refertilization is now segregated into two major silicate reservoirs, the mantle and the crust. The crust, oceanic and continental, is volumetrically and gravimetrically minor, representing only ca. 1% of the present day Bulk Silicate Earth. This, combined with the much higher abundances (by up to 3 orders of magnitude) of HSE in mantle peridotites compared to typical crustal rocks, implies that extraction of the continental crust has had very little influence on the average HSE composition of the present day mantle, making mantle peridotites the most suitable lithology for constraining the HSE and Os isotopic compositions of the Primitive Bulk Silicate Earth, hereafter referred to as the Primitive Mantle.
Two types of natural samples of mantle rocks are found at the surface of the Earth. The first consists of km-size slices of mantle material tectonically emplaced within the crust. Such peridotite massifs include orogenic, ophiolitic, and abyssal peridotites. The second type consists of relatively small (up to several kg) nodules or “xenoliths” of peridotites, which are typically brought to the surface by mafic magmas (primarily alkali basalt or kimberlite/lamproite) and more rarely by felsic magmas (e.g., andesitic–dacitic and phonolitic volcanism). Both occurrences offer advantages and disadvantages when trying to understand the composition of the terrestrial mantle. As a consequence of their kilometric size, peridotite massifs provide a more global view and allow us to assess the 3D relationships of the different lithologies present in the Earth’s mantle, i.e., peridotites, including lherzolites, harzburgites, and dunites, and pyroxenite veins of a variety of compositions. However, due to their tectonic emplacement and long residence time within the crust, they have experienced re-equilibration at relatively low temperatures, and are often strongly serpentinized. On the other hand, peridotite xenoliths are at best metre-sized nodules, but more commonly less than 20-cm-sized rounded fragments, which may show reaction rims (few cm thick) with their host lavas. Peridotite xenolith occurrences are geographically more widespread than are peridotite massifs, which are confined to orogenic zones and abyssal plains. Because of their rapid ascent from mantle level to the Earth’s surface, peridotite xenoliths are also generally much less strongly serpentinized than massif peridotites, and thus in certain respects may better preserve the mineralogical and chemical signatures of their mantle source (see Pearson et al. 2003 and refs. therein). Though xenoliths are almost always found in volcanic host rocks of Phanerozoic age, most likely for reasons of preservation, the mantle sampled by xenoliths spans a much larger age range (Mid-Archean to Holocene) than the mantle represented by peridotite massifs. Kimberlitic to lamproitic volcanism has excavated many Archean peridotite xenoliths, both within cratons and in mobile or orogenic belts directly surrounding cratonic blocks (see Aulbach et al. 2016, this volume, for detailed review). On the other hand, alkali-basaltic volcanism, like orogenic massifs, provides samples of mantle peridotite of Proterozoic to Holocene age. Basalt-hosted peridotite xenoliths have been found in a wide variety of mostly non-cratonic tectonic settings. On the continents, these include zones of incipient rifting (e.g., East Africa Lorand et al. 2003a; Southwestern United States, Burton et al. 1999; Harvey et al. 2011; Baikal rift zone, Russia, Pearson et al. 2004), backarc extensional and cordillera regions broadly associated with subduction (e.g., Brandon et al. 1999; Peslier et al. 2000a,b), orogenic mobile belts at the periphery of cratons (e.g., Trans North China Orogen; Gao et al. 2002; Liu et al. 2010), and regions that have experienced lithospheric thinning and replacement (e.g., eastern China, Wu et al. 2003; Zheng et al. 2007; Liu et al. 2011; Sierra Nevada, California, Lee et al. 2000). In the oceans, peridotite xenoliths are found in volcanism associated with active hot spots such as Hawaii (Bizimis et al. 2007), in ocean–ocean collision zones such as Kamchatka (e.g., Widom et al. 2003) and Papua New Guinea (McInnes et al. 1999) and in more complex settings such as Kerguelen (e.g., Hassler and Shimizu 1998; Lorand et al. 2004).
Ultimately, mantle peridotite massifs and lava-hosted peridotite xenoliths allow us to look at the Earth’s mantle from different perspectives, providing complementary information on its structure and composition and possible secular evolution. Nevertheless, peridotite massifs and peridotite xenoliths only sample the uppermost part of the upper mantle (maximum 250 km depth) and as such can be used to reconstruct the composition of only the Primitive Upper Mantle (PUM), a hypothetical fertile upper mantle reservoir that did not experience crustal extraction or other differentiation processes. This composition can be extrapolated to the entire Primitive Mantle (i.e., Primitive Bulk Silicate Earth) composition only if it is assumed that the terrestrial upper and lower mantle are similar in terms of HSE and Os isotope composition or any other chemical element of interest. Furthermore, considering the long time span, sometimes exceeding 3 billion years, since their stabilization in the Earth’s lithospheric mantle, the sources of both ultramafic massifs and xenoliths have recorded a complex petrological history, possibly including multiple partial melting events and episodes of interaction with percolating melts or fluids. This multistage petrological history and its possible effects on the peridotite geochemical and isotopic composition have to be considered before estimating the primitive mantle composition and characterizing the nature of the late veneer and its implications for the formation and evolution of our solar system.
This Chapter aims to provide a comprehensive review of how the study of non-cratonic peridotite xenoliths has contributed over the past five decades to the evolution of our understanding of HSE and Os isotope systematics in the Earth’s mantle. This review is based on ca. sixty studies presenting whole-rock, mineral and/or in situ HSE ± Re–Os isotope data; only a very few of these combined petrographic information about the HSE host phases with geochemical and isotopic approaches. In order to avoid prejudging tectonic context, we define non-cratonic peridotite xenoliths as all of those brought to the surface by basaltic volcanism, or very rarely, more evolved magmas. This chapter is directly complementary to Lorand and Luguet (2016, this volume), focusing on the chalcophile/siderophile trace elements in mantle rocks, Aulbach et al. (2016, this volume), which explores our knowledge of the HSE and Os isotopes in cratonic mantle, Becker and Dale (2016, this volume), whose topic is the Re–Os and HSE signatures in tectonically emplaced mantle peridotites, and Harvey et al. (2016, this volume), which deals with the Re–Os–Pb systematics of mantle sulfides and their utility in mantle geochronology, and Gannoun et al. (2016, this volume), which focuses on Re–Os and PGE behavior of basaltic volcanism.
PETROLOGY AND LOCATION OF NON-CRATONIC PERIDOTITE XENOLITHS
Alkali-basaltic volcanism brings nodules of widely varying lithology to the surface. Mantle peridotites are thus commonly associated with other mantle (pyroxenites) and crustal (gabbros, granitoids) xenoliths. In contrast to cratonic peridotite nodules, which are very frequently garnet bearing, non-cratonic basalt-borne peridotite xenoliths are exclusively spinel peridotites, indicating derivation from above the garnet–spinel transition at about 80 km. Petrographically, they range from “fertile” lherzolite to “highly depleted” harzburgite and dunite and consist primarily of olivine, orthopyroxene, clinopyroxene and spinel. Their whole-rock major element compositions and compatible trace element abundances, along with the major element compositions of the modally-dominant minerals (especially olivine), suggest that mantle xenoliths generally experienced less than 15–20% partial melting, mostly at shallow depths. Nevertheless, detailed petrographic investigations often reveal the presence of minor to trace abundances of volatile-rich minerals such as amphibole, phlogopite, or apatite, as well as carbonates, in addition to non-volatile-bearing minor phases such as Ti-oxides. These accessory minerals have been observed in nearly all peridotite-xenolith localities but not always in all the xenoliths studied from a given locality. The occurrence of such minerals along with incompatible lithophile trace element-enriched patterns (e.g., relative enrichment in large ion lithophile elements (LILE) such as Rb and in light rare earth elements (LREE) as evidenced by (La/Yb)N > 1, N = primitive mantle-normalization after McDonough and Sun, 1995) argue for a fairly ubiquitous overprinting of the sampled lithospheric mantle by percolating melts or fluids. The frequent association of such enriched patterns with “depleted” Rb–Sr, Sm–Nd, and Lu–Hf isotopic signatures (i.e., indicating long-term residence in a mantle source depleted in incompatible elements) suggests that much of this overprinting was recent, and thus possibly related to the tectono-magmatic processes that brought the xenoliths to the surface, for example by infiltration of host lavas (Downes 2001). The large variety, in both nature and degree, of incompatible trace element enrichments and isotopic signatures and volatile-rich minerals reflects 1) the mode of melt/fluid percolation, which may range from channel-like to porous or diffuse percolation, and which can occur at variable melt/rock ratios and 2) the large compositional range of fluids and melts, varying from silicate to carbonatitic melts, with various degrees and patterns of volatile enrichments, to oxidizing S- and Cl-bearing fluids or vapor phases. All of these melts and fluids are able to modify the post-melting composition of the Earth’s mantle and profoundly complicate estimations of the Primitive Upper Mantle (PUM) composition. These multistage complex petrological histories should thus be understood and their effects on the HSE and Os isotope systematics firmly constrained before attempting to reconstruct the PUM composition.
Geographically, non-cratonic peridotite xenoliths sampled by basaltic volcanism and whose HSE and/or Os isotope signatures have been investigated spread over the 7 continents and 3 oceanic plates. Sampling locations include:
North and Central America: along the Rio Grande Rift structure (Mitchell and Keays 1981; Morgan et al. 1981; Morgan 1986; Meisel et al. 1996, 2001; Hart and Ravizza 1996; Burton et al. 1999; Harvey et al. 2011; Byerly and Lassiter 2012); in the Basin and Range and Sierra Nevada provinces of California (Lee et al. 2000; Lee 2002; Armytage et al. 2014); and in the Northern US–Canadian Cordillera (Brandon et al. 1996, 1999; Peslier et al. 2000a,b)
Australia: in the Quaternary Newer Volcanic Area in SE Australia (Mitchell and Keays 1981; McBride et al. 1996; Handler et al. 1997, 1999; Handler and Bennett 1999; Alard et al. 2000, 2002), in the New South Wales Snowy Mountains (Morgan and Lovering 1967; Powell and O’Reilly 2007) and in Northern Queensland (Morgan et al. 1981; Morgan 1986; Handler and Bennett 1999; Handler et al. 1999, 2005; Meisel et al. 2001)
Europe: often in close association with the Cenozoic European Rift System (Meisel et al. 1996, 2001; Lorand and Alard 2001, Pearson et al. 2002; Alard et al. 2002, 2011; Schmidt et al. 2003; Ackerman et al. 2009; Harvey et al. 2010; Fischer-Gödde et al. 2011; González-Jiménez et al. 2013, 2014)
Asia: including the lithosphere of eastern China (Gao et al. 2002; Wu et al. 2003, 2006; Reisberg et al. 2005; Zheng et al. 2007; Xu et al. 2008a; Zhang et al. 2008, 2011, 2012; Chu et al. 2009; Hong et al. 2012; Yu et al. 2012); central to western China, including localities such as Hannuoba in the Trans North China Orogen (Gao et al. 2002; Becker et al. 2006; ; Xu et al. 2008b; Zhang et al. 2009, 2012; Liu et al. 2010, 2011; Fischer-Gödde et al. 2011; Sun et al. 2012; Chen et al. 2014), and southern China (Zhang et al. 2008; Liu et al. 2012a,b, 2013). Additional studies of the Chinese lithosphere can be found in the Chinese literature. Results are also available for xenolith suites from South Korea (Lee and Walker 2006); North Korea (Yang et al. 2010), Taiwan (Wang et al. 2003, 2009), the Tariat depression in the Baikal Rift (Meisel et al. 1996; Pearson et al. 2004; Ionov et al. 2006; Wang et al. 2013), and the Japan–Kamchatka and Papua New Guinea volcanic arcs (McInnes et al. 1999; Kepezinhas et al. 2002; Widom et al. 2003; Saha et al. 2005)
Africa: within the Cameroon Line on the SE side of the Benue Trough (Rehkämper et al. 1997), in the Anza Graben at the intersection of the East African and Gregory Rifts in southern Ethiopia, (Lorand et al. 2003a; Reisberg et al. 2004) and within the Middle Atlas Rift system of Morocco (Wittig et al. 2010)
Antarctica: in Marie Byrd Land area (Handler et al. 2003)
Atlantic Ocean islands: in particular the Canary Islands (Simon et al. 2008).
A BRIEF REVIEW OF HSE AND Os ISOTOPE ANALYTICAL METHODS AND HSE NORMALIZATION VALUES
Analysis of HSE concentrations and Os isotope compositions is technically challenging for several reasons. These include 1) the very low abundances of these elements in the silicate Earth, 2) the possibility of the HSE to be concentrated in highly refractory trace phases that may be randomly and unevenly distributed in the rock, 3) the volatile nature of the most oxidized chemical species, OsO4, preventing use of classical oxidizing acid digestion techniques, and 4) the extremely high first-ioniziation potential of Os, preventing mass spectrometric analysis by thermo-ionization of the ion Os+. The studies included in this review cover five decades, and over that period analytical methods have continuously evolved to meet these challenges, through both improvements in chemical digestion/dissolution and separation techniques and new developments in mass spectrometry. The first whole-rock HSE analytical technique, instrumental neutron activation analysis (INAA) (Crocket et al. 1968), which produced concentration data for only a subset of the HSE, was largely supplanted in the 1990s by inductively coupled plasma ionization mass spectrometry (ICPMS), usually coupled with isotope dilution techniques, permitting acquisition of precise concentration data for all of the HSE. At about the same time, the introduction of the negative thermal ionization mass spectrometry technique (NTIMS; Creaser et al. 1991; Volkening et al. 1991) revolutionized the analysis of Os isotope compositions. These developments in mass spectrometry were accompanied by improvements in sample digestion techniques, such as dissolution in highly oxidizing solutions using Carius tubes (Shirey and Walker 1995) or more recently, high pressure ashers (HPA) (e.g., Ishikawa et al. 2014, see Meisel and Horan 2016, this volume), as well as simplifications of separation and purification procedures for Os and the other HSE. (e.g., Birck et al. 1997). Since 2000, in situ measurements of HSE contents and Os isotopic composition in HSE host minerals have been developed, and are performed respectively by laser-ablation (LA)-ICPMS and LA-multi-collector-(MC)-ICPMS (Pearson et al. 2002; Nowell et al. 2008). Proton probe measurements have also provided information on HSE distribution at the scale of their mineral hosts via elemental mapping and subsequent estimations of concentration (Guo et al. 1999). For a detailed understanding of different chemical and measurement techniques, and a discussion of their advantages and possible limitations, the reader is referred to Meisel and Horan (2016, this volume).
Whole-rock HSE data obtained over the last four decades by widely varying analytical procedures and measurement techniques are included in this present review chapter unless 1) they yielded high procedural blanks, resulting in significant contamination of the HSE concentrations and/or 2) the international and in-house standard geological material of peridotitic composition (e.g., UB-N or JPP-1) reveal analytical issues (low Os concentrations due to volatility-related loss of OsO4, incomplete BMS or PGM digestion, incomplete spike-sample equilibration for isotope dilution analyzes, or large isobaric interferences during mass spectrometric measurement). We stress that the potential limitations of each analytical method must be considered when interpreting the results obtained from the many HSE and Os isotopic studies that have been performed, especially when comparing data from studies that employed different techniques.
Highly siderophile element signatures are conventionally presented as CI-chondrite normalized patterns in the literature. Such concentration data have been obtained for the CI-chondrites Orgueil and Ivuna. The latter systematically show lower HSE contents than Orgueil, which was the CI-chondrite first used for normalization of HSE concentrations. The HSE data obtained since Crocket et al. (1967) are in good agreement considering the small amounts of CI-chondrite available for analysis. Nevertheless, only Anders and Grevesse (1989) and Fischer-Gödde et al. (2011) reported concentrations of all the 8 HSE. The Orgueil HSE concentrations of Anders and Grevesse (1989) are systematically higher for Ir, Pd, and to a lesser extent Pt than the Orgueil analyzes of Fischer-Gödde et al. (2011); the latter are in excellent agreement with other Orgueil HSE determinations (e.g., Horan et al. 2003) (see also Day et al. 2016, this volume for detailed review). Consequently, in this review, we have chosen to normalize the HSE concentrations to the Orgueil average HSE data of Fischer-Gödde et al. (2011) (n = 2 for Os and Re, n = 3 for all the other HSE). The CI-Chondrite (Orgueil) normalising values used here are the following: 463 ng.g−1 Os, 422 ng.g−1 Ir, 629 ng.g−1 Ru, 131 ng.g−1 Rh, 878 ng.g−1 Pt, 572 ng.g−1 Pd, 36.65 ng.g−1 Re, and 175 ng.g−1 Au.
The CI-chondrite normalized HSE patterns of the non-cratonic peridotites presented in this chapter are plotted in the following order: Os, Ir, Ru, Rh, Pt, Pd, Re, and Au. The order from Os to Pd reflects the decreasing temperature of condensation of these elements and follows their increasing incompatibility trend during partial melting of the Earth’s mantle. Rhenium and Au are placed last in the sequence, as they appear to be even less compatible than Pd.
HOST MINERALS OF HIGHLY SIDEROPHILE ELEMENTS IN NON-CRATONIC PERIDOTITE XENOLITHS
Nature of the host minerals
The first hint at the nature of host minerals of HSE in terrestrial rocks came in the early 1970s from HSE analyzes in pyrrhotite (Fe1−xS, with x = 0–0.2) and pentlandite ((Fe,Ni)9S8) from the Sudbury nickel sulfide ores that revealed μg.g−1 level contents of Ir and Pd within these Fe–Cu–Ni sulfides, also known as base metal sulfides (BMS) (Keays and Crocket 1970). For terrestrial mantle rocks, similar HSE host minerals to those recognized in ore deposits were suggested by the high Ir and Au contents obtained from sulfide-rich fractions of a Kilbourne Hole peridotite xenolith (Jagoutz et al. 1979), although these authors did not consider base metal sulfides to play a major role as HSE host phases. The first firm evidence that BMS were the major Ir, Au, and Pd carriers in mantle peridotites was established by Mitchell and Keays (1981), who demonstrated the dominant BMS contribution to the whole-rock budget of these three HSE. Analyzes of BMS in peridotite xenoliths from the Rio Grande Rift (namely the Kilbourne Hole locality: Morgan and Baedecker 1983; Hart and Ravizza 1996; Burton et al. 1999; Harvey et al. 2011) confirmed 1) the μg.g−1-level concentrations of Os and Re in the BMS, which are ~3 orders of magnitude higher than in the whole-rock peridotites and 2) that Fe–Ni–Cu sulfides controlled ~90% of the whole-rock Os mass balance. The major role of BMS in HSE systematics was also recognized over the same period by proton probe analyzes in non-cratonic peridotite xenoliths (Guo et al. 1999) and analyzes of HSE in hand-picked BMS fractions from a Pyrenean orogenic peridotite (Pattou et al. 1996). Pattou et al. (1996) further showed that 1) BMS were the main hosts of gold and 2) that the whole-rock HSE fractionations were perfectly mimicked by those of the BMS. The development of HSE measurements in BMS from non-cratonic xenoliths (and other settings of mantle peridotites) by LA-ICPMS (Alard et al. 2000, 2002, 2011; Lorand and Alard 2001; Pearson et al. 2002) and the ever-growing database on BMS and their respective whole-rock HSE signatures further demonstrate that bulk-rock HSE budgets and signatures are controlled by BMS.
Likewise, HSE measurements of silicates and oxides (spinel) confirm the minor contribution (< 5–20%) of the modally major minerals to the HSE whole rock budget (Mitchell and Keays 1981; Hart and Ravizza 1996; Burton et al. 1999; Handler and Bennett 1999; Harvey et al. 2010, 2011). Rhenium differs from the other HSE as it partitions more significantly into silicates and oxides by a factor of 3–4 (Handler and Bennett 1999) although still behaving as an incompatible element in these main peridotitic minerals (Mallmann and O’Neill 2007). Burton et al. (1999) found that silicate phases could account for about 35% of the Re present in the Kilbourne Hole xenolith that they studied in detail. The somewhat lithophile character of Re is also evidenced by the systematically higher Re/Os ratios of the silicates and oxides compared to the corresponding whole-rock peridotite values (Burton et al. 1999; Harvey et al. 2010, 2011).
In addition to BMS, Mitchel and Keays (1981) and more recently Alard et al. (2011) and Delpech et al. (2012), identified platinum group minerals (PGM) as other HSE carriers in non-cratonic BMS-bearing spinel-peridotite xenoliths (see detailed review by O’Driscoll and González-Jiménez et al. 2016, this volume) using scanning electron microscope (SEM), laser-ablation ICPMS or nuclear microprobe (NMP). In a peridotite xenolith from Mount Porndon (SE Australia), Mitchell and Keays (1981) recognized two Pt- and Pd-rich PGM within a pentlandite-rich BMS grain, whose nature was not unambiguously identified (possibly cooperite PtS and paolovite Pd2Sn). Alard et al. (2011) in a spinel-peridotite from Montferrier (Languedoc, France) and Delpech et al. (2012) in a spinel-peridotite from Kerguelen (Indian Ocean) each recognized an As–Pt ± Pd-rich PGM microphase included in or on the outer rim of pyrrhotite–pentlandite grains. Additionally, Delpech et al. (2012) identified a micrometric Pd–Te–As-rich PGM enclosed in a massive pentlandite grain, a micrometric Au nugget within the matrix and a larger number of Pt–Te–Bi-rich PGM generally associated with the Cu-rich BMS. In other non-cratonic peridotite nodules from SE Australia and the Czech Republic, the occurrence of Pt-bearing PGM was suspected on the basis of the poor whole-rock reproducibility of Pt and Pd, and their decoupling from Ir (Handler and Bennett 1999; Ackerman et al. 2009) but such phases were not firmly identified. All these PGM–BMS associations demonstrate that PGM also play a role in controlling the abundances and fractionations of HSE and Os isotopes at the whole-rock scale in BMS-bearing non-cratonic peridotite xenoliths.
The question regarding which host phases control HSE in non-cratonic BMS-free peridotitic xenoliths remains unanswered, at least from the perspective of indisputable high resolution imaging observation and mineralogical identification. However, Lorand et al. (2004) suggested on the basis of the whole-rock and BMS HSE patterns of harzburgites from Kerguelen island that highly refractory platinum group minerals such as laurite and Os–Ir–Ru-rich alloys stabilize the HSE in these peridotitic residues, similarly to what has been demonstrated for highly depleted orogenic peridotites (Luguet et al. 2007; Lorand et al. 2010).
Petrography of the Base Metal Sulfides
Abundances, distribution and sulfide mineral assemblages
Base metal sulfides are accessory minerals in mantle peridotites, independent of the mode of emplacement of the host rock at the Earth’s surface (i.e., tectonic emplacement of peridotite massifs or volcanic emplacement of peridotite xenoliths), with modal abundances routinely < 0.1 wt%, exceptionally reaching 0.2 wt% as in Montferrier peridotite xenoliths (Mitchell and Keays 1981; Lorand and Conquéré 1983; Szabó and Bodnar 1995; Lee 2002; Liu et al. 2010; Alard et al. 2011; Lorand and Luguet 2016, this volume, Harvey et al. 2016, this volume). Base metal sulfides are heterogeneously distributed at the whole-rock scale, which translates into large modal variations (0–0.2 wt%) between thin sections when several thin sections of a given peridotite are investigated.
In mantle peridotites as a whole, BMS occur as micrometric grains with a large range of dimensions, from <10 μm to a few hundreds of μm in diameter or, in rare cases, to a few mm in length (MacRae 1979; Mitchell and Keays 1981; Lorand and Conquéré 1983; Lorand 1987; Szabó and Bodnar 1995; Lorand and Alard 2001; Lee 2002; Alard et al. 2002, 2011; Lorand et al. 2003a, 2004; Wang et al. 2009; Liu et al. 2010; Lorand and Luguet 2016, this volume; Harvey et al. 2016, this volume) (Fig. 1). If present at the grain boundary between silicates and spinel, they are referred to as interstitial or intergranular BMS. Such BMS show a relatively irregular, sometimes elongated morphology, and can show cuspate to linear rims with the minerals they are in contact with. Alternatively, BMS can occur as inclusions within silicates and melt pockets. BMS inclusions are characterized by either a rounded globule-like or more rarely polyhedral morphology. Non-cratonic spinel peridotite xenoliths typically have both interstitial and enclosed BMS; however, interstitial BMS have been reported to be predominant in the peridotites from Hannuoba (Liu et al. 2010), Kilbourne Hole (Burton et al. 1999), SE and Northern Australia (Alard et al. 2000) and in southern France at the locality of Montferrier (Alard et al. 2002, 2011).
Base metal sulfide may be composed of one to several sulfide phases. In most cases, the sulfide phases described in peridotite xenoliths are made up of S-Fe-Ni and Cu, and consist mainly of monosulfide solid solution (MSS: Fe1−xS, Ni1−xS), pyrrhotite (Fe(1−x)S with x = 0–0.2), pentlandite ((Fe,Ni)9S8) and chalcopyrite (CuFeS2), sometimes associated with minor amounts of mackinawite ((Fe,Ni)1+xS), chalcocite (Cu2S), bornite (Cu5FeS4) or millerite (NiS) (Mitchell and Keays 1981; Lorand and Conquéré 1983; Lee 2002; Lorand et al. 2003a, 2004; Wang et al. 2009; Harvey et al. 2010; Liu et al. 2010; Alard et al. 2011; Delpech et al. 2012; Lorand and Luguet 2016, this volume) (Fig. 1). Furthermore, interstitial BMS grains and BMS enclosed within silicates but connected to the interstitial medium via cracks are pseudomorphosed into magnetite, valleriite (hydrous Fe–Cu sulfide), and iron oxyhydroxides [FeO(OH).5–6H2O] (Luguet and Lorand 1998) (Fig. 1-d,f), sometimes leading to complete pseudomorphic replacement of all BMS grains present in a given sample (Mitchell and Keays 1981; Luguet and Lorand 1998; Lee 2002; Lorand et al. 2003b; Harvey et al. 2010; Alard et al. 2011). These compounds form in response to the extremely oxidizing conditions present during low-temperature alteration. As such, the replacement of the sulfides by iron oxyhydroxides has been attributed to supergene alteration related to percolation of meteoric waters. These iron oxyhydroxides are what is left of the original BMS after the loss of S as mobile sulfates (Lorand 1990; Luguet and Lorand 1998; Lorand et al. 2003b).
Base metal sulfide types and their textural relationships with the peridotite minerals and components
At least four BMS types have been distinguished in the past in non-cratonic spinel-peridotite xenoliths (Lorand and Conquéré 1983; Szabó and Bodnar 1995; Alard et al. 2000, 2002, 2011; Lorand and Alard 2001; Lee 2002; Lorand et al. 2003a, 2004; Delpech et al. 2012), but these distinctions also apply to BMS in all types of mantle peridotites (see Lorand et al. 2013, Lorand and Luguet 2016, this volume; Harvey et al. 2016, this volume). Most importantly, these BMS types are not based solely on specific habits (included vs. intergranular) but integrate the habits, textural relationships of the BMS with the silicates, oxides, volatile-rich minerals, and melt pockets of the host peridotite, and the sulfide assemblage or the bulk sulfide composition of the BMS. The classification presented here is based on the original Type 1 and Type 2 BMS identified by Alard et al. (2000), and has grown and developed due to the increasing knowledge of BMS that has been acquired over the last 15 years.
Type 1 BMS are characterized by a bulk Fe–Ni-rich, S-rich sulfide assemblage dominantly consisting of MSS (or MSS1 and MSS2: two Ni-rich pyrrhotites plotting within the MSS compositional field, see Craig 1973; Lorand and Grégoire 2006), pyrrhotite (Po), pentlandite (Pn) and minor amounts of Cu-rich sulfides, in particular chalcopyrite (Cpy). The pentlandite generally occurs as flames within the MSS or Po and the chalcopyrite is generally confined to the rims of the grains. Type 1 BMS is ubiquitous and can be found as rounded inclusions typically 1) within olivine porphyroclasts in spinel peridotite xenoliths from Hannuoba (Liu et al. 2010; Fig. 1a); North (Mt Quincan) and Southeast Australia (Allyn River, Mt Gambier) and southern France (Montferrier) (Alard et al. 2000, 2002, 2011) and 2) within clinopyroxene in East African porphyroclastic (i.e., deformed) spinel-peridotite xenoliths and spinel-peridotite xenoliths from the Massif Central Province, but exclusively in those from the Southern Massif Central (latitude < 45º 30′) (Alard et al. 2002; Lorand et al. 2003b). Type 1 BMS may be surrounded by decrepitation features (Lorand et al. 2003b; Liu et al. 2010).
Type 2 BMS are characterized by a S-poorer, Cu–Ni-richer sulfide assemblage with greater amounts of Pn and Cpy and smaller amounts of MSS/Po in comparison to Type 1 BMS. Type 2 BMS bulk composition has previously been described as a Cu-rich pentlandite (Alard et al. 2000). In these Type 2 BMS, Cpy may occur as individualized patches widespread within the BMS grain (see Alard et al. 2011). These Type 2 BMS are found as inclusions within metasomatic clinopyroxene showing either a poikilitic texture or spongy rims in the Northern Massif Central location of Montboissier (Lorand and Alard 2001; Alard et al. 2002), Kerguelen (Lorand et al. 2004) and Panshishan in East Central China (Reisberg et al. 2005). Moreover, Type 2 BMS occur within silicate-melt pockets in the peridotites from the Massif Central location of Mont Briançon (Lorand and Alard 2001; Alard et al. 2002; Harvey et al. 2010) and from Sidamo in the East African Rift (Lorand et al. 2003a). More particularly, in the Kerguelen peridotites (Lorand et al. 2004), carbonate-melt pockets host Type 2 BMS grains (Fig. 1b–c), while in the peridotite 1026V from Big Creek in the Sierra Nevada (Lee 2002), the only BMS grain whose sulfide assemblage is not replaced by iron oxyhydroxides appears to be an olivine-enclosed Type 2 BMS. Pentlandite-Cpy ± MSS sulfide assemblages were observed both as inclusions in olivine, orthopyroxene and clinopyroxene and as interstitial grains, sometimes located at triple junctions in Chinese mantle peridotites from Qilin and Hannuoba (Fig. 1d) (Guo et al. 1999, Liu et al. 2010). Further occurrences of intergranular Type 2 BMS were described 1) at triple junctions of the silicate matrix in Kerguelen peridotites (Lorand et al. 2004), in Massif Central peridotites especially in the sulfide-poor lherzolites from Montferrier (Alard et al. 2002) and in peridotites from Panshishan in eastern China (Reisberg et al. 2005), and 2) in close association with clinopyroxene in the East African Rift (Lorand et al. 2004), SE Australia (Alard et al. 2002) and Massif Central peridotites (Lorand and Alard 2001; Alard et al. 2002).
Type 3 and Type 4 BMS are similar in composition to Type 1 BMS, ranging from almost exclusively MSS/Po (95%) to 50% Po with Pn (up to 45%) intergrown as flame-like structures and with very small volumes of Cu-rich sulfides (max 6 vol. %), typically Cpy, confined to the rims of the Po–Pn grains (Lorand et al. 2004, see Fig. 1e). However, the resemblance to Type 1 BMS stops here as Types 3 and 4 occur as intergranular components or as inclusions within metasomatic minerals. Intergranular Type 3 BMS were described in granular peridotites from the East African Rift (Lorand et al. 2003a) and in protogranular and poikilitic harzburgites from Kerguelen (Lorand et al. 2004; Delpech et al. 2012), where they occur in close association with metasomatic clinopyroxene and glass pockets, and more rarely at opx–ol–ol junctions. In these peridotites, Type 3 BMS also appear as inclusions in metasomatic clinopyroxene or as a connected network of veinlets outlining silicate grain boundaries; these veinlets sometimes intrude the silicates following fracture planes. Type 4 BMS have been so far exclusively described in Kerguelen Island dunite xenoliths (Lorand et al. 2004; Delpech et al. 2012) and in the Montferrier xenoliths (Alard et al. 2011). The Type 4 BMS are exclusively intergranular phases sometimes associated with polycrystalline dolomite grains (Fig. 1e) but never with a melt pocket. According to Delpech et al. (2012), the morphology of a BMS grain depends on the geometry of the intergranular space it occupies; for the Type 4 BMS, the rims curve inward but never exhibit dihedral angles < 60º (Fig. 1f). In addition to their different bulk sulfide composition, the high dihedral angles of these Type 4 BMS contrast strongly with those of the intergranular Type 2 BMS from the same xenoliths, which show generally cuspate edges forming low dihedral angles with the surrounding minerals.
Although information on the bulk composition of the BMS was not provided, Burton et al. (1999) and Harvey et al., (2010) described rounded BMS grains found within silicate melt pockets or small BMS inclusions cogenetic with CO2-fluid inclusions within major silicate phases such as olivine crystals in the Kilbourne Hole and the Mont Briançon peridotites. Burton et al. (1999) additionally observed interstitial BMS located at the triple junctions between silicates. These textural features could highlight the presence of Type 2, -3 and/or -4 BMS in these peridotite xenoliths.
Origin of the Base Metal Sulfides and Platinum Group Minerals
Understanding the origin of BMS is crucial for correct interpretation of the HSE systematics as well as the Os-isotopic signatures at both the bulk peridotite scale and the mineral scale. A realistic assessment must integrate the mineralogical assemblage and the textural relationships with the silicate, oxide, and volatile-rich minerals, as well as the specific origin of these latter.
Type 1 BMS, which are included within silicate porphyroclasts (i.e., olivine), showing typical refractory composition (high Mg numbers) and possible deformation features (i.e., kink bands) are interpreted as phases residual to partial melting (Alard et al. 2002). Their Fe–Ni-rich sulfide assemblages, reflected in high MSS (or Po) and Pn vs. Cpy proportions, result from the fractional crystallization of a high temperature sulfide liquid and its complex sub-solidus re-equilibration history. These mineralogical evolution paths have been described on the basis of experimental investigations (Craig and Kullerud 1969; Kullerud et al. 1969; Misra and Fleet 1973; Dutrizac 1976; Fleet and Pan 1994; Peregoedova et al. 2002) or detailed studies of BMS from non-cratonic spinel mantle peridotite xenoliths as well as peridotites from massifs (Lorand and Conquéré 1983; Lorand 1987; Szabó and Bodnar 1995; Guo et al. 1999). Readers are referred to these publications for an in-depth review. Briefly, from the high-temperature mantle sulfide liquid, a monosulfide solid solution compositionally close to FeS crystallizes first at 1192 °C, and coexists with a Ni- and Cu-rich sulfide liquid. As temperatures decrease down to ca. 1000 °C, the MSS field extends toward compositions richer in Ni and Cu. Reaction at 860–900 °C between the MSS and the Cu–Ni-rich liquid produces an Intermediate Solid Solution (ISS) and a high temperature Heazlewoodite Solid Solution (HzSS: (Ni,Fe)3S2). Chalcopyrite crystallizes from the ISS below 557 °C, while pentlandite forms at ca. 610 °C as a product of the reaction between HzSS and MSS. Below 300 °C, Ni-poor MSS1, Ni-rich MSS2 (>30 wt% Ni), pyrrhotite, and pentlandite exsolve at the expense of the MSS.
Although having a Fe–Ni-rich sulfide assemblage similar to that of the Type 1 BMS inclusions described above, Types 3 and 4 BMS have a more complex origin. Though they differ from one another, both types are metasomatic in nature as highlighted by their peculiar spatial association with secondary silicates or volatile-rich components. Lorand et al. (2004) suggested that Type 3 BMS, which occur exclusively as interstitial grains associated with metasomatic clinopyroxene, represent former Type 1 BMS (inclusions) that were remobilized by pervasive melt percolation and silicate recrystallization. As such, this process would erase all trace of former Type 1 BMS inclusions, remobilizing and disguising them as intergranular Type 3 BMS. Within an individual suite (e.g., Sidamo or Kerguelen), the remobilization of the BMS is likely coupled with a partial dissolution of the BMS, implying reaction with a S-undersaturated silicate melt, since the peridotites showing extensive silicate recrystallization (poikilitic textures) have lower BMS modal abundances (Lorand et al. 2003a, 2004) for similar fertility indices. On the other hand, Type 4 BMS were interpreted as sulfidation products due to the reaction of a S-rich CO2 vapor phase with the magnesian silicates (Lorand et al. 2004; Delpech et al. 2012). This interpretation is based on the systematic association of these BMS with CO2-fluid inclusion trails and with dolomite grains as well as the absence of liquid–liquid immiscibility features, in contrast to what is observed in Type 2 BMS from the same locality, which testifies to precipitation of the Type 4 BMS as solid phases (Lorand et al. 2004).
The S-poorer, Cu-richer Type 2 BMS result from the crystallization of highly evolved sulfide melts, similar in composition to the residual sulfide liquid left after the crystallization of the MSS from the high-temperature sulfide melt, which is assumed to be the precursor of mantle BMS. According to Lorand and Alard (2001), these melts would also likely resemble the first sulfide melt generated by the incongruent partial melting of the mantle BMS, which would leave behind in the peridotitic residue MSS-like residual phases (i.e., residual Type 1 BMS). The metasomatic origin of the Type 2 BMS found as either inclusions or intergranular components is supported by their systematic association with metasomatic clinopyroxene, sometimes displaying poikilitic textures, as well as their close contact with silicate- and carbonate-glass pockets. The dominant ovoid or rounded morphology of these Type 2 BMS (especially in the melt pockets) indicates a clear immiscibility of the sulfide melts within the silicate or carbonatitic melt. These S-saturated percolating silicate or carbonatitic melts may have been either low degree partial melts or evolved melts of higher partial melting degrees, whose original composition could have been modified through fractional crystallization and melt–rock reactions. As such, the repeated association between metasomatic clinopyroxene and Type 2 BMS highlights a genetic link. In fact, the melt–rock reactions forming clinopyroxene at the expense of orthopyroxene are generally considered to be melt-consuming (e.g., Bedini et al. 1997), triggering a “chain reaction” of sulfur saturation, sulfide melt exsolution products and clinopyroxene–BMS coprecipitation.
Of course, considering the reaction rims present at the xenolith-host lava contact, one can wonder if Type 2, Type 3 and Type 4 BMS could be due to lava contamination. This can however be ruled out on numerous lines of evidence (see Harvey et al. 2016, this volume). Mainly, Burton et al. (1999), Alard et al. (2002) and Harvey et al. (2010) highlighted the extremely Fe–S-rich (60 wt% Fe) but Ni–Cu-free composition of the sulfides, close to pure monoclinic pyrrhotite, present in alkali basalts from Kilbourne Hole, SE Australia and the Massif Central. This contrasts with the S-poorer and Cu and Ni-richer compositions of Type 2, Type 3 and Type 4 BMS. Besides, lava-derived sulfides occur exclusively as veinlets, while Types 2, 3 and 4 BMS are found as both inclusions and intergranular components but only occasionally as veinlets.
The Pt-tellurides, so far recognized in the Montferrier and the Kerguelen peridotite xenoliths (Alard et al. 2011, Delpech et al. 2012), are likely products of the fractional crystallization of a Cu-Ni-rich sulfide melt and thus have the same origin as the Type 2 BMS in which they are enclosed. The MSS/Cu–Ni sulfide melt partition coefficients determined experimentally suggest that Pt, Pd, Se, Te, Bi will preferentially partition into the Cu–Ni-rich sulfide melt (Helmy et al. 2007, 2010). For high Pt–Pd concentrations in the Cu–Ni-rich sulfide melt, Helmy et al. (2007) suggested the possible immiscibility of a telluride melt and a sulfide melt. Nevertheless, owing to the more likely low Te concentrations, Lorand et al. (2010) interpreted the Pt-(Bi) tellurides from the Lherz orogenic peridotites as late-stage exsolution from the Cu–Ni-rich sulfide melts. The reaction of these Pt-tellurides with As-bearing vapor could be at the origin of the PtAs2 nuggets (Delpech et al. 2012) also observed in Montferrier, Kerguelen and Mount Porndon (Alard et al. 2011; Delpech et al. 2012; Mitchell and Keays 1981, respectively).
HIGHLY SIDEROPHILE ELEMENTS AND 187Os/188Os RESULTS FROM NON-CRATONIC PERIDOTITE XENOLITHS
Our knowledge of HSE and Os-isotopic behavior in mantle peridotites comes both from whole-rock studies and from analyzes of individual mineral phases, especially BMS. Base metal sulfide data have been obtained both in situ, by LA-(MC)ICPMS and proton probe, and by extraction of individual BMS grains followed by chemical separation and analysis of Os isotopes and HSE abundances. Bulk analyzes of mineral separates of silicate phases and spinel have placed constraints on partitioning of the HSE between the various non-sulfide phases that dominate the modal composition of the rock. Whole-rock and BMS analyzes provide complementary information about HSE behavior in the mantle that needs to be reconciled.
Summary of results from whole-rock studies
In order to obtain a global view of HSE whole-rock systematics, we have compiled a database of 187Os/188Os ratios and HSE concentrations based on ~60 studies of peridotite xenoliths from alkali basalts erupted in widely varying tectonic and geographic locations (electronic supplement available upon request to the authors). This database includes about 980 analyzes of Os-isotopic compositions and Os and Re concentrations, as well as about 450 concentration analyzes of other HSE, with the exception of Rh and Au, two mono-isotopic elements that cannot be analyzed by isotope dilution, for which only ~100 analyzes are available. Whole-rock total HSE concentrations range from 1 pg.g−1 to 20 ng.g−1, more exceptionally extending to 20–40 ng.g−1 (Table 1). Apart from Re, the scatter in concentrations is more pronounced (by one extra order of magnitude) for highly depleted mantle residues (i.e., those with Al2O3 contents <~1.5% in Fig. 2). When compared to tectonically emplaced peridotites, about half of the peridotitic xenoliths overlap in terms of HSE concentrations, while the other ~50% show considerable scatter, mostly to lower values but sometimes to higher values than the PUM concentration estimates of Becker et al. (2006) or the narrow concentration interval defined by orogenic, ophiolitic, and abyssal peridotites, which is here based on only three studies comprising ~90 analyzes. However, as these include 12 different localities this database is nevertheless likely to be representative (see also the comprehensive compilation of orogenic, ophiolite and abyssal peridotites in Becker and Dale (2015), this volume). Average HSE concentrations from the two types of peridotite settings are compared in Table 1. As this comparison shows, HSE contents of basalt-borne xenoliths are considerably lower than those of massif peridotites, though inter-element ratios are fairly similar in most cases. Large variations are observed for the most incompatible HSE, namely palladium and rhenium whose concentrations vary over 3 orders of magnitude (Fig. 2). PdN/IrN ratios (0.85 ± 0.96) are generally lower and more variable than those of massif peridotites (1.15 ± 0.33) but do not show the extreme PdN/IrN fractionation observed in cratonic peridotites (0.44 ± 0.47; Pearson et al. 2004). Although also an incompatible HSE, the behavior of Pt is more complex, with Pt contents and particularly PtN/IrN ratios showing very large variations among the xenoliths. Strikingly, Os, which behaves compatibly, is as variable as Pd and Re in terms of the range of concentrations displayed. As has been noted in previous studies, Os concentrations of alkali-basalt-hosted xenoliths (2.3 ± 1.6 ng.g−1) are in most cases substantially lower than those of massif peridotites (4.1 ± 1.6 ng.g−1) (Fig. 2). However the mean OsN/IrN ratio of all of the basalt-hosted xenoliths (0.89 ± 1.39) is only slightly lower than the mean massif peridotite value (1.09 ± 0.32), suggesting that the very low OsN/IrN ratios reported for some xenolith suites and often ascribed to late-stage Os loss during and after eruption (Handler et al. 1999; Pearson et al. 2004) are not ubiquitous. Osmium contents of non-cratonic xenoliths are also lower and less variable than those of cratonic peridotite xenoliths (4.2 ± 2.7 ng.g−1; Pearson et al. 2004; Aulbach et al. 2016, this volume), which are generally brought to the Earth’s surface via kimberlitic or lamproitic volcanism.
Positive binary HSE correlations are routinely observed between Ir, Ru, and Rh in individual xenolith localities as well as at the scale of the full database of non-cratonic peridotite xenoliths (Fig. 3). The binary trends possibly highlight slightly supra-chondritic RhN/IrN and RuN/IrN ratios (Table 1) (Pearson et al. 2004; Becker et al. 2006). Only broad positive Os vs. Ir variations (Fig. 3) exist (e.g., Becker et al. 2006; Liu et al. 2010; Fischer-Gödde et al. 2011; for Hannuoba xenoliths), which contrast with the much stronger correlations observed among tectonically emplaced peridotites. Correlations between the IPGE (Iridium Group PGE: Os, Ir, Ru; Barnes et al. 1985) and Rh; and Pt and Pd are far more scattered (Vitim: Pearson et al. 2004; Yangyuan xenoliths: Liu et al. 2010, 2011; Rio Grande Rift xenoliths: Becker et al. 2006; Harvey et al. 2011) or even non-existent (Sidamo: Lorand et al. 2003a; Atlas: Wittig et al. 2010; SE Australia: Handler and Bennett 1999; Eifel: Fischer-Gödde et al. 2011) (Fig. 3). These observations support the strongly coupled behavior of Ru–Rh–Ir, and decoupling or less coherent behavior between these elements and Pt and Pd as well as with Re and Os. Coherent behavior of Ir–Ru–Rh implies that they are hosted in a single phase, i.e., BMS. The decoupling between Pt and Pd from Ir and Ru argues for the existence of several minerals (BMS and PGM) possibly controlling Pt and Pd and/or different generations of BMS/PGM in terms of origin (residual vs. metasomatic) and HSE composition (e.g., different PdN/IrN). Such interpretations are supported by the observations of Pd–Pt-rich PGM in the specific xenolith suites of SE Australia, Montferrier and Kerguelen (Mitchell and Keays 1981; Alard et al. 2011; Delpech et al. 2012) as well as the poor reproducibility of duplicate analyzes for Pt, Pd, PtN/IrN and PdN/IrN (see Lorand and Alard 2001; Lee 2002; Lorand et al. 2004; Ackerman et al. 2009; Alard et al. 2011). As will be discussed below, the incoherent behavior of Re relative to the other HSE may also be related to the fact that a significant fraction of this element is hosted by silicate and oxide phases, while the erratic behavior of Os may reflect mobility under oxidizing conditions.
Correlations of HSE concentrations with indices of melt extraction (e.g., Al2O3, Yb, Lu, CaO) are quite rare and can be quite variable for a given HSE or within a group of HSE (e.g., within IPGE). Negative trends with Al2O3 are observed for Os in the Kozakov xenoliths (Czech republic: Ackerman et al. 2009), but are absent for the other compatible IPGE (Ru and Ir) or for Pd. In contrast, Pd concentrations broadly decrease with the progressive depletion of the mantle peridotite xenoliths from the Newer volcanic field in SE Australia (Handler and Bennett 1999), in the Sidamo peridotite xenoliths (Lorand et al. 2003a), in the Hannuoba xenoliths (Becker et al. 2006), as does Re (and the Re/Os ratios) in the mantle peridotite xenoliths from the Newer volcanic field in SE Australia (Handler and Bennett 1999). Gold, for which fewer data are available than for the other HSE, shows broadly increasing contents as extent of melt extraction decreases (Fischer-Gödde et al. 2011). Finally, HSE abundances or their relative fractionation may follow S or Se contents or S/Se ratios—as seen in the Massif Central xenoliths (Lorand and Alard 2001; Alard et al. 2011)—although such behavior is far from systematic (Ackerman et al. 2009), possibly partly due to sensitivity of S and BMS to syn- and/or post-eruption disturbance.
The great majority of 187Os/188Os whole-rock ratios of peridotite xenoliths from non-cratonic settings fall in the range 0.11–0.13. Osmium isotopic compositions more radiogenic than those of chondritic meteorites (including 187Os/188Os values not only for carbonaceous chondrites but also for enstatite and ordinary chondrites, which range up to 0.1305: Walker et al. 2002a; Day et al. 2016, this volume) are quite rare but are occasionally observed, for example in the Montferrier locality of southern France (Alard et al. 2011). Conversely, 187Os/188Os less than 0.11, indicative of Archean melt extraction episodes, are also nearly absent among whole rock analyzes of alkali-basalt-borne peridotite nodules. Correlations between 187Os/188Os and 187Re/188Os are usually weak and do not define even approximate isochrons in most cases, with the Hannuoba xenoliths (Gao et al. 2002) being a notable exception. On the other hand, it has long been noted that whole-rock 187Os/188Os ratios of non-cratonic peridotite xenoliths (Meisel et al. 1996, 2001), as well as of massif peridotites (Reisberg and Lorand 1995) are often well correlated with indices of degree of melt extraction such as bulk-rock Al2O3, Yb, or Lu contents, modal clinopyroxene content, olivine Mg# ([Mg2+/(Mg2++Fe2+)] on a molar basis) or spinel Cr# ([Cr3+/(Cr3++Al3+)] on a molar basis). In fact, correlations between 187Os/188Os and Al2O3 have acquired the nickname of “aluminachrons” as one possible interpretation of these trends is that they represent isochron analogs, developed by radiogenic ingrowth since a partial-melting event that left peridotites variably depleted in Re. This and other hypotheses for the origin of such correlations will be discussed below. Several typical examples from our database of correlations between 187Os/188Os and Al2O3, ranging in quality from very strong to nearly non-existent, are shown in Figure 4.
Trends between whole rock 187Os/188Os and melt depletion indices are quite common in non-cratonic peridotite suites from a wide variety of tectonic contexts, and have even been found in oceanic settings, such as the Salt Lake Crater in Hawaii (Bizimis et al. 2007). Nevertheless, these trends are sometimes perturbed towards more radiogenic values, particularly among harzburgitic samples (Peslier et al. 2000b). This has been attributed to metasomatism and melt percolation processes. Certain suites show no trends at all, such as the Mont Briançon samples of the French Massif Central (Harvey et al. 2010), an observation that the authors ascribed to multiple episodes of partial melting and melt percolation. In other cases, such as those of the xenolith suite from the Rangrim massif of North Korea, the absence of such trends may be related to lithospheric removal or delamination, leading to the addition of juvenile material from the convecting mantle (Yang et al. 2010). Finally, we note that cratonic peridotite xenoliths carried by kimberlite pipes, unlike the majority of the non-cratonic samples considered here, do not typically show correlations between Os-isotopic composition and melt extraction indices (e.g., Carlson et al. 1999; Meisel et al. 2001; Pearson et al. 2002, 2004; Aulbach et al. 2016, this volume), displaying instead a wide range of 187Os/188Os coupled with a limited spread in Al2O3 values.
Summary of results from Base Metal Sulfides and other Mineral Phases
Most of our knowledge of the HSE composition of the BMS and PGM hosts in non-cratonic spinel-bearing peridotite xenoliths has been obtained via laser-ablation analyzes (Alard et al. 2000, 2002, 2011; Lorand and Alard 2001; Pearson et al. 2002; Delpech et al. 2012) and proton probe investigations (Guo et al. 1999). Handpicked BMS and aggregates of silicate and oxide material have also been processed for Os isotope analysis by thermal ionization mass spectrometry (Burton et al. 1999; Harvey et al. 2010, 2011). Advantages and disadvantages of the various analytical methods for individual BMS grains are discussed in detail in Harvey et al. (2016), this volume. Most of these studies that investigated Re–Os-isotopic systematics at the BMS scale do not provide mineralogical and textural information for the analyzed BMS, and therefore cannot be considered for the following discussion.
HSE compositions of the various Base Metal Sulfide phases
Base metal sulfides contain several hundred ng.g−1 to several hundred μg.g−1 of HSE, with concentrations 50–100,000 times higher than those obtained in the respective whole-rock peridotites, demonstrating that although BMS are accessory minerals in mantle peridotites, they host more than 80% of the HSE (Mitchell and Keays 1981; Hart and Ravizza 1996; Burton et al. 1999; Guo et al. 1999; Alard et al. 2000, 2002, 2011; Lorand and Alard 2001; Pearson et al. 2002; Harvey et al. 2010, 2011; Griffin et al. 2012; González-Jiménez et al. 2013).
Because of their relatively small size and their complex mineralogical microstructures due to the exsolution and intricate intergrowth of sulfide phases during subsolidus re-equilibration (e.g., MSS + Pn), as well as the extensive replacement of the BMS by iron oxyhydroxides during supergene alteration, pure sulfide phases (e.g., Pn, Cpy) have rarely been analyzed. To date pure analyzes of only four MSS (Type 1 BMS) and one isocubanite (Icb) have been obtained, all exclusively in the BMS of the Massif Central spinel-peridotite xenoliths (Alard et al. 2000; Lorand and Alard 2001). Any other analyzes of HSE compositions at the BMS scale most likely represent mixtures reflected in the bulk composition of the BMS. Despite this, either considering the purely mineralogical aspect (MSS vs. Icb vs. Pn) or the bulk composition (Type 1 vs. Type -2 BMS), BMS show a large range of fractionation between the most compatible HSE (Os, Ir, Ru, Rh) and the more incompatible HSE (Pt, Pd, Re, and Au) but also within the compatible or incompatible HSE groups (e.g., Pt vs. Pd). Note that the terms “compatible” and “incompatible” as used here refer to the global behavior of these elements during partial melting, and are not intended to describe the affinity of any of these elements for any specific phase, these aspects of HSE behavior being discussed more fully in Brenan et al. (2016, this volume).
Monosulfide solid solutions are generally characterized by high HSE concentrations (e.g., 10–100 μg.g−1 Os, 5–45 μg.g−1 Pd, see Lorand and Alard 2001). They commonly display smooth but convex-upwards CI-chondrite normalized HSE patterns characterized by a slightly positive Os–Ru segment and a progressive and gentle decrease from Rh or Pt to Pd. When hosted by xenoliths, MSS yield systematically sub-chondritic PdN/IrN (Fig. 5). Such patterns are very similar to those of MSS from other environments such as inclusions within megacrysts or diamonds (Bulanova et al. 1996; Alard et al. 2000; Aulbach et al. 2016, this volume). Although Au is depleted compared to the compatible HSE, it shows variable abundances when compared to Pd, its closest HSE in terms of incompatible geochemical behavior, and it also varies independently of any other HSE. While the pure MSS patterns are all remarkably similar (note that 3 MSS patterns are overlapping), their large range in HSE concentrations may relate to the MSS composition, more specifically the metal/sulfur ratio (Lorand and Alard 2001) and be controlled by the availability of vacancies in the sub-metal lattice (Mackovicky et al. 1986; Barnes et al. 1997).
Pure isocubanite, a Cu-rich sulfide (Lorand and Alard 2001), has strongly contrasting abundances and CI-chondrite normalized HSE patterns compared to MSS and Pn (Fig. 5). It shows extremely low concentrations of the compatible HSE (Os–Ir–Ru), in the range of a few tens to hundreds of ng.g−1 (e.g., 50 ng.g−1 Os). These abundances are thus at least 500 times lower than in the HSE-poorest pure MSS. The Icb HSE pattern displays an overall steep positive slope, due to the significant enrichment of Pt, Pd, and Au over the compatible HSE. Importantly, the Pd contents of this Cu-rich sulfide overlap those of the MSS.
Intergrowths of MSS and Cu-rich sulfides yield HSE compositions intermediate to both the Icb and the MSS HSE composition (see MBR11-S1a,b of Lorand and Alard 2001; Fig. 5). The MSS + Cu-rich sulfide grains display low Os–Rh abundances, with a CI-normalized flat segment, closely mimicking that of the MSS although at much lower HSE concentrations. The HSE-patterns of such composite BMS also display steep positive slopes from Rh to Pd, which likely reflect the high Pd and Pt contents of the Cu-rich sulfides in comparison to the MSS.
Pure pentlandite has never been analyzed in non-cratonic peridotite xenoliths, but HSE compositions for MSS + Pn (Lorand and Alard 2001) and Pn + Cpy (Alard et al. 2000) mixtures are routinely determined (Fig. 5). MSS + Pn mixtures (likely representing Type 1 BMS) typically display a similar HSE concentration range to that of the MSS for the compatible HSE (Os to Rh) but their CI-chondrite normalized patterns tend to be flatter. Conversely the compatible HSE in the Pn + Cpy mixtures (most likely Type 2 BMS) yield much lower concentrations and more fractionated inter-element ratios (OsN/IrN =1.8–58.5 and PdN/IrN = 3.4–1143). The variability of the absolute concentrations and fractionations likely reflects the relative proportions of Pn and Cpy. Despite this, both the MSS + Pn and Pn–Cpy mixtures are characterized by systematically negative Pt and Au anomalies compared to their respective neighbors Rh and Pd, and Pd and Re. The negative Pt anomalies in the Pn-bearing sulfide mixtures are generally interpreted as reflecting the rejection of Pt from the octahedral sites of the pentlandite, while Ru, Rh, and Pd are accommodated in these crystallographic sites (Mackovicky et al. 1986; Czamanske et al. 1992). Platinum is in fact present within pentlandite as Pt-rich nuggets (Alard et al. 2011), which exsolved upon cooling and sub-solidus re-equilibration (Lorand and Alard 2001). These nuggets are sometimes revealed either directly by high-resolution proton probe imaging or indirectly by spikes in the signal intensity during LA-ICPMS analysis. The Pn-bearing bimodal BMS show a large range of Pd concentrations spanning 1.8–35 μg.g−1, regardless of the second sulfide present in the mixture. However, PdN/IrN is systematically sub-chondritic in the MSS + Pn mixtures (0.13–0.84) and supra-chondritic in the Pn + Cpy mixtures (1.4–17). Rhenium concentrations in the two available analyzes differ substantially, but Re appears to vary with Pd, meaning that Pd-poor Pn are most likely to show a concomitant Re depletion and vice versa.
Os-isotope compositions of the various Base Metal Sulfide and silicate phases
As the preceding discussion shows, BMS are by far the dominant host for Os in peridotites, and as such have relatively high Os concentrations, typically in the range 10–100 μg.g−1 (e.g., González-Jiménez et al. 2013). Thus in addition to BMS mineral separates (Hart and Ravizza 1996) these high Os concentrations make it possible to analyze individual BMS grains, either by LA-ICPMS (Alard et al. 2002; Pearson et al. 2002) or by physical separation of the BMS grains followed by chemical separation of Re and Os and mass spectrometric analysis (Burton et al. 1999; Harvey et al. 2006, 2010, 2011, 2016, this volume). In the case of laser ablation analyzes, Os-isotopic compositions can be directly linked to the petrographic context of the BMS grains as well as their mineralogical composition (e.g., Type 1 and Type 2 BMS). Though more difficult, this can also be done for bulk BMS analyzes, by carefully extracting the grains of interest from a thin section or by identifying the likely grain location (intergranular or included, although both Type 1 and metasomatic Type 2 BMS can be found as inclusions) on the basis of grain morphology and determining the concentrations of Fe, Ni, and Cu in order to estimate the mineralogical assemblages (e.g., Pn–Cpy–Po). The in situ 187Os/188Os dataset of BMS from non-cratonic peridotite xenoliths was mostly acquired on Type 1 BMS (exclusively occurring as inclusions) and Type 2 BMS, which are dominantly found in intergranular positions although Type 2 BMS enclosed in clinopyroxene have also been analyzed (see Alard et al. 2002).
Analyzes of individual BMS by either technique have demonstrated the extreme heterogeneity in Os-isotopic composition that can be found among BMS grains from the same xenolith or even from the same thin section. For example, Harvey et al. (2011) found 187Os/188Os ratios ranging over 0.124–0.373 in BMS extracted from a single peridotite xenolith from Kilbourne Hole, New Mexico. Since the earliest studies, which combined Re and Os concentration measurements with simultaneous determinations of Os-isotope compositions (Alard et al. 2002), it has been recognized that globally, Type 2 BMS (both included or intergranular) have higher 187Os/188Os ratios, higher Re concentrations, and lower Os concentrations than BMS included in residual phases, in particular olivines. This likely results from the higher modal proportions of Os-poor, but Re-rich, Cpy at the BMS grain scale in Type 2 BMS compared to Type 1 BMS (Alard et al. 2002). There are, however, many exceptions to the “Type 2 BMS = more radiogenic” rule, and wide variations of Os-isotopic composition are observed in BMS from both Type 1 and Type 2 settings in a given rock (Alard et al. 2000, 2002, 2011; Wang et al. 2009). This extreme isotopic heterogeneity, and the tendency of Type 2 BMS to have higher 187Os/188Os ratios, is usually attributed to the introduction of metasomatic BMS with radiogenic Os compositions and / or high Re contents into peridotites that have experienced ancient melt extraction events (Alard et al. 2002). Though there is a general tendency for these Type 2 BMS with elevated 187Os/188Os to have higher 187Re/188Os ratios, there is no systematic relationship between 187Re/188Os and 187Os/188Os and negative correlations have even been reported (Alard et al. 2002). The positive and negative 187Os/188Os vs. 187Re/188Os variations have been ascribed to the recent addition of metasomatic BMS components of different origins (melts vs. fluids) (Wang et al. 2009 ; Harvey et al. 2010).
Though BMS largely dominate the Os budget in spinel-peridotite xenoliths, the contribution of the modally major phases (olivine, orthopyroxene, clinopyroxene, and spinel) has also been examined (Hart and Ravizza 1996; Burton et al. 1999; Harvey et al. 2010, 2011). Because of the 4–5 order-of-magnitude lower Os concentrations in these phases relative to those of mantle BMS, even minute amounts of included BMS could significantly bias the measured results on silicates and spinels. Nevertheless, systematic differences in Os and Re concentrations among the silicate phases and spinel have been found in carefully prepared mineral separates, suggesting that this problem has been avoided. The studies cited above are consistent with generally increasing contents of Os and Re from olivine to orthopyroxene to clinopyroxene to spinel (Harvey et al. 2010, 2011). While mass-balance calculations based on these analyzes confirm that only a few percent of the Os in the whole-rock can be accommodated by these modally major phases, they may, however, represent a significant reservoir for Re, representing for example about 35% of the total Re budget in xenoliths studied by Burton et al. (1999) and Handler and Bennett (1999). These studies also show that the 187Os/188Os and 187Re/188Os ratios of silicate phases and spinels from non-cratonic mantle xenoliths are nearly always higher than those of corresponding whole rocks. In their investigation of a Kilbourne hole xenolith, Burton et al. (1999) found that the non-sulfide phases, as well as the interstitial BMS, were in isotopic equilibrium, but were, however, more radiogenic than the enclosed BMS. In contrast, isotopic equilibrium between the silicate phases and spinel was not observed in other Kilbourne hole xenoliths (Harvey et al. 2011), or in peridotite xenoliths from the Massif Central (Harvey et al. 2010).
Reconciling 187Os/188Os results from whole-rock and Base Metal Sulfide analyzes
At first glance, the Os-isotopic results from whole rocks and BMS summarized above seem to provide conflicting messages. Whole rock 187Os/188Os ratios are limited to a fairly restricted range, only rarely exceeding chondritic values. In the majority of cases, these ratios are correlated with indices of degree of melt extraction. Among the samples that do not fall on the correlation trends, most plot towards more radiogenic values (Fig. 4). In contrast, individual BMS analyzes portray a much more chaotic picture. Though there is a general tendency for interstitial BMS (exclusively Type 2 BMS here) to have more radiogenic compositions than enclosed BMS (most dominantly Type 1 BMS), a very large range of 187Os/188Os, extending to highly suprachondritic values, can be found in BMS from a single xenolith (Alard et al. 2002; Wang et al. 2009).
A first point to emphasize is that the existence of Os-isotopic heterogeneity among the BMS from a single xenolith does not automatically imply introduction of a metasomatic component. As explained by Burton et al. (1999), Os diffusion in peridotite xenoliths is likely to be too sluggish to permit full isotopic equilibration between all of the BMS and silicate phases in the rock. This is not necessarily because of low diffusion coefficients (which are unknown for Os in silicate phases), but because the extremely low Os concentrations of silicate phases impose very low Os concentration gradients, incapable of driving rapid diffusion. This concept is clearly explained by Hofmann and Hart (1978) for the analogous case of Sr diffusion across an olivine barrier. Because of very slow diffusion, BMS enclosed in silicate crystals are at least partially shielded from Os-isotopic re-equilibration with the rest of the rock. As Type 2 BMS generally have higher 187Re/188Os than Type 1 BMS, radiogenic ingrowth alone could produce substantial 187Os/188Os heterogeneity among the different BMS from a given xenolith. In the simplest case of a single partial melting episode followed by long term residence in a cool lithosphere, without melt percolation or metasomatism, the whole-rock 187Os/188Os ratios of peridotite xenoliths from a given suite are expected to vary systematically with the extent of melting experienced by each xenolith. Within each of the xenoliths, the enclosed BMS, dominantly Type 1 BMS, are expected to have low Re/Os and thus will develop very little radiogenic Os over time. As a result, enclosed BMS are expected to have low 187Os/188Os ratios corresponding to Re-depletion ages (TRD) approaching the age of the partial melting event (see “Obtaining age information from Os isotopes of whole rocks” as well as Walker et al. 1989 and Harvey et al. 2016, this volume, for an explanation of TRD ages). In contrast, the interstitial BMS (most likely Type 2 BMS) characterized by higher Re/Os should yield more radiogenic and likely variable 187Os/188Os corresponding to TRD ages that cannot be simply interpreted.
Several processes may complicate this simple scenario. While Os-isotopic differences between and among intergranular (Type 2 BMS) and enclosed (here Type 1) BMS do not require the addition of metasomatic BMS, this process is nevertheless thought to be common. In this case, whole rock 187Os/188Os ratios will represent mixed signatures, and therefore their value for constraining the timing of melt extraction may be compromised. Nevertheless, enclosed Type 1 BMS that have not been affected by metasomatism could still provide information about the age of the melting event. As the Type 2, mostly intergranular, BMS tend to have lower Os contents than the Type 1 BMS, their contribution to the whole-rock Os-isotopic budget is assumed in many cases to be quite minor (Harvey et al. 2011). These two simple assumptions (i.e., that the intergranular BMS have radiogenic Os compositions but low Os concentrations), have however, been challenged in cratonic peridotites where the oldest TRD ages and most unradiogenic 187Os/188Os have been reported for intergranular BMS, associated with kimberlite-related metasomatic silicates, such as phlogopite, or secondary clinopyroxene (Wainwright et al. 2015). Alternatively, heating and/or deformation without the addition of a metasomatic component could lead to partial internal Os-isotopic re-equilibration between the BMS within a given rock. In this case, the individual BMS model ages may have little meaning, while systematic relationships between whole rock 187Os/188Os ratios and indices of extent of melting will still remain. Finally, some peridotites may have been affected by several episodes of partial melting. If full re-equilibration does not occur during the later events, enclosed Type 1 BMS may retain 187Os/188Os compositions acquired during earlier events. For example, the TRD age of enclosed Type 1 BMS might reflect an ancient melting event in the convecting mantle, rather than the time of lithospheric formation (Pearson et al. 2004). In this case, the Os-isotopic data obtained from both whole rock and individual BMS analyzes will have ambiguous interpretations.
These considerations suggest that neither whole-rock nor individual BMS analyzes consistently yield 187Os/188Os ratios with simple interpretations. The two types of analyzes are complementary; neither can be viewed as “better” than the other. Only a thorough petrographic and geochemical study of a xenolith suite can allow the information from the two methods to be reconciled.
THE LIFE OF A XENOLITH, AS RECORDED IN HSE- AND Os-ISOTOPE SYSTEMATICS
The HSE and Os-isotope characteristics of a peridotite xenolith reflect a large number of processes and factors, including the nature of the original source material, the mechanisms and extent of partial melting, later melt percolation and metasomatism, and late-stage modifications linked to eruption and surface residence. In the following sections, we will examine each of these aspects.
The HSE and 187Os composition of the Primitive Upper Mantle
In their pioneering paper that laid the basis of Re–Os isotope geochemistry, Allègre and Luck (1980) found that mantle-derived samples of different ages follow a roughly chondritic Os-isotopic evolution curve. As discussed in the introduction, the most widely accepted hypothesis proposed to explain both this observation and the chondritic relative HSE abundances of fertile mantle samples is that nearly all of the HSE are derived from the late addition of meteoritic material after the end of core formation (Kimura et al. 1974; Chou 1978). However, the material included in this “late veneer” was not necessarily homogeneous. In fact, the HSE (Horan et al. 2003; Fischer-Gödde et al. 2010) and Os-isotopic (Walker et al. 2002a; Day et al. 2016, this volume) compositions of chondritic meteorites vary considerably, and it is possible that the late additions included material with HSE compositions extending beyond the measured meteoritic range (Walker 2009). Thus the very existence of a well-defined primitive mantle composition is not assured, and depends on the ability of the Hadean mantle to homogenize the potentially heterogeneous HSE-bearing extraterrestial material provided. Partial melting and model age calculations are based on the inherent assumption that the mantle has a uniform composition that can be characterized, but this may be only an approximation.
Crustal extraction will cause the average bulk mantle composition of all elements to differ from that of the Bulk Silicate Earth. However in the case of the HSE and Os isotopes, this difference should be extremely small, both because of the low mass of the crust relative to that of the mantle (~0.6%) and because of the very low HSE concentrations in the crust. Based on the analysis of loess samples, Peucker-Ehrenbrink and Jahn (2001) estimated an Os concentration of ~30 pg.g−1 coupled with a 187Os/188Os ratio of ~1.4 for the upper continental crust. It is difficult to obtain a robust estimate of the average Os content of the lower continental crust, but lower crustal xenoliths yield median Os concentrations (North Queensland, 49 pg.g−1, Saal et al. 1998; Central Arizona, 31 pg.g−1, Esperança et al. 1997) that are similar to the upper crustal estimate. Assuming PUM-like Os parameters ([Os] = 3.9 ng.g−1; 187Os/188Os = 0.1296; Meisel et al. 2001; Becker et al. 2006) mass balance calculations show that extraction of the crust should decrease the 187Os/188Os ratio of the remaining mantle by <0.1% while increasing the absolute concentrations of Os by 0.6%. Thus the effect of crustal extraction on the HSE composition of the primitive mantle is insignificant.
Several methods have been employed to characterize the average Os-isotopic composition of the convecting upper mantle reservoir that presumably provided the source of the lithospheric mantle later sampled by peridotite xenoliths. Meisel et al. (1996, 2001) found that the frequently observed correlations between Al2O3 and 187Os/188Os in peridotite massifs and alkali-basalt-borne peridotite-xenolith suites converge near the approximate alumina content of an unmelted primitive mantle peridotite (taken by these authors as 4.2 %, Meisel et al. 2001). They thus suggested the corresponding 187Os/188Os ratio (0.1296 ± 0.0008) to be that of the PUM. More recent studies (Becker et al. 2006) tend to confirm the work of Meisel et al. (2001), as most peridotite-xenolith suites from alkali basalts do indeed define broad correlations trending toward the proposed PUM value, though many of these trends are only weak and show the combined effects of several processes (Fig. 4). As noted by Meisel et al. (2001), the fact that fertile samples from peridotite suites of different ages define a single temporal evolution curve, with a 187Re/188Os ratio of 0.435 (calculated assuming the same values for initial 187Os/188Os ratio, age of the Earth, and 187Re decay constant used by Shirey and Walker (1998) to define the chondritic evolution curve), is consistent with the limited transfer of Re to the continental crust implied by mass balance calculations. The proposed PUM 187Os/188Os ratio of Meisel et al. (2001) is comparable to those of enstatite and ordinary chondrites, but is somewhat more radiogenic than the carbonaceous chondrite range (Walker et al. 2002a). Walker et al. (2002b) used chromitites from ophiolites to estimate an average 187Os/188Os value of 0.12809 ± 0.00095 for the convecting upper mantle, although O’Driscoll et al. (2012) suggested that ophiolitic chromitites most likely represent the upper limit rather than the average composition of this reservoir. Walker et al. (2002b) attributed the ~1.2% difference between the average chromitite value and the Meisel et al. (2001) PUM estimate to the long-term isolation of subducted ocean crust with relatively high Re contents. Based on analyzes of picrites from West Greenland and Baffin Island, Dale et al. (2009) proposed a present-day 187Os/188Os ratio of 0.1276 ± 0.0007 for the upper convecting mantle reservoir, similar to the value proposed by Walker et al. (2002b). Abyssal peridotites have also been used to constrain the average upper mantle 187Os/188Os ratio. As summarized by Harvey et al. (2006), whole rock analyzes of these rocks yield a wide range of values, with an average 187Os/188Os ratio (~0.1246) that is substantially lower than the PUM value of Meisel et al. (2001). In contrast, mid-oceanic ridge basalt (MORB) glasses free of known analytical artifacts yield an average 187Os/188Os ratio of 0.133 ± 0.009 (Gannoun et al. 2007, 2016 this volume). However, these authors argue that most MORB glasses have been contaminated by assimilation of older oceanic crust, since Os-rich BMS separated from MORB glasses yield a much less radiogenic average composition of 0.1263 ± 0.0012 (Gannoun et al. 2007, 2016 this volume, much closer to the abyssal peridotite average.
In summary, plausible arguments can be made to support values anywhere in the range 0.125–0.130 for the average present-day 187Os/188Os ratio of the upper part of the convecting mantle. This 4% ambiguity must be borne in mind when trying to define the composition of the source that melted to produce the peridotites later sampled as xenoliths in alkali basalts.
The HSE abundances of the PUM were estimated using the co-variations of HSE concentrations and inter-element ratios with lithophile indices of melt depletion (e.g., Al2O3 and Lu) in mantle peridotites (Morgan 1986; McDonough and Sun 1995; Becker et al. 2006; Fischer-Gödde et al. 2011). The estimates of Morgan (1986) are based purely on non-cratonic alkali-basalt-hosted spinel-bearing peridotites from the historical Basaltic Volcanism Study Project (1981) (e.g., Kilbourne Hole, New Mexico; Mount Quincan, Queensland, Australia; San Carlos and Cochise Crater, Arizona), but the subsequent three studies integrated data for mantle peridotite xenoliths from cratonic and non-cratonic environments along with data from peridotite massifs (McDonough and Sun 1995; Becker et al. 2006; Fischer-Gödde et al. 2011). The Fischer-Gödde et al. (2011) study used the same samples and broad analytical approach as Becker et al. (2006) but adapted these techniques to allow analysis of the mono-isotopic elements Rh and Au.
When considered together (Morgan 1986; Becker et al. 2006; Fischer-Gödde et al. 2011), it appears that only broad correlations exist between absolute concentrations of any of the HSE and whole-rock Mg# or Al2O3 contents of the peridotites. Large variations in concentration existing at any given Mg# or Al2O3 content, assumed to result from partial melting alone are in fact a collage of many different petrological processes. Nevertheless, if inter-element ratios rather than concentrations are considered, much more systematic behavior is observed, allowing the HSE composition corresponding to fertile mantle peridotite to be constrained. The picture emerging from these investigations confirms that the PUM contains ng.g−1 level concentrations of the HSE (Fig. 6). When comparing the primitive mantle composition estimated by Morgan (1986), McDonough and Sun (1995) and Becker et al. (2006) with Fischer-Gödde et al. (2011) significant differences appear. Between 1986 and 2011, the estimated HSE concentrations of the PUM have increased by 10–15% for the heavy HSE (Os, Ir, Pt), 25–40% for Ru, Rh, and Re and by 70–80% for Au and Pd. Overall, the corresponding range of concentrations is 3.1–3.9 pg.g−1 Os, 3.2–3.5 pg.g−1 Ir, 5–7 pg.g−1 Ru, 0.9–1.19 pg.g−1 Rh, 7.1–7.6 pg.g−1 Pt, 3.9–7.1 pg.g−1 Pd, 1–1.7 pg.g−1 Au, and 0.26–0.35 pg.g−1 Re. All three PUM estimates overlap within one standard deviation uncertainty except for Ru and Pd, the two most chalcophile HSE. One may argue that these PUM concentration variations may reflect the evolution of analytical methods from 1986 (INAA) to 2006/2011 (High-pressure/high-temperature digestions and ICPMS/TIMS measurements), especially since the high-pressure dissolution techniques used by Becker et al. (2006) and Fischer-Gödde et al. (2011) are known to result in a more efficient dissolution of the refractory Os–Ir–Ru–Pt alloys, which are prone to form in some mantle peridotites and are known to be difficult to digest (Luguet et al. 2007; Lorand et al. 2008). However, several of the Kilbourne Hole samples (e.g., UM8, UM5) were analyzed by both Morgan (1986) and Becker et al. (2006) and the results are in excellent agreement, arguing against an analytical artefact. The variations in the PUM estimates relate instead to the database itself, namely its limited size, and the provenance of the mantle peridotites (i.e., whether only xenoliths or all types of mantle peridotites are considered). In the following plots, we will use the combined estimates of Becker et al. (2006) and Fischer-Gödde et al. (2011) to represent the PUM composition.
Whole-rock observations on samples representing melting residues
To understand the consequences of partial melting for the mantle residue, attention has been focused on lherzolites and harzburgites whose whole-rock major element, trace element, and mineral compositions (Mg number in olivine) indicate that they are partial melting residues that experienced no or very limited melt-rock reactions. Porphyroclastic to protogranular textures were also part of the selection criteria; poikilitic peridotites were avoided, as these latter are considered to have formed from reaction and re-equilibration with large volumes of silicate melt (e.g., wall rocks of melt pathways) (see Grégoire et al. 1997; Xu et al. 1998). From our compilation of ca. 400 samples with both HSE and Os-isotopic data, 33 peridotites seem to be appropriate targets for constraining the effect of partial melting on the HSE and 187Os signatures (database available upon request to the authors). These samples come from both oceanic and continental settings and represent localities that are geographically widely separated.
When compared to the HSE pattern of the primitive mantle (Becker et al. 2006; Fischer-Gödde et al. 2011), these lherzolites and harzburgites show a relatively flat segment from Ir to Ru or sometimes Rh and a marked depletion of Pt–Pd–Re in comparison (Fig. 7). The Ir, Ru, Rh, and Pt concentrations define relatively narrow intervals. Interestingly, in the case of Ir and Ru, the concentrations in the mantle residues overlap with the PUM estimated composition but also extend toward lower concentrations those of the PUM (Fig. 7) and of the orogenic, ophiolitic, and abyssal peridotites (Fig. 2). Progressive depletion is noticeable from Pt to Pd and to Re. These patterns mirror those traditionally obtained in partial melts such as MORB or OIB (Ocean Island Basalts) lavas which show progressive enrichment from Pt to Re (see Bézos et al. 2005; Day 2013; Lissner et al. 2014; Gannoun et al. 2016, this volume). It is worth noting that while these HSE patterns systematically show a steep Rh–Pd segment considered to be characteristic of partial melting, some HSE patterns are characterized by a flat Pd–Re segment, which argues for a post-melting disturbance of Re, although the ReN/OsN is still sub-chondritic. When the HSE are plotted against partial melting proxies, such as the whole-rock Al2O3 contents, although there is a relatively large scatter in the data, some trends appear. A clear positive co-variation is observed for Re with Al2O3 (r2 = 0.58), and a fainter one for Pd with Al2O3 (r2 = 0.21) (Fig. 8). On the other hand, although shifted quite systematically toward lower than PUM concentrations, Rh and Ru concentrations seem to increase as the residues become increasingly depleted. In the larger Ru database, we see that the behavior of Ru is bimodal, with constant Ru concentrations for residues with Al2O3 > 1.5 wt% and a sharp but steady increase in Ru concentrations with decreasing Al2O3 for harzburgites with Al2O3 < 1.5 wt% (Fig. 8). Osmium and Ir behave similarly to Ru, and as will be explained below, the observed increasing abundance of these compatible HSE with decreasing Al2O3 content is expected in residues of melt extraction. Nevertheless, preservation of a partial melting trend for the Os concentrations is quite surprising considering that the lower than PUM Os concentrations of non-cratonic peridotite xenoliths likely result from secondary processes greatly post-dating the original melting event (see “Syn- to post-eruptive processes”). Platinum offers an even more complex behavior, with scattered concentrations for mantle residues with Al2O3 > 1.5 wt%, while for residues with Al2O3 < 1.5 wt%, two trends emerge, highlighting a Pt enrichment in highly depleted peridotites similar to that observed for Os, Ir, and Ru, or in contrast a marked depletion as seen for Pd and Re. Generally, the HSE ratios may be more robust tracers of partial melting signatures than HSE concentrations. As such, the RuN/IrN ratios remain chondritic in moderately depleted mantle residues (Al2O3 > 2 wt%), but extend toward suprachondritic values in highly depleted residues (Fig. 8). In contrast, ReN/IrN and PdN/IrN decrease with increasing partial melting, although the PdN/IrN variations show more scatter than the ReN/IrN ratios.
Because of the increasing Re depletion during melting, partial melting should leave residues with variably depleted Re/Os ratios that with time will develop variable 187Os/188Os ratios. The 33 selected peridotites from our compilation that show HSE patterns typical of partial melting residues, have 187Os/188Os-isotopic ratios that are systematically unradiogenic, ranging over 0.1140–0.1246 for 187Re/188Os ratios of 0.01–0.32. However, only very rough correlations between 187Os/188Os and 187Re/188Os are observed, which is not surprising given that data from suites of various melt extraction ages are combined. As noted by Meisel et al. (2001) and discussed above, stronger correlations are often observed between 187Os/188Os and various indices of degree of melt extraction if individual suites are considered (Fig. 4).
Mineralogical control of HSE fractionation during partial melting
To understand the effect of partial melting on the HSE, we should first consider their behavior during partial melting of BMS, their main hosts in mantle peridotites. BMS have low partial-melting temperatures, in the range 850–1200 °C for Cu–Ni rich BMS to Fe-rich BMS, respectively (Craig and Kullerud 1969). This, added to the low viscosity of sulfide melts in mantle peridotites at the FMQ oxygen buffer (Gaetani and Grove 1999), implies that sulfide components are very likely to be extracted within the partial melt even at low degrees of partial melting. The solubility of S in silicate melts, which is a complex function of P, T, sulfur and oxygen fugacities, and iron activities (Haughton et al. 1974; Wallace and Carmichael 1992; Mavrogenes and O’Neill 1999; O’Neill and Mavrogenes 2002), will be the main factor controlling BMS consumption during partial melting. Assuming that non-cratonic spinel peridotites experienced partial melting at 1–2 GPa, a S solubility of 1000 μg.g−1 in the silicate melt (Mavrogenes and O’Neill 1999), typical of MORB melts (Jenner et al. 2010; Lissner et al. 2014) can be assumed. Thus the S content in the peridotitic residue can be modelled using the following equation Co = Cr * (1-F) + Cl * F, where Co is the S content in the fertile mantle source, F is the degree of partial melting and Cl is the S solubility in the partial melts. As S contents of the fertile mantle have been estimated to be ca. 250 ± 20 μg.g−1 (Lorand 1991), batch partial melting would result in a S-free mantle residue for 25% of partial melting while this will be achieved for melting degrees of 17–18% if the melting is considered to be fractional. So BMS and subsequently the whole-rock S abundances will decrease with increasing partial melting and thus will be correlated with lithophile indices of melt depletion.
The behavior of HSE and any possible fractionation among the HSE will depend on the state of the BMS during partial melting. Two scenarios can be envisaged. First, it can be assumed that at the high temperature and relatively high pressure at which melting takes place in the mantle, the BMS will exist as a molten sulfide matte (e.g., Fonseca et al. 2011; Mungall and Brenan 2014). The behavior of the HSE would thus be controlled by their partition coefficients between sulfide liquid and silicate liquid (Dsulf.liq/sil.liq.). Experimental studies since the 1990s have indicated partition coefficients generally in excess of 10,000, confirming the strong affinity of the HSE for sulfide liquids over silicate melts (Brenan et al. 2016, this volume). Nevertheless, while the absolute values of the partition coefficients between sulfide liquids and silicate melts range over 10,000–1,000,000, strikingly, Os, Ir, Pt, and Pd all have very similar Dsulf.liq/sil.liq. regardless of the experimental conditions and composition of the sulfide liquid (Stone et al. 1990; Fleet et al. 1996, 1999; Mungall et al. 2005; Mungall and Brenan 2014). A similar conclusion can also be drawn for the Dsulf.liq/sil.liq. extrapolated from MORB sulfide globules and silicate glasses (Peach et al. 1990). Equilibration between a silicate liquid and a molten sulfide matte would not create relative fractionations among Os, Ir, Pt, and Pd in the sulfide, because the absolute values of their partition coefficients are all similarly high (i.e., Dsulf.liq/sil.liq. > 10,000). Rhenium partition coefficients between sulfide matte and silicate liquid have been investigated less thoroughly, but range from 16,000 in the experimental work of Sattari et al. (2002) to much lower values (ca. 1,000–46) when determined from natural MORB samples (Roy-Barman et al. 1998, Gannoun et al. 2004). The experimental work and parametrisation of Fonseca et al. (2007) supported a similarly low Dsulf.liq/sil.liq. for Re (1–50). Brenan (2008), on the other hand, suggested somewhat higher Dsulf.liq/sil.liq.values (400–800) on the basis of his experimental results, but these are still much lower than the Dsulf.liq/sil.liq.values of the other HSE.
Modelling of the HSE concentrations in mantle residues is based on the same assumptions as modelling of the S content, except that the HSE concentrations in the mantle residues depend on both the residual amount of sulfides (Msulf) and the concentrations of the HSE in these sulfides (Csulf) (see Lorand et al. 1999 for detailed explanations). For the Dsulf.liq/sil.liq. partition coefficients now available (e.g., Mungall and Brenan 2014), increasing partial melting should result in an increase of the Os, Ir, Ru, Pt, Pd concentrations in the mantle residues until the essentially complete extraction of all BMS, which will result in a sudden and efficient stripping of these elements from the peridotitic residues (see Fig. 8 for Ru and Pd). Before the complete exhaustion of BMS is reached, this partial melting scenario would however create barely any fractionation between compatible and incompatible HSE or even among incompatible HSE (see Fig. 8 for RuN/IrN and PdN/IrN). Such a partial modelling scenario is unable to account for fractionations for example of Pd/Ir ratios or decreasing Pd concentrations with increasing extents of melting (Fig. 8). To mimic the HSE patterns of the natural mantle residues requires Dsulf.liq/sil.liq.for Pt and Pd of 2000 and 200 respectively; these are 5,000 and 2,500 times lower than the most recently determined set of Dsulf.liq/sil.liq. (Mungall and Brenan 2014). In contrast, rhenium would be less strongly retained by the sulfide matte and thus progressively extracted into the partial silicate melt (Fig. 8) considering its low Dsulf.liq/sil.liq.(Roy-Barman et al. 1998; Fonseca et al. 2007; Brenan 2008), and would thus fractionate from the other compatible (Ir) and incompatible HSE (Pt, Pd) in a way that would tend to resemble the partial-melting trend (Figs. 7 and 8).
Alternatively, the second partial-melting scenario assumes that BMS 1) host 100% of all HSE and 2) melt incongruently producing a Cu–Ni-rich sulfide melt and leaving a solid “monosulfide solid solution” behind. The physical extraction of these highly mobile Cu–Ni-rich sulfide melts (Gaetani and Grove 1999) along with the partial silicate melts would thus lead to fractionation of the HSE owing to their contrasting partitioning preferences between these two sulfide phases. Such incongruent melting of the BMS finds support in the experimentally determined concomitant stability of two sulfide phases (one solid, one liquid) of the Fe–Ni–Cu–S system at the P–T conditions of the upper convective mantle (Bockrath et al. 2004; Ballhaus et al. 2006) as well as in observations of the co-existence of different sulfide phases (e.g., MSS vs. Cu–Ni-rich sulfides) in natural samples (see Lorand and Conquéré 1983; Lorand 1987; Szabó and Bodnar 1995; Guo et al. 1999). The partition coefficients of the HSE between the MSS and the Cu–Ni-rich sulfide melt (DMSS/sulf.liq.) demonstrate that MSS will concentrate the compatible HSE (Os–Ir–Ru) and Re, which have DMSS/sulf.liq. between 1 and 10 (Brennan 2002). For a given set of experimental conditions, the partition coefficients of the compatible HSE and Re differ at most by a factor of two (e.g.,DMSS/sulf.liq. of 3.4 ± 1.1, 4.5 ± 1.5, 5.1 ± 1.2 and 5.3 ± 1.2 for Re, Os, Ir, and Ru for the run OS-2 of Ballhaus et al. 2006). In contrast, the incompatible HSE (PPGE: Pd, Pt) have a higher affinity for the Cu–Ni-rich sulfide melts, as suggested by their very similar DMSS/sulf.liq. ranging over 0.11–0.16 for Pt and 0.14–0.16 for Pd (Li et al. 1996; Bockrath et al. 2004; Ballhaus et al. 2006 for experimental runs involving sulfide phases with compositions similar to those obtained in natural peridotite samples). Considering the large difference of DMSS/sulf.liq. between the compatible HSE and Re, and the incompatible HSE, incongruent melting of the BMS would result in significant fractionations of Os–Ir–Ru–Re from Pt–Pd even for low degrees of partial melting and partial melting residues would be enriched in Re, Os, Ir, Ru over Pd and Pt. In mantle residues, the Pd/Ir fractionation would reflect the ratio of the DMSS/sulf.liq. of these two elements and the relative abundances of the MSS and Cu–Ni-rich sulfide phase. For extensive partial melting of an originally fertile peridotite with chondrite-like HSE composition, Bockrath et al. (2004) predicted the melting residue to contain 0.5 ng.g−1 Pd, 3.7 ng.g−1 Ir and have a sub-chondritic PdN/IrN ratio of 0.1 after complete extraction of the Cu–Ni-rich sulfide melt. These predictions match very well the Pd and Ir concentrations observed in non-cratonic harzburgite xenoliths having experienced extensive partial melting.
The incongruent melting of BMS cannot however reproduce the extreme Re depletion of the harzburgites, their sub-chondritic ReN/OsN or suprachondritic PdN/ReN ratios, nor the increasing depletion from Pt to Pd observed in the partial melting residues (Fig. 7). To account for these geochemical features within the framework of the BMS incongruent melting model, it must be assumed that Re and Pt are not purely hosted in the MSS or in the Cu–Ni-rich sulfide melt. In addition to being hosted in BMS, Re is also incorporated in the modally major minerals of a peridotite, with increasing Re concentrations detected from olivine to orthopyroxene, clinopyroxene and spinel (Burton et al. 1999; Harvey et al. 2010, 2011). Upon partial melting at relatively low pressure in the spinel stability field, Re will partition into the silicate melt, because even though it is hosted by solid silicate phases, it will behave an incompatible element during mantle partial melting as its partition coefficients between the silicates and oxides, and the silicate melt are systematically ≤ 0.1 at FMQ (Mallmann and O’Neill 2007). On the other hand, the more compatible behavior of Pt during partial melting likely reflects its stabilization as Pt-rich alloy during the partial melting event. In fact, the experiments of Peregoedova et al. (2004) demonstrated that a mantle-like MSS undergoing desulfidation (fS2 decrease) as expected during the consumption of BMS during partial melting, would exsolve micrometric Pt-Ir alloy. The presence of this alloy would prevent the Pt from partitioning fully into the Cu-Ni-rich sulfide melt and explain the high Pt concentrations observed in some harzburgites as well as the Pd–Pt decoupling during partial melting.
Both models fail to explain the persistence of HSE in mantle residues even when the degree of partial melting is so high that all the BMS are extracted into the partial melt. Both scenarios predict no HSE left in BMS-free mantle residues, while for example the Kerguelen peridotites OB-98-58 and OB-98-426 (Lorand et al. 2004) and MBRX from the Massif Central (Lorand and Alard 2001), which are BMS free still contain 16–26 ng.g−1 total HSE. The occurrence of refractory platinum group minerals (PGM), such as laurite–erlichmanite or Ru–Os–Ir-rich alloys could account for the persistence of HSE in BMS-free residues. These high-temperature PGM, that exsolve in response to the oversaturation of the compatible HSE in the disappearing BMS during partial melting, have been optically identified in highly depleted mantle harzburgites from orogenic massifs (Luguet et al. 2007; Kogiso et al. 2008), where they account for 50–100% of the whole-rock HSE abundances (Luguet et al. 2007). LA-ICPMS analyzes of an MSS bleb from the non-cratonic harzburgite GM 453 from Kerguelen (Delpech et al. 2012), whose whole-rock HSE pattern however differs from those typical of partial melting residues, reveal an unusual composition characterized by elevated OsN and RuN compared to IrN at much higher concentrations than “normal” MSS. For Delpech et al. (2012), this reflects the existence of laurite/Ru–Os–Ir alloys, which were subsequently captured within a metasomatic BMS and redissolved as the latter can accommodate a few hundred μg.g−1 of IPGE (Fonseca et al. 2012). At present, we can only speculate that the sharp change of the Os, Ir, Ru, and Pt behavior with whole-rock Al2O3 at 1.5 wt% (see Fig. 8) marks the beginning of the exsolution of these high temperature PGM during high degrees of partial melting or that if on the contrary it reflects a post-melting metasomatic event, during which these PGM were captured and fully redissolved within the added metasomatic BMS.
Transposing mineralogical scale processes to the whole-rock scale poses additional challenges. Testing and understanding which fine-scale partial melting models may be the most applicable is clearly linked to our understanding of the broader scale correlations used to discuss these processes, which in turn depend on a reasoned screening of the HSE- and Os-isotope data. As noted for both the 187Os/188Os compositions and the HSE concentrations, correlations with whole-rock major element parameters (e.g., Al2O3 content) are often faint. As such, the HSE systematics may not purely reflect partial melting, but also offer a record of post-melting events, during which HSE or 187Os were disturbed. Moreover, the whole-rock Al2O3 contents may also have been modified during mantle refertilization processes, when the percolation of large volumes of melt led to the crystallization of new minerals (olivine, clinopyroxene) (see Bodinier et al. 1988; Le Roux et al. 2007).
Ancient melt-extraction events recorded by 187Os/188Os systematics
The preceding discussion of processes occurring at the mineralogical scale may help explain the well-known observation that during partial melting of peridotite, Os behaves compatibly while Re is moderately incompatible (Shirey and Walker 1998). The bulk partition coefficient of Re is similar to that of the heavy rare earth elements (HREE) or aluminium. However this similarity is largely fortuitous, as the geochemistry of these elements is quite different. The HREE and Al are lithophile elements, while Re is both moderately lithophile and chalcophile, displaying a partitioning behavior in sulfides and silicates that depends strongly on oxygen and sulfur fugacity (Righter and Hauri 1998; Fonseca et al. 2007; Mallmann and O’Neill 2007; Brenan 2008). Nevertheless, though the reasons for the similarity in global behavior are complex, the overall result is that Re concentrations should be roughly proportional to those of the HREE and Al in peridotites that have experienced variable degrees of melt extraction. Osmium, on the other hand, is extremely chalcophile (e.g., Fleet et al. 1996, 1999; Mungall and Brenan 2014), being hosted almost completely by BMS in lherzolites. As long as a discrete sulfide phase remains during partial melting, essentially all of the Os will remain in the residue, and bulk-rock Os concentrations should increase as a result of mass loss during melting. Together, these considerations suggest that in a suite of peridotites that experienced variable degrees of melt extraction, bulk-rock Re/Os ratios should vary in a roughly linear manner with Al2O3, HREE, and other indices of melt extraction. Assuming that all of the rocks in the suite have approximately the same 187Os/188Os ratio at the time of melting, either because of a similar pre-melting history or because of isotopic equilibration during melting, radioactive decay with time will lead to the development of a positive correlation between 187Os/188Os and Al2O3 (for example) that can be viewed as an isochron analog (Fig. 4). As will be explained in the section on “Obtaining age information from Os composition of whole rocks”, such isochron analogs, which were first identified in massif peridotites (Reisberg and Lorand 1995), can be used for estimating model ages of melt extraction.
Though the isochron analog model provides a simple framework for the interpretation of whole-rock Os isotope systematics in peridotite xenoliths carried by alkali basalts, it clearly cannot explain the many cases where little or no correlation is observed between 187Os/188Os and melt-extraction indices (e.g., Handler et al. 2005; Ackerman et al. 2009; Harvey et al. 2010; Alard et al. 2011). It does not take into account abundant textural and geochemical evidence suggesting that many peridotites have been affected by melt percolation and/or refertilization processes (e.g., Lenoir et al. 2001; Le Roux et al. 2007). It also does not consider the evidence for isotopic heterogeneity among BMS from a single sample (e.g., Alard et al. 2002; Zheng et al. 2007; Harvey et al. 2011; González-Jiménez et al. 2013, 2014), which is usually attributed to addition of a radiogenic intergranular component (though as noted above at least some of this heterogeneity may result from radiogenic ingrowth). Traces of older events may also be preserved in BMS shielded in silicate phases (e.g., Pearson et al. 2004) if later melting events do not succeed in completely re-equilibrating Os-isotopic compositions at the scale of the rock. Simple partial-melting models are also inconsistent with the HSE systematics observed in certain suites, such as the common presence of flat or erratic relationships between Os or Ir, and Al2O3, rather than the expected shallow negative trends (Fig. 4). Thus melt extraction alone is not sufficient to explain many of the HSE features observed in peridotite xenolith suites.
Post-melting petrological history
After undergoing partial melting, the former convecting upper mantle may be stabilized in the subcontinental lithosphere, where it likely resides for hundreds of millions to billions of years, and is subjected to possible interaction with ascending fluids and melts. Primary silicate melts are produced through partial melting of mantle sources of various compositions, via variable degrees of partial melting and over different ranges of pressure. These melts may evolve compositionally through melt/rock reactions with the lithospheric mantle, and as a result may exsolve a volatile-rich fluid. These percolating melts and fluids react with the lithospheric mantle, which triggers modifications of the post-melting composition of the mantle peridotites, dependent upon the melt and fluid compositions and the percolation process. Based on detailed petrological and geochemical investigations on spinel-peridotite xenoliths (e.g., Bedini et al. 1997; Xu et al. 1998), as well as on orogenic and ophiolitic mantle massifs (e.g., Van der Wall and Bodinier 1996; Dijstra et al. 2003), the bottom of the lithosphere near the asthenosphere is considered to have a higher porosity (2–3%) than the mantle present at shallower depth (< 1%) closer to the Moho. Such a negative porosity gradient implies that percolation of large volumes of asthenospheric melts may have taken place at the base of the lithospheric mantle, whereas smaller melt fractions may be able migrate to shallower levels.
Metasomatism linked to percolation of S-undersaturated silicate melts
Among the worldwide collection of spinel-peridotite xenoliths, those having experienced percolation of large volumes of melts are generally characterized by poikilitic, poikiloblastic, granular to equigranular (micro)textures with mostly strain-free silicates and have equilibrated at high temperatures (Wells 1977; Xu et al. 1998, Massif Central peridotites; Bedini et al. 1997, East African mantle xenoliths, Grégoire et al. 1997; Moine et al. 2004, Kerguelen peridotites; Ackerman et al. 2007, Bohemian Massif; Xu et al. 2003, East China mantle xenoliths). Their trace element patterns, along with the homogeneous mineral composition within the whole lherzolite–harzburgite suite and the absence of amphibole or phlogopite argue for extensive re-equilibration of a harzburgitic residue with asthenospheric melts with a basaltic composition.
In the xenoliths that show these petrological and geochemical characteristics, the HSE concentrations are mostly lower than those of the PUM, in most cases by a factor of 2 or 3 (i.e., Os: 1.41 ± 1.44 (2 stdev) ng.g−1 vs. 3.9 ng.g−1; Ir: 1.37 ± 1.44 ng.g−1 vs. 3.5 ng.g−1; Ru: 3.00 ± 3.04 ng.g−1 vs. 7.0 ng.g−1; Rh: 0.47 ± 0.38 ng.g−1 vs. 1.2 ng.g−1 Rh; Pt: 6.00 ± 17.1 ng.g−1 vs. 7.6 ng.g−1; Pd: 2.18 ± 7.6 ng.g−1 vs. 7.1 ng.g−1; Re: 0.028 ± 0.035 ng.g−1 vs. 0.35 ng.g−1; Au: 0.5 ± 0.53 ng.g−1 vs. 1.7 ng.g−1 Au). The corresponding HSE patterns display a broad arch-shape, being depleted in Os, Pt, and Pd in comparison to Ir and Ru. Their arch shapes resemble those resulting from melt depletion alone but differ in their larger scatter of Ir–Ru–Rh–Pt concentrations, which are mostly below the PUM estimates (Fig. 9). The combination of low HSE abundances, low S and Se concentrations for chondritic S/Se ratios and the rarity of BMS in thin sections from such peridotite xenoliths suggest that the BMS originally present have been removed during melt/rock reactions at high melt/rock ratios (Lorand and Alard 2001; Lorand et al. 2003a,b, 2004). The high temperatures of re-equilibration recorded by these xenoliths (T ≫ 1100 ºC) are well above the liquidus temperatures of both Cu-rich sulfides and MSS (e.g., Craig and Kullerud 1969; Fleet and Pan 1994; Ballhaus et al. 2006). This, along with the higher fO2 recorded during melt/rock reactions (Lorand and Alard 2001; Hanger et al. 2015) would facilitate remobilization of the BMS, which is suggested by the numerous veinlets of Cu-rich sulfides and the association of intergranular MSS with neoblasts as observed in the East African peridotites (Lorand et al. 2003a). The decrease of the HSE concentrations (Fig. 9) additionally implies the partial dissolution of the remobilized BMS within the percolating melt. Primary partial melts generated at depth, such as OIB-like melts, would likely become S-undersaturated during their ascent due to decreasing pressure (Mavrogenes and O’Neill 1999), and thus will have the ability to scavenge BMS as well as the S, Se, and HSE that they host, resulting in a decrease of the HSE, S, and Se concentrations in such highly reacted mantle rocks (e.g., Harvey et al. 2015b).
The extent of the melt/rock reactions and the composition of the post-melting peridotitic residue will account for the large range of HSE patterns observed as a result of these pervasive melt-percolation processes. Preferential loss of Cu–Ni-rich sulfides over MSS grains (if the silicates do not recrystallize and instead act as a repository for the MSS inclusions) would be reflected in MSS-like HSE patterns with high Ir–Ru–Rh concentrations but low Pt–Pd contents. Delpech et al. (2012) also proposed that reaction of the BMS with such a melt would lead to the desulfidation of residual MSS, liberating Pt–Ir–Os alloys and ultimately leaving behind a Ru–Rh bearing MSS. Peridotites that experienced extensive interaction at high melt/rock ratios may display OIB-melt-like HSE patterns, with a slight positive slope from Os to Pd (Lorand et al. 2003a), possibly reflecting the occurrence of MSS directly precipitated from the percolating melts.
Peridotite xenoliths from the Sidamo region of the East African rift system (Bedini et al. 1997; Lorand et al. 2003a; Reisberg et al. 2004) provide a good example of the effects of pervasive melt percolation on HSE and Os-isotope systematics. In this region, ancient (> 2 Ga) melt extraction produced variably depleted harzburgites and lherzolites that preserve a rough correlation between 187Os/188Os and Al2O3 (Fig. 4 k,l). However, both lithologies display two textural types, granular and deformed, that correspond respectively to extensive re-equilibration with large volumes of basaltic melt near the bottom of the lithosphere, and overprinting by smaller amounts of evolved, volatile-rich melts at higher levels. The higher degrees of melt percolation evident in the granular peridotites correlate with low total PGE contents as well as PGE patterns and low Pd/Ir ratios suggestive of interaction of MSS with large volumes of OIB-like melts. This process may also have increased the 187Os/188Os ratios of the granular lherzolites, with one clinopyroxene-rich sample (ET32) displaying a 187Os/188Os ratio similar to PUM. However as there are no firm constraints on the compositions prior to melt percolation, and no systematic relationships between the amount of metasomatic clinopyroxene and the Os-isotopic composition, the importance of this process is difficult to quantify.
Further examples of the effects of pervasive melt percolation can be found in eastern China (Reisberg et al. 2005). In the Subei Basin region, xenoliths from two sites separated by ~6 km (Lianshan and Panshishan) define a correlation between 187Os/188Os and indices of melt extraction (Fig. 4 e,f) suggestive of ancient melt depletion in the underlying mantle lithosphere. However, despite the close proximity of the two sites, xenoliths from Lianshan display much lower Os, Re, sulfur, and modal BMS contents than those from Panshishan for similar extents of melt depletion. Unlike in Sidamo, the textures of the xenoliths from both localities are similar, being predominantly protogranular, so textural evidence for extensive melt percolation is lacking. Nevertheless, the much lower chalcophile element abundances in Lianshan are most easily explained by pervasive percolation of S-undersaturated melts. It is possible that this process also explains the internally heterogeneous and radiogenic Os-isotope ratios observed for a few Lianshan samples (though an additional fluid-based metasomatic process cannot be excluded), but overall the Os-isotopic systematics seem to have been preserved. The lack of correlation between Os-isotopic composition and 187Re/188Os, which contrasts with the much better correlation between 187Os/188Os and Yb suggests that the melt percolation leading to the perturbation of the Re/Os ratios was geologically recent. A very similar example may be found in the Hannuoba and Yangyuan xenolith suites from the Trans-North China Orogen (Liu et al. 2010). As in the Subei basin example, both suites define a single correlation in a plot of 187Os/188Os vs. Al2O3 (Fig. 4c,d), suggesting that the lithosphere beneath both localities experienced the same ancient magmatic episode, but one locality (Yangyuan) has much lower S, Se, Os, Pd, and Re contents than the other (Hannuoba). Highly siderophile element patterns for the Yangyuan samples consistently show a marked hump in Ir and Ru, coupled with depletion in Os, and progressive depletion from Pt through Re, while the Hannuoba samples show much flatter patterns similar to those observed in massif peridotites. Based on these observations and on oxygen fugacity determinations, Liu et al. (2010) suggest that the HSE features of the Yangyuan peridotites reflect the recent percolation of S-undersaturated oxidizing melts, resulting in incongruent sulfide breakdown. They also note that this process has apparently had little effect on 187Os/188Os.
Metasomatism linked to percolation of volatile-rich silicate melts
Percolation of volatile-rich melts may be directly linked to the reactive porous flow percolation of S-undersaturated OIB-like melts that infiltrated the base of the lithosphere. The melt/rock reactions at high melt/rock ratios (i.e., large melt volume) lead to precipitation of various mineral phases, including metasomatic phases (i.e., amphibole, phlogopite, apatite, rutile) and volatile-free typical peridotite minerals (i.e., clinopyroxene, garnet, spinel, orthopyroxene, and olivine); the latter type of metasomatism has been referred to as “stealth metasomatism” (O’Reilly and Griffin 2012) because it is not always simple to detect. Phase precipitation consumes the melt and shifts its composition towards incompatible-element-richer compositions, as documented in Kerguelen (e.g., Grégoire et al. 1997; Moine et al. 2004), or East African peridotite xenoliths (Bedini et al. 1997). The resulting smaller melt fractions consist of evolved silicate melts that are likely S-saturated or close to S saturation. In the Kerguelen and East Africa peridotites, these were probably carbonatitic melts and as such may have produced the geochemical signatures generally ascribed to “carbonate metasomatism”. Primary silicate melts enriched in volatiles such as S may also be produced from low degrees of partial melting of the mantle. Regardless of the melt origin, melt/rock reactions with S-saturated silicate melts result in precipitation in the peridotitic rocks of Cu–Ni-rich BMS (Type 2 BMS) generally associated with metasomatic clinopyroxene as well as minerals/components more symptomatic of carbonatitic melts, e.g., carbonatitic melt pockets (see “Origin of base metal sulfides and platinum group minerals”). However, the extent of these melt/rock reactions from the perspective of the silicate phases may be quite variable. Percolation of S-saturated silicate melts may contribute to modal metasomatism, in particular, precipitation of metasomatic clinopyroxene. If extensive, the modal metasomatism can be described as refertilization (modification of the whole-rock composition) converting harzburgitic protoliths into “fertile” lherzolites, as shown by Le Roux et al. (2007). These authors suggest that such a process would imply the existence of positive trends between the degree of fertility of the peridotites (expressed as Al2O3) and the HSE abundances as well as the 187Os/188Os signatures, and argue that these trends would reflect mixing but definitely not melting. Alternatively, melt/rock reaction may operate in a more cryptic manner (low melt/rock ratios) without affecting significantly the mineralogical modes of the peridotites or leading to refertilization. In this configuration, HSE concentrations as well as 187Os/188Os signatures will vary independently of the degree of fertility of the peridotites.
Peridotites showing evidence of percolation by S-saturated silicate melts typically have HSE abundances of Ir to Pt overlapping with the most recent estimate for the Primitive Upper Mantle composition (Becker et al. 2006; Fischer-Gödde et al. 2011) (Ir: 3.46 ± 1.28 (2 SD) ng.g−1 vs. 3.5 ± 0.8 ng.g−1; Ru = 6.84 ± 2.85 ng.g−1 vs. 7.0 ± 1.8 ng.g−1, Pt = 7.07 ± 3.4 ng.g−1 vs. 7.6 ± 2.6 ng.g−1, respectively), without distinctive deviations in term of HSE inter-elemental fractionations (RuN/IrN and PtN/IrN = 1.33 ± 0.47 and 0.99 ± 0.49 vs. 1.34 ± 0.24 and 1.04 ± 0.42 for PUM) (Fig. 10). In contrast, Pd and Re show a much larger span of concentrations as the relative standard deviations (2 SD) rise to 70 and 90% (40–50% for the other HSE), and tend to lower concentrations in comparison to the PUM (Pd = 5.66 ± 3.97 ng.g−1 vs. 7.1 ± 2.6 ng.g−1; Re = 0.194 ± 0.84 ng.g−1 vs. 0.35 ± 0.12 ng.g−1) (Fig. 10). Additionally, the whole-rock 187Os/188Os ratios are systematically non-radiogenic (0.1199–0.1278). For Hannuoba, which has been extensively analyzed for the HSE and 187Os signatures at the whole-rock scale (Becker et al. 2006; Fischer-Gödde et al. 2011; Gao et al. 2002; Liu et al. 2010), the 187Os/188Os increases with the fertility degree of the peridotites (Al2O3) (r2 = 0.84), while PdN/PtN and PdN/IrN display much weaker trends (r2 = 0.45 and 0.48, respectively). The other peridotites showing clear signs of reactions with S-saturated silicate melts are fewer per xenolith locality (e.g., n = 5 or 6 in SE Australian xenoliths from Handler and Bennett 1999, and n = 5 in the Kerguelen peridotites analyzed by Lorand et al. 2004). None of these localities exhibits positive trends between Al2O3 and PdN/IrN or PdN/PtN, but the data for 187Os/188Os are scarce.
Lower than PUM Pd and Re concentrations, in combination with subchondritic PdN/IrN and ReN/OsN and, for example, PdN/IrN and 187Os/188Os vs. Al2O3 positive trends, could be interpreted as evidence of partial melting. However partial melting alone can explain neither the flat or positively-sloped Pt–Pd sections of the PGE patterns (average PdN/PtN = 1.23 ± 0.67) nor the occurrence, although rather infrequent, of suprachondritic PdN/IrN and ReN/OsN and the decoupling of PdN/IrN with whole-rock Al2O3 (Fig. 11). Conversely, one may argue that some of the samples showing these features (e.g., MQ19, Handler and Bennett 1999; DMP 56; Becker et al. 2006; Atl-3B and Atl-3V; Wittig et al. 2010) may be typical PUM-like peridotitic material. However the major element compositions of these four samples (Al2O3 = 1.98–3.49 wt% vs. 4.2 wt% for PUM estimates, Meisel et al. 2001) or their 187Os/188Os compositions require 5–15% partial melting (Fig. 11). This comparison worsens if all the peridotites showing flat to positive sloped Pt–Pd segments are considered, as the whole-rock Al2O3 contents extend down to 0.44 wt% (peridotite MM-94-101 from Kerguelen; Lorand et al. 2004).
Considering the clues provided by BMS petrography, the HSE patterns are best explained if we assume that most of these peridotites have experienced partial melting and subsequent precipitation of Type 2 BMS during the percolation of volatile-rich melts, evolved or primary in composition (Lorand et al. 2003a, 2004; Reisberg et al. 2004, 2005; Ackerman et al. 2009; Wang et al. 2009; Wittig et al. 2010; Alard et al. 2011). Type 2 BMS, are mainly made up of pentlandite and Cu-rich sulphides (i.e., chalcopyrite). Laser ablation ICPMS analyzes have demonstrated that Type 2 sulfides are characterized by chondritic to suprachondritic PdN/IrN likely dependent upon the relative proportions of pentlandite to chalcopyrite (Alard et al. 2000, 2011) (Fig. 5). Furthermore, irrespective of their type (i.e., Type 1 and Type 2 BMS), positive correlations exist between 187Os/188Os and 187Re/188Os among all of the BMS present in single peridotite xenoliths from SE and eastern Australia (Alard et al. 2002; Pearson et al. 2002; Powell and O’Reilly 2007), from the Massif Central (Alard et al. 2002, 2011; Pearson et al. 2002) and Taiwan (Wang et al. 2009). When the petrographic information on the BMS is available (Alard et al. 2002; Pearson et al. 2002), the metasomatic intergranular Type 2 BMS show suprachondritic 187Re/188Os and more radiogenic 187Os/188Os (but not systematically radiogenic signatures) than the Type 1 BMS, which are enclosed in refractory silicates (e.g., olivine) and interpreted as residual. These latter are characterized by sub-chondritic 187Re/188Os and unradiogenic 187Os/188Os. Alard et al. (2002) additionally showed that three interstitial Type 2 BMS in the peridotite “Gam VL11” exhibit a large range of 187Re/188Os but rather similar 187Os/188Os compositions varying from 187Os/188Os = 0.1257 ± 0.0053 to a PUM-like 187Os/188Os = 0.1282 ± 0.0008. This could reflect the relatively recent precipitation of these Type 2 BMS as they do not show post-metasomatic ingrowth of 187Os, that would result in a positive 187Re/188Os and 187Os/188Os. This also shows that the peridotite Gam VL11 reacted with a mantle-derived silicate melt. The origin of the melt can also be constrained if several Type 2 BMS from a given peridotite show post-metasomatic 187Os ingrowth, since the BMS with the lowest 187Os/188Os and 187Re/188Os constitute a record of the maximum isotopic composition of the percolating silicate melts.
When compared to the 187Os/188Os signatures of the BMS, the whole-rock composition plots on the trend defined by the different BMS populations, either more toward the middle, in between the enclosed and intergranular BMS (Alard et al. 2011) or in some cases closer to the enclosed BMS endmember (Harvey et al. 2010, 2011). When replicate whole-rock 187Os/188Os data are available, the replicates move along this trend. This indicates that the whole rock corresponds to a mixture of the Types 1 and 2 BMS, with variable relative contributions between the replicates (see peridotites Gam VL11 and MGam1: Alard et al. 2002). If enclosed BMS dominate the whole-rock Os budget (due to their higher Os concentrations compared to the intergranular BMS), the whole-rock 187Os/188Os signatures would resemble that of the BMS inclusions. Alternatively, when the relative modal abundances of the intergranular BMS increase, the whole-rock 187Os/188Os signatures are controlled by both the enclosed and intergranular BMS and shift toward the 187Os/188Os signature of the intergranular BMS endmember.
In order to model the effect on the HSE systematics of the precipitation of metasomatic Type 2 BMS in partial-melting residues during the percolation of volatile-rich silicate melt fractions (Fig. 12), the S-free peridotite KH03-16 from Kilbourne Hole (Harvey et al. 2015b) is chosen as the partial melting residue. Type 2 BMS AR10f from SE Australia peridotite xenoliths (Alard et al. 2000) and MBS-1-s6 from the Montboissier xenolith locality in the Massif Central (Lorand and Alard 2001) represent a likely Type 2 BMS endmember composition. These two Type 2 BMS show smooth positively sloped HSE patterns, which likely correspond to the bulk composition of Type 2 BMS (Pn + Cpy mixtures and Pt–Te nuggets exsolved during subsolidus reactions). The absolute Os–Ir–Ru concentrations of the Type–2 BMS endmember are chosen to be similar to those of grains MBS1-s2, s4; 80-24-s1, s2, s3 (Lorand and Alard 2001) and Mtf-s4 (Alard et al. 2011), from the Mont Briançon and Montferrier localities in France, respectively. The composition of the metasomatic Type 2 BMS component is therefore 16.2 μg.g−1 Os, 12.2 μg.g−1 Ir, 24.9 μg.g−1 Ru, 40 μg.g−1 Pt, 45 μg.g−1 Pd and 0.9 μg.g−1 Re. Using these parameters, the HSE patterns of the peridotites that reacted with volatile-rich small melt fractions can be explained by addition of 0.005–0.02 wt% of Type 2 BMS into an effectively S-free harzburgite (Fig. 12). These weight fractions of Type 2 BMS convert into an overall addition of 16–66 μg.g−1 S, which is the “normal” S content for peridotitic xenoliths and perfectly overlaps with the S contents of the peridotites, showing evidence of melt/rock reactions with S-saturated silicate melts (e.g., 5–320 μg.g−1 S for Hannuoba, the Massif Central, Kerguelen and the deformed East African peridotitic xenoliths: Lorand and Alard 2001; Lorand et al. 2003b, 2004; Becker et al. 2006; Fischer-Gödde et al. 2011). Of course, these estimates of metasomatic Type 2 BMS additions are dependent on the HSE concentrations in the Type 2 BMS. Even when choosing a HSE-poorer Type 2 BMS component similar to the AR10f BMS of Alard et al. (2000), the amount of metasomatic BMS needed to reproduce the whole-rock peridotite signatures is still realistic (max 0.055 wt% sulfides or 200 μg.g−1 S). Addition of Type 2 BMS will also translate into higher Se contents and higher Cu contents, and a S/Se ratio trending toward the chondritic value (Lorand et al. 2003b; Alard et al. 2011).
The whole-rock 187Os/188Os signatures can be similarly modelled, assuming various 187Os/188Os compositions for the Type 2 BMS component (Fig. 13). One model assumes the Type 2 BMS component to have a radiogenic 187Os/188Os composition (0.14), whereas the second considers that the Type 2 BMS component precipitates from a mantle-derived silicate melt (187Os/188Os = 0.1282), as deduced from the 187Os/188Os vs. 187Re/188Os variation of the metasomatic Type 2 BMS in the SE Australian GamVL 11 peridotite (see above). In any case, both these 187Os/188Os-isotopic compositions are typical of OIB-like (see Widom et al. 1999; Day 2013 and refs. therein) or MORB-like melts (see Escrig et al. 2005; Gannoun et al. 2007). As seen for the whole-rock HSE patterns, the whole-rock 187Os/188Os of the peridotites can be explained by addition of 0.005–0.02 wt% Type 2 BMS, whose initial 187Os-isotopic composition is in equilibrium with typical mantle derived melts. This isotopic signature reflects the exsolution of the Type 2 BMS component from evolved small melt fractions or primary silicate melt fractions, which are both S-saturated (Fig. 13).
Timing of Refertilization/Metasomatism by S-saturated silicate melts
The preceding discussion provides strong arguments for melt/rock reaction processes, based on textural and mineralogical relationships as well as HSE systematics. This is complemented by recent studies of massif peridotites (Le Roux et al. 2007), which unlike peridotite xenoliths have the advantage of allowing 3-D structural relationships between rock types to be observed directly. The observed field relationships provide further evidence of the importance of refertilization processes in creating the range of existing lithologies and geochemical variations. An often unstated assumption is that refertilization occurred recently, in a geologic sense, or at least that it was unrelated to the melt extraction process, postdating it by hundreds of millions of years. For example, Le Roux et al. (2007) argue that the lherzolites observed in the Lherz massif formed by the refertilization of ancient harzburgites, and suggest that the linear correlation between 187Os/188Os and Al2O3 observed in Lherz and other eastern Pyrenean massifs (Reisberg and Lorand 1995) may result from this process. This suggestion has also been made to explain similar correlations in peridotite-xenolith suites (e.g., Pearson et al. 2004). In this model, melts or melt-derived cumulate phases bearing both Al2O3 and radiogenic 187Os/188Os are added to a highly depleted precursor peridotite with an unradiogenic 187Os/188Os isotope composition. However the details of how this mechanism might operate remain unclear. A central but sometimes unstated requirement of this hypothesis is that partial melting leaves only harzburgites, which constitute the highly depleted endmember of the mixing trend. If true, this scenario has profound consequences for models of the formation of subcontinental lithosphere, so resolution of this issue has implications well beyond the field of HSE geochemistry.
Though the evidence for refertilization is strong, the assumption that this process greatly postdated melt extraction raises some serious complications, mostly linked to the fact that this requires coupled mobility of Os and Al. This seems unlikely, given the highly contrasting geochemistry of these two elements. If the correlation between 187Os/188Os and Al2O3 is caused by the addition of radiogenic BMS together with metasomatic clinopyroxene and spinel, a strong positive correlation should also exist between Os concentration and Al2O3. For example, mass balance calculation shows that to obtain a PUM-like 187Os/188Os ratio of ~0.13 in fertile lherzolite, starting with a harzburgite with a 187Os/188Os ratio of 0.12 and adding BMS with a 187Os/188Os ratio of 0.14, it would be necessary to double the Os concentration of the rock, i.e., if the harzburgite contained 3 ng.g−1 Os it would be necessary to add another 3 ng.g−1 of Os with 187Os/188Os of 0.14. This calculation is completely independent of the concentration of Os in the added BMS or the details of how they are introduced. If the added BMS had a 187Os/188Os ratio of 0.135, the Os concentration would have to be tripled from the harzburgite to the fertile lherzolite. Such strong positive correlations between Os concentration and Al2O3 are never observed in peridotite xenoliths, with the possible exception of the suite from the Middle Atlas, Morocco studied by Wittig et al. (2010). Instead, scattered, flat, or slightly negative correlations, the latter being the expected result of partial melting, are observed between Os, or its sister element Ir, and Al2O3 (Fig. 4). This is observed in the Hannuoba suite (Liu et al. 2010), cited above as a possible example of refertilization.
If late refertilization is assumed to result from melt-residue mixing without involvement of metasomatic BMS, the coupled mobility of Os and Al implicit in this model still poses a problem. Namely, the enriched endmember must have a remarkably high Os concentration in order to produce a linear mixing trend passing through the PUM composition on a 187Os/188Os vs. Al2O3 diagram. This point is illustrated in Figure 14a. As Os is a highly compatible element during melting, mantle melts typically have much lower Os concentrations than residual peridotites (Shirey and Walker 1998), and so define highly curved mixing trajectories. Improbably elevated Os concentrations in the melt phase (8 ng.g−1 in the example shown in Fig. 14a), and/or extremely radiogenic compositions (187Os/188Os = 0.175 in the example shown) are required to match the observed trends (other parameters given in figure caption). Given the frequent occurrence of these trends, melts with such odd compositions would have to be nearly ubiquitous and highly reproducible, despite the fact that they are almost never observed. Similar observations were made by Rudnick and Walker (2009), who also raised the question of why the trends rarely extend beyond PUM. On the other hand, it has been suggested that mixing does not occur with the melts themselves, but rather with cumulate phases (clinopyroxene, spinel, and BMS) left by the passing melts. However, this would only exacerbate the problem, as sulfide deposition would rapidly strip the melt of Os, causing the trend to deviate further from the required linear trajectory. This is shown in Figure 14b, which models the effect of the progressive crystallization of a clinopyroxene, spinel, and BMS assemblage from a basaltic melt (see figure caption for parameters). The Os/Al2O3 value of the melt drops to < 1% of its original value after only <5% crystallization, even if a rather conservative choice (Dsulf.liq/sil.liq. = 5 × 104) is made for the Os partition coefficient. The much higher D values (~106) obtained in the recent study of Mungall and Brenan (2014) would cause Os to be stripped even more rapidly. For all of these reasons, it seems unlikely that refertilization involving addition of radiogenic Os can explain the observed 187Os/188Os vs. Al2O3 trends. Nevertheless, it is now known that Os mobility during melt percolation does indeed occur, as described in the previous sections. However, it seems more likely that this process would perturb rather than create correlations between Os isotopes and major element contents.
One possible way to reconcile the evidence for refertilization processes with the arguments against coupled introduction of radiogenic Os and Al2O3 is to suggest that refertilization did indeed occur, but that is was contemporaneous with, or occurred shortly after the melt extraction event. This process is described as “auto-refertilization” by Rudnick and Walker (2009), who argue that it is inherent in the melt extraction process, as harzburgites produced by large extents of melting towards the top of a melting column are traversed by melts derived from greater depths. According to this hypothesis, Re rather than Os is added in a coupled manner with Al2O3 and other magmaphile elements. This could occur either through the direct addition of melt trapped during percolation, or through the addition of BMS plus clinopyroxene and spinel precipitated from the percolating melt. This would produce correlations between 187Re/188Os and Al2O3 and other melt extraction indices, which in time would grow into correlations with 187Os/188Os. This suggestion is intrinsically simpler than the introduction of radiogenic Os because, unlike Os, Re is known to behave as a moderately incompatible element during magmatic processes. As discussed earlier, this is due largely to the less chalcophile nature of Re at the oxygen and sulfur fugacity conditions relevant to mantle melting (Brenan 2008). Figure 14b shows the variation of the Re/Al2O3 ratio of the melt during progressive crystallization of the clinopyroxene, spinel and BMS phase assemblage mentioned above, assuming Dsulf.liq/sil.liq. = 50 (Fonseca et al. 2007). It can be seen that this ratio remains fairly constant, implying that crystallization of such a phase assemblage from a percolating melt could indeed produce the required, and observed, correlated enrichment of Re and Al2O3. Furthermore, intergranular BMS left by such percolating melts would be expected to have higher 187Re/188Os ratios than enclosed BMS, and so with time would be expected to develop higher 187Os/188Os ratios. Though this model may not be applicable in all cases, it is appealing as it combines many of the key points of the melt extraction and refertilization models.
Metasomatism involving highly oxidizing fluids and slab-derived melts
In addition to silicate melts, fluids have often been described as potential metasomatizing agents of the lithospheric mantle in tectonic environments ranging from stable Archean cratons to destructive margins. Such fluids may be related to the dehydration of the subducting plate in subduction zones or may result from fluid/melt immiscibility during melt percolation and melt/rock reaction in intraplate contexts. In any case, fluids have been considered to play an important role in the transfer of metals and are generally proposed as an efficient transport agent in formation models of ore deposits (Stein 2014).
Supra-subduction-zone peridotite xenoliths have been sampled by volcanism associated with subduction of young (15–25 Myr), hot oceanic crust such as in the northern Kamchatka volcanic front (Kepezinhas et al. 2002; Widom et al. 2003), and the US and Canadian Cordillera (Brandon et al. 1999; Peslier et al. 2000a,b; Lee et al. 2000; Lee 2002) but also with subduction of old (~100–150 Myr), cold oceanic crust such as in Papua New Guinea (Lihir Island, McInnes et al. 1999), in the Japan volcanic front (Brandon et al. 1996) and the southern Kamchatka volcanic arc (Kepezhinskas et al. 2002; Widom et al. 2003; Saha et al. 2005). All of these, as well as peridotite xenoliths from the Kerguelen oceanic island (Lorand et al. 2004; Delpech et al. 2012) and the Montferrier locality in the Massif Central (Alard et al. 2011), display HSE systematics different from those reflecting partial melting or metasomatism with silicate melts. At least two groups of HSE patterns can be distinguished (Figs. 15 and 16).
Xenoliths from Montferrier and Kerguelen, as well as those associated with hot young subduction (Valovayam volcano in the northern Kamchatka arc: Kepezhinskas et al. 2002 and Big Creek volcano in the Southern US Cordillera, Sierra Nevada: Lee 2002), are characterized by saw-tooth CI-chondrite normalized HSE patterns with positive Os and Pd anomalies, but not necessarily high absolute Pd concentrations, as Pd concentrations are mainly below PUM estimates (except for Sierra Nevada 1026V with 96 ng.g−1 Pd: Lee 2002; and Kerguelen MM-94-54 with 13.2 ng.g−1 Pd: Lorand et al. 2004) (Fig. 15). Osmium concentrations are generally in the ng.g−1 concentration level (2.3–7 ng.g−1) in the Kerguelen (apart for MM-94-54), Montferrier, Sierra Nevada (Big Creek and Oak Creek) and northern Kamchatka peridotites (see Fig. 15). Xenoliths derived from the subduction settings of the northwestern US and Canadian Cordillera (Brandon et al. 1996; Peslier et al. 2000a), for which only Re–Os systematics were studied, contrast from those of the Kamchatka and Sierra Nevada by their Os contents extending toward much lower values (0.050–4.47 ng.g−1).
Subtle contrasts in the HSE patterns appear when specific localities are singled out. The Montferrier xenoliths (Alard et al. 2011) show the smoothest HSE patterns with overall PUM-like HSE concentrations but slight to moderately suprachondritic OsN/IrN and PdN/PtN (1.04–1.81 and 1.22–1.81, respectively) (see Fig. 15). No systematic behavior of Re with Pd or Os is observed. The Kerguelen peridotites are characterized by overall lower HSE contents, but with pronounced Os and Pd enrichments (OsN/IrN = 1.35–2.63 and PdN/PtN = 2.24–5.37). This shift toward lower HSE concentrations, especially evident for the MM-94-54 peridotite reflects the superposition of S-undersaturated silicate melt and fluid percolations. The Kerguelen BMS show suprachondritic OsN/IrN and PdN/PtN ratios and HSE patterns perfectly mimicking the whole-rock signatures, independent of their mineralogy (Po–Pn or Pn–Cpy) (Delpech et al. 2012). Among the subduction-zone-related xenoliths for which the full HSE data are available (Kepezhinskas et al. 2002; Lee 2002), only one peridotite from the Sierra Nevada, but both peridotites collected in the Valovayam volcano in the northern Kamchatka volcanic front, show similar HSE patterns to the xenoliths from Montferrier and Kerguelen. The Valovayam xenoliths have the most supra-chondritic OsN/IrN (4.9 and 6.1) but variable PdN/PtN, including both a sub- and a supra-chondritic value (0.77 and 1.92). They also seem to point toward a Au-enrichment concomitant with those of Os and Pd. The Sierra Nevada peridotite displays chondritic OsN/IrN (0.98) and suprachondritic PdN/PtN (2.9), but also suprachondritic PtN/IrN in contrast to all the other xenoliths.
The Kerguelen peridotites and all those associated with young hot subduction zones show 187Os/188Os values ranging from PUM-like to extremely radiogenic compositions (0.1274 to 0.1755). Duplicate analyzes vary widely (0.1755 vs. 0.1360 for Pg-6 peridotite and 0.1523 vs. 0.1275 for Pg-43; Alard et al. 2011) in the Montferrier locality, demonstrating that the Os carrier phases are multiple and heterogeneously distributed. In this study, smaller aliquots (1–2 g) were used for the Os-isotopic analyzes than for the HSE analyzes, thus the nugget effect may explain the greater variability of the 187Os/188Os data (see Meisel and Horan 2016, this volume). Additionally, for Pg-43, the positive relationship between Os concentration and 187Os/188Os and the negative co-variation with 187Re/188Os suggest that the Os-rich host phase contains radiogenic 187Os but is not particularly rich in Re. This contrasts with the earlier evidence for overprinting of whole-rock 187Os signatures provided by the positive correlation of 187Os/188Os with 1/Os in the Canadian Cordillera harzburgitic xenoliths (Peslier et al. 2000a). On this basis, these authors argued that the mantle wedge under Canada was overprinted by a metasomatizing agent characterized by a low Os content and a radiogenic Os-isotope signature.
The Kerguelen and Montferrier peridotitic xenoliths also differ from the peridotites clearly overprinted by S-undersaturated and S-saturated silicate melts when the chalcogens (S, Se, Cu) are considered. The percolation of S-undersaturated melt strips the BMS originally present, and thus results in low BMS modal abundances and extremely low whole-rock S, Se and Cu contents. By contrast, low melt/rock ratios lead to precipitation of Type 2 BMS, made up of pentlandite and Cu-rich sulfides, which result in relatively high bulk rock S, Se, Cu contents, but chondritic S/Se ratios. In contrast, the Kerguelen and Montferrier peridotites that experienced percolation of fluids are characterized by very high whole-rock S contents, suprachondritic S/Se (up to 11,000, i.e., approximately 4 times the chondritic values) and moderate Cu contents (Lorand et al. 2004; Alard et al. 2011). This type of metasomatism was originally labelled COPPS (Cu–Os–Pt–Pd–S; Lee 2002). The combination of the coupled Os and Pd enrichments with the systematic suprachondritic OsN/IrN ratios cannot be explained by magmatic processes owing to 1) the opposing partitioning behavior of Pd and Os between alloys and sulfide melt or MSS and Cu–Ni-sulfide melt, and 2) the similar partitioning behavior of Os and Ir between sulfide melt and silicate and between MSS and Cu–Ni-sulfide melt. In contrast, the suprachondritic S/Se ratios typically observed in hydrothermal sulfide deposits (e.g., Thierault and Barnes 1998; Luguet et al. 2004; Lorand and Alard 2010) provide support for metasomatic overprinting related to fluid percolation. Thermodynamic calculations (Wood 1987) and experimental work (Fleet and Wu 1995) performed at magmatic temperatures (> 1000 °C) under oxygen fugacity conditions near FMQ point to the highly volatile character of Os oxides and of Pd chlorides and indicate the moderately volatile nature of Ru possibly in both oxide and chloride form. The volatile character of these HSE increases at higher fO2, suggesting that oxidizing Cl-bearing slab fluids can efficiently transport Os and Pd, in contrast to Ir and Pt (Wood 1987; Xiong and Wood 2000). In the Kerguelen peridotites, Lorand et al. (2004) reported the association of minute BMS blebs associated with CO2-rich fluid inclusions, all forming continuous trails crosscutting the silicate grains. They also noted the absence of liquid–liquid immiscibility features, which were routinely observed for BMS blebs included in carbonatitic silicate melt pockets and precipitated during the percolation of volatile-rich small melt fractions (see above). All of these observations provide strong support for the transport of HSE by a fluid or vapor phase in the Kerguelen peridotites, although this fluid or vapor probably resulted from unmixing from an evolved silicate melt during the final stages of reactive porous flow percolation (Grégoire et al. 2000). In the subduction environment of northern Kamchatka, Widom et al. (2003) attributed the whole-rock Os and Re concentrations and the radiogenic 187Os signatures to the interaction of the mantle wedge with highly oxidizing slab melts, likely adakitic in nature.
The whole-rock radiogenic Os isotope signatures in Kerguelen and in the xenoliths associated with young hot subduction demonstrate that the source of these fluids possesses long-term high Re/Os ratios. In subduction environments, the best candidate for a reservoir with a high Re/Os ratio and a time-integrated, radiogenic Os-isotope signature is the subducting oceanic plate. While this seems perfectly logical for peridotite xenoliths associated with arc volcanism (e.g., Brandon et al. 1996), involvement of a subduction-derived fluid in the Montferrier xenoliths is not supported by their Nd–Sr–Pb isotopic compositions (Dautria et al. 2010). Alternatively carbonatitic melts have been suggested to have affected the Montferrier xenoliths on the basis of their whole-rock high field-strength element (HFSE: Nb, Ta, Hf, Zr) signatures. Carbonatitic melts, owing to their very radiogenic 187Os/188Os ratios, which range up to 9.31 ± 0.21 (Pearson et al. 1995; Blusztajn and Hegner 2002; Escrig et al. 2005) could thus account for the radiogenic Os-isotopic signatures reported in the Montferrier peridotite xenoliths.
This group consists exclusively of subduction-related xenoliths associated with destruction of cold old oceanic plates. These are sampled in the southern Kamchatka volcanic front (in the Avachinski volcano: Kepezhinskas et al. 2002; Widom et al. 2003; Saha et al. 2005; in the Bakering volcano: Widom et al. 2003; and in the Shiveluch and Karchinski volcanoes: Saha et al. 2005), the Japan volcanic arc (Brandon et al. 1996) and on Lihir Island, Papua New Guinea (McInnes et al. 1999). The Avachinski peridotites (southern Kamchatka) show a large range of HSE patterns. Half of the samples display pronounced enrichments in Ru and Pt, and minor enrichments in Rh, while the Os and Pd signatures appear as relative negative anomalies (Fig. 16A). These patterns broadly mirror the ones obtained in the Volovayam volcano in the northern Kamchatka volcanic arc that exhibit Os, Pd and Ru–Rh enrichments over Ir and Pt. Intermediate HSE patterns between the Os–Pd + Rh ± Ru enriched (Valovayam volcano in the northern Kamchatka volcanic arc) and the Ru–Pt ± Rh enriched (southern Kamchatka) patterns are also observed. The Lihir Island peridotite HSE patterns are characterized by Os–Rh–Pt–Au-enrichments (Fig. 16B). Their whole-rock 187Os/188Os compositions vary from unradiogenic to slightly radiogenic relative to PUM (0.12170–0.13348). The 187Os/188Os range in the Avachinski xenoliths is narrower (0.1275–0.1299). Ultimately, peridotites from Ichinomegata (Japan; Brandon et al. 1996) and the Shiveluk volcanoes in the northern Kamchatka volcanic front (Saha et al. 2005) show a similar range of 187Os/188Os compositions. Neither these southern Kamchatka nor Lihir Island peridotites show 187Os/188Os compositions as radiogenic as in the northern Kamchatka or Montferrier xenoliths. The picture becomes more complicated when xenoliths from the Kharchinski and Bakering volcanoes are considered. The peridotites from the Kharchinski volcano in southern Kamchatka have similar 187Os/188Os compositions to those of the nearby Shiveluk and Avachinski arc volcanoes but extend to radiogenic values (0.1265–0.1585; Saha et al. 2005). At the end of this trend sits the Bakering volcano, which is located behind the arc front in southern Kamchatka, with exclusively radiogenic 187Os signatures (0.1352–0.1566; Widom et al. 2003).
Overall, in southern Kamchatka, 187Os/188Os signatures vary perpendicular to the subduction zone, evolving from unradiogenic to radiogenic with increasing depths of the subducting slab, likely highlighting the varying nature of the metasomatic agents (Widom et al. 2003). Under the Avachinski volcano, the mantle wedge has likely experienced metasomatism by hydrous slab fluids while the mantle wedge under the Bakering volcano, located further from the subduction front, where the subducted plate is deeper (70–100 km), is inferred to have reacted with slab- and sediment-derived carbonated melts. On the same basis, Widom et al. (2003) attributed the similarity of the 187Os/188Os ratios observed in Lihir Island, the Japan arc, and the Avachinski volcano to the existence of similar metasomatic processes, and by extension similar metasomatising agents, that affect the mantle wedge in a uniform manner above relatively cold, old (100–150 Ma) subducting slabs at slab depths of ~100–150 km. The variability of the HSE patterns in the Avachinski and Lihir volcanos nevertheless presents a more complicated picture possibly not entirely consistent with the conclusions of Widom et al. (2003). The variable HSE signatures, whose origin is not yet properly understood, indicate the variability of the overprinting processes and/or metasomatizing agents possibly due to the non-uniform nature of the subducted material (slab + sediments) and the geometry of the subduction zone.
Syn- to post-eruptive processes
A striking feature of about half of the non-cratonic basalt-hosted spinel-peridotite xenoliths is their subchondritic OsN/IrN compared to those of orogenic, ophiolitic, and abyssal peridotites, as well as cratonic kimberlite-borne peridotite xenoliths (Table 1). Considering the similarity of the Os and Ir partition coefficients between sulfide liquids/silicate melts, MSS/Cu–Ni-rich sulfide melt and alloy/MSS, this Os/Ir fractionation cannot be attributed to magmatic processes and may instead reflect near-surface processes. Alkali-basalt-hosted xenoliths also have low sulfur contents (Lorand 1990; Ionov et al. 1992), undoubtedly due in part to late stage oxidative weathering of the BMS and their pseudomorphosis by iron-oxyhydroxides. This replacement is on average more pronounced in breccia-hosted xenoliths compared to lava-hosted xenoliths (Luguet and Lorand 1998), suggesting that porosity of the basaltic host rocks was a factor in controlling the extent of the S loss. This mineralogical replacement led to the mobilisation of S as sulfates due to the infiltration of meteoric water (Lorand 1990). While Dreibus et al. (1995) suggested that Se and Cu are less easily oxidized or their oxide compounds are immobile, making the abundances of Se and Cu insensitive to supergene weathering, Harvey et al. (2015b) recently challenged this view for Se, as Se appears to be mobilised along with S during supergene weathering in the Kilbourne Hole xenoliths.
The positive Ir/Os vs. Cu/S trend obtained in xenoliths from SE Australia was originally taken to imply coupled Os and S losses (Handler et al. 1999). Consequently, one could argue that Os was lost from the peridotite xenoliths during the supergene weathering of the BMS and leached from the iron oxyhydroxide pseudomorphs along with sulfates. However, this explanation may be challenged as 1) low Os concentrations and subchondritic OsN/IrN ratios are sometimes obtained for mantle xenoliths whose BMS are free of iron oxyhydroxide replacement (Ackerman et al. 2009) and 2) some xenoliths with extensive supergene weathering of the BMS show PUM-like Os contents and chondritic OsN/IrN ratios (Alard et al. 2011). Hence, supergene weathering appears unlikely to be the major factor controlling the Os–Ir systematics.
Alternatively, the high temperature and high fO2 experienced by the xenoliths upon their entrainment by the basaltic lavas may trigger simultaneous losses of S and Os, as long as interstitial BMS are present in the mantle peridotite (Handler et al. 1999). Under these conditions, volatile loss of Os as OsO4 and immobility of Ir (Wood 1987) could result in the Os/Ir fractionations observed in basalt-borne peridotitic xenoliths. However, the solubility calculations of Wood (1987) predict similar volatility of Os and Ru, but no sub-chondritic RuN/IrN values are observed in mantle xenoliths, rather the contrary. Assuming volatile loss of OsO4 is indeed a viable explanation for the low OsN/IrN ratios in basalt-borne peridotite xenoliths, the absence of sub-chondritic OsN/IrN ratios in cratonic kimberlite hosted xenoliths (Pearson et al. 2004) may be attributed to the less-oxidizing conditions upon kimberlite eruption, or to the fact that Os is hosted in a more refractory host phase than BMS. As cratonic peridotites have experienced extremely high degrees of partial melting that likely exhausted all the BMS from the residue, stabilization of Os along with Ir and Ru in high-temperature HSE alloys could insure the non-volatization of Os upon kimberlite eruption (see Luguet et al. 2015; Wainwright et al. 2015).
Additional explanations for the low S contents of these xenoliths have been proposed (Ionov et al. 1992), including extensive percolation of S-undersaturated melts (Lorand and Alard 2001; Lorand et al. 2003a). Recently, Ackerman et al. (2009) suggested that during such reactions at high melt/rock ratios, the removal of IPGE bearing MSS would explain the subchondritic Os/Ir ratios in metasomatized peridotites. If correct, one could thus expect a correlation between the OsN/IrN ratio and the extent of the BMS removal (e.g., concentration in Ir or Ru). However, Kerguelen peridotite xenoliths, which show variable overprinting by large fractions of S-undersaturated melts, have chondritic OsN/IrN regardless of the extent of BMS removal.
The origin of the sub-chondritic OsN/IrN thus remains enigmatic and requires more detailed and systematic studies integrating petrographic characterisation of the BMS with whole-rock and BMS studies of the HSE, 187Os/188Os, S, and Se signatures.
CHRONOLOGICAL INTERPRETATION OF Os-ISOTOPIC DATA AND TECTONIC IMPLICATIONS
One of the primary advantages of the Re–Os-isotopic system is that it offers the most reliable means of dating magmatic events in mantle peridotites. This is because of the compatible nature of Os during mantle melting, which results in high Os concentrations in peridotites relative to those of most melts and metasomatic fluids. As the preceding discussion shows, Os is now known to be much more mobile than originally believed, so its resistance to perturbation during metasomatic and melt percolation events is only relative. Nevertheless, in many cases Os isotopes can provide information about the timing of mantle melting events that can only rarely be obtained from Sr, Nd, or even Hf isotopes, though the latter sometimes yield useful age data (e.g., Bizimis et al. 2007; Liu et al. 2012a,b).
Obtaining age information from Os isotopes of whole rocks
It is very unusual to obtain robust Re–Os isochrons or even well-constrained errorchrons from peridotite massif/xenolith suites, either among the whole rocks of a given suite or at the scale of the HSE host phases from a given sample. This is in part because the Re–Os system may be subject to open system behavior, but also because the different rocks of a peridotite suite and even their constituent HSE host phases may have experienced different histories. For this reason, model ages, which estimate the moment when a sample’s isotopic evolution began to deviate from that of an assumed mantle evolution curve, are often used, as they allow the possibly different times of melting experienced by the different samples to be estimated. The model age concept has long been applied to incompatible lithophile element based isotopic systems (e.g., Sm–Nd and Lu–Hf), and a more complete description of its application to the Re–Os system can be found in Shirey and Walker (1998) and Harvey et al. (2016, this volume). We stress that for all isotopic systems, the geologic reality of a model age depends on several factors: 1) the suitability of the model, in the case of the Re–Os isotope system, the assumption that the peridotite is the residue of a single melt extraction event from a source with an isotopic composition on the mantle evolution curve; 2) the unproven assumption that melting was profound enough to allow the peridotite to fully equilibrate isotopically with the surrounding rocks, thus erasing any small scale heterogeneity that may have existed; 3) the choice of the mantle evolution curve used for the calculation.
This latter point merits particular discussion. While it is generally acknowledged that the 187Os/188Os ratio of the mantle followed a roughly chondritic evolution curve, chondritic Re–Os values span a certain range (Walker et al. 2002a; Day et al. 2016, this volume), and the appropriate values to use are subject to debate. Furthermore, as discussed above, the evolution of the upper convecting mantle from which peridotites are derived may deviate from a purely chondritic path. The resulting ambiguity becomes progressively worse as the sample age decreases and the various possible evolution curves diverge, with the consequent uncertainty reaching more than 500 million years for samples less than a billion years old (Fig. 17). The choice of the evolution curve is also critical to the calculation of the commonly used γOs(t) parameter, the percentage difference of the initial 187Os/188Os ratio of a sample from the contemporaneous 187Os/188Os ratio of the “chondritic” reference curve [γOs(t) = 100 × (187Os/188Ossample(t)/187Os/188Oschondrite(t) - 1)]. As not all authors use the same reference curve parameters, caution is needed when comparing γOs(t) values from different studies, and these values should not be taken too literally, e.g., a sample with a γOs(t) value of +1 is not necessarily “enriched” relative to the true but unknown average mantle Os-isotopic composition. The most commonly used evolution curve parameters (187Os/188Os = 0.1270, 187Re/188Os = 0.40186; Walker et al. 1989; Shirey and Walker 1998) are based on typical carbonaceous chondrite values, but other evolution curves, notably that of the inferred PUM reservoir (187Os/188Os = 0.1296, 187Re/188Os = 0.435; Meisel et al. 2001) are also sometimes employed. On the other hand, uncertainties on the 187Re decay constant (λ187Re) have only a very minor impact on the evolution curve, with the preferred values of Smoliar et al. (1996) (λ187Re = 1.666 × 10−11 yr−1, based on meteorite analyzes) and of Selby et al. (2007) (λ187Re = 1.6668 × 10−11 yr−1 based on magmatic ores) yielding 187Os/188Os compositions that differ by only 0.04%, after 4.567 Gyr of radiogenic ingrowth, all other parameters being equal.
Several types of Re–Os model ages are defined. The first, TMA, are analogous to the model ages used for the Sm–Nd and Lu–Hf systems. TMA ages are calculated by using the measured 187Re/188Os ratio to extrapolate the sample’s 187Os/188Os ratio back in time until its intersection with the assumed mantle evolution curve. However, Re is known to be a fairly mobile element, so the measured 187Re/188Os ratio of a sample may have been modified since the original partial melting event. For that reason, Walker et al. (1989) defined the “Re depletion age” (TRD). TRD is calculated by directly comparing the measured 187Os/188Os ratio of the sample with the mantle evolution curve (which is equivalent to assigning a 187Re/188Os value of zero to the sample). For mantle xenoliths, the TRD age concept is often refined (TRD eruption), by taking into account the radiogenic ingrowth of 187Os that has occurred since magmatic emplacement of the xenoliths at the surface. In this case it is assumed that all of the Re in the xenolith was added by the host magmas, and the measured 187Re/188Os value of the sample is used to extrapolate its 187Os/188Os ratio back in time to the eruption age (Pearson et al. 1995). This ratio is then compared with the mantle evolution curve to obtain the TRD at the time of eruption. All TRD ages are by definition minimum estimates of the age of melt extraction. For highly depleted samples such as harzburgites that are expected to contain very little Re after melt removal, TRD ages are likely to approach the true age of melt extraction, and for this reason the TRD concept has been applied quite successfully to the study of cratonic xenoliths (e.g., Walker et al. 1989; Pearson et al. 1995). However, lherzolites, which are quite common in non-cratonic terrains, have experienced lesser degrees of melt extraction and are thus expected to contain non-negligeable Re contents. Therefore TRD ages obtained from lherzolites almost inevitably underestimate true ages of melt extraction by hundreds of millions of years. For example, a lherzolite with a TRD age of 0.5 Ga cannot be viewed as younger than a depleted harzburgite with a TRD age of 1.5 Ga, yet this type of error is often made. Thus while a TRD age provides a firm minimum estimate for the time of melt extraction in all cases, in lherzolites such ages should never be equated with the true age of melting. It may even be advisable to avoid providing TRD ages for any samples except harzburgites, as they are often misinterpreted. The frequently presented graph showing Re–Os TRD model ages vs. Al2O3 (as an alternate axis to 187Os/188Os) may also be misleading, especially to non-specialists.
Nonetheless, if limited to harzburgites the TRD concept can be quite useful. The oldest TRD age in a suite of harzburgites may provide a good estimate of the approximate time of melt extraction, with younger harzburgite TRD ages possibly representing samples that have been affected by metasomatism or refertilization. An alternative but essentially equivalent method of estimating the time of melt extraction is the isochron analog (sometimes referred to as the “aluminachron”) method (Reisberg and Lorand 1995). In this case, 187Os/188Os is plotted against Al2O3 or another index of melt extraction, which is expected to vary systematically with 187Re/188Os. If a reasonably well-defined correlation exists, it is possible to estimate the 187Os/188Os ratio at the value of the proxy corresponding to a 187Re/188Os value of zero (i.e., where all of the Re has been removed from the residue by partial melting). For Al2O3, Handler et al. (1997) proposed that this value was about 0.7 wt%, at least in their sample suite. The thus estimated 187Os/188Os ratio is analogous to the initial ratio of the true 187Re/188Os isochron that would have existed if the Re concentrations had not been perturbed. It can therefore be compared to the mantle evolution curve to obtain a model age of melt extraction. The age obtained from such a correlation is independent of whether the correlation itself is viewed as the product of radioactive decay after an event leaving variably depleted residues, or as the result of recent refertilization of a highly depleted residue. In both cases it would represent the age of the melt extraction event. Use of the isochron analog method has the advantage that it can be applied to suites where harzburgitic samples are lacking or where their Os-isotopic compositions have been modified by metasomatic processes (e.g., the Canadian Cordillera; Peslier et al. 2000a,b). It also has the advantage that many samples contribute to the definition of the correlation line, and thus the 187Os/188Os value of the depleted endmember used to define the age of melting may be better constrained than if TRD ages of harzburgites are used directly, i.e., the estimated age is less affected by aberrant values from one or two odd samples. On the other hand, this method is clearly limited to cases where a correlation can be well defined. Also, this technique implicitly assumes that a single melting event is being dated, and thus is not applicable to cases where a rough correlation results from the tectonic juxtaposition of unrelated domains.
Obtaining age information from Os isotopes of Base Metal Sulfides
This subject is discussed in detail in Harvey et al. (2016, this volume) and therefore only summarized, for completeness, here. The concepts of TMA and TRD Os model ages defined for whole rocks can also be applied to individual BMS grains, analyzed either in bulk after chemical separation of Re and Os (Burton et al. 1999; Harvey et al. 2010, 2011), or in situ, by laser ablation ICPMS. The caveats discussed above relative to the use of model ages of course apply to BMS as well as whole rocks. BMS from individual peridotites typically display a wide range of 187Os/188Os ratios (e.g., Alard et al. 2002; Harvey et al. 2011), and chronological interpretation of these values can be challenging. One of the most important tasks is selecting the BMS likely to yield meaningful age data. As noted by Griffin et al. (2002), BMS with elevated 187Re/188Os values (> 0.07 according to their modelling) may have suffered metasomatic enrichment of Re and possibly Os, and so their TMA ages are unlikely to be meaningful. Wang et al. (2009) argue that the TRD ages of BMS with elevated 187Re/188Os ratios may still be reliable if these BMS define horizontal arrays on 187Os/188Os vs. 187Re/188Os plots, as this might suggest that only Re, but not radiogenic Os, had been recently added. In any case, it should be noted that in situ Os-isotopic measurements of BMS with high 187Re/188Os ratios are intrinsically delicate, because of the extremely large corrections that must be applied to correct for the isobaric interference of 187Re on 187Os.
Even after exclusion of data from BMS that may have been modified by late metasomatic processes, interpretation of Os-isotopic data from BMS poses many challenges. Enclosed BMS, which tend to have high Os contents and unradiogenic Os-isotopic compositions, are often thought to represent residues of ancient melting events. As is also true of residual whole rocks, the model ages of such residual BMS only have chronological meaning if the melting event was sufficiently intense to homogenize the Os-isotopic composition of the source. Interstitial BMS generally (but not always; see González-Jiménez et al. 2014; Wainwright et al. 2015) have more radiogenic compositions, and are usually interpreted to have been left by later metasomatic or refertilization events. In this case, the TRD ages of the BMS are meaningful only if the Os-isotopic compositions of the percolating melts had 187Os/188Os ratios equivalent to the value of the contemporaneous mantle evolution curve.
It is often assumed that the wide range of Os-isotopic compositions found among BMS from a given peridotite results from the introduction of several generations of BMS (e.g., Wang et al. 2009; González-Jiménez et al. 2013). However, a part of this variation could simply result from radiogenic ingrowth, coupled with variable extents of diffusive re-equilibration with the surrounding BMS and silicate phases. In this case, the TRD age obtained from a given BMS might not have any genetic significance. This complexity is illustrated by considering the results of Burton et al. (1999). In a lherzolite xenolith from Kilbourne Hole, these authors found that intergranular BMS and silicate phases had equilibrated isotopically, while the enclosed BMS were less radiogenic. In such a case, the TRD ages of the intergranular BMS would have no chronological significance. The enclosed BMS could potentially indicate the age of melt depletion, since the surrounding silicate material with extremely low Os concentrations serves as a barrier against diffusion (see discussion above in “reconciling 187Os/188Os results from whole-rock and base metal sulfides” and Burton et al. 1999). Nevertheless, some degree of equilibration with the surrounding silicate cannot be excluded, which would modify the 187Os/188Os ratio and thus the model age. More evidence of the possible importance of this process is provided by the silicate Re–Os data of samples from the Massif Central (Harvey et al. 2010) and Kilbourne Hole (Harvey et al. 2011). Unlike in the Burton et al. (1999) study, these authors did not find intermineral isotopic equilibrium among the silicate phases, or between the silicates and the intergranular BMS. However the 187Os/188Os ratios of the silicate phases in both localities were much less radiogenic than would be expected (nearly all < 0.15), given the extremely high 187Re/188Os ratios of the silicate phases and the > 1 Ga lithospheric residence times suggested by the whole-rock and some of the BMS TRD ages. This might suggest that much of the 187Os produced by radioactive decay in the silicates diffused into the BMS phases, thus altering their 187Os/188Os ratios. On the other hand, it is possible that some of the Re in the silicates was recently introduced, and thus did not have any effect on the 187Os/188Os ratios of the BMS. While it is not obvious which of these processes is more likely (and both may occur in the same sample), this issue should be considered when interpreting BMS model ages.
For all of these reasons, obtaining meaningful chronological information from individual BMS in peridotite xenoliths is no simpler than obtaining reliable age data from whole-rock Os analyzes. Nevertheless, in some cases, enclosed, residual BMS in olivines may provide evidence of ancient melting events not detected in the whole-rock data. Furthermore, probability density plots of TRD ages obtained from in situ laser ablation analyzes of BMS from a given locality yield age peaks that can often be correlated with known tectonic events (González-Jiménez et al. 2013, 2014; Wang et al. 2009). This suggests that despite the numerous difficulties, useful age information may indeed be obtained from Os-isotopic analyzes of BMS, and used to constrain tectonic processes.
Tectonic interpretation of Os model ages
In early studies of non-cratonic peridotites brought to the surface both in orogenic massifs (Reisberg and Lorand 1995) and as xenoliths in alkali basalts (Handler et al. 1997; Peslier et al. 2000a) it was assumed that Os model ages represented the time of lithosphere formation. However, studies of abyssal peridotites (e.g., Harvey et al. 2006), forearc peridotites (Parkinson et al. 1998) and ultramafic xenoliths in ocean island basalts (Bizimis et al. 2007; Hassler and Shimizu 1998) have shown that unradiogenic Os compositions, corresponding to whole-rock Os TRD ages > 1.5 Ga and individual BMS TRD ages as old as 2 Ga, can be preserved in peridotites assumed to have been recently extracted from the convecting mantle. Rough correlations between whole-rock 187Os/188Os and indices of melt extraction have even been observed (Bizimis et al. 2007; Liu et al. 2008). Therefore ancient TRD ages found in peridotite xenoliths carried by alkali basalts do not necessarily indicate the formation age of the subcontinental lithosphere. Instead, they may reveal a pre-lithospheric history in the convecting mantle. Furthermore, mantle convection could juxtapose domains of different ages, and possibly create apparent trends that could be misinterpreted as representing a single melting event.
Nevertheless, while the sub-oceanic mantle does clearly include older components, the great majority of peridotite samples, from this reservoir, including both xenoliths (e.g., samples from Canary Islands; Simon et al. 2008) and abyssal peridotites, have whole rock 187Os/188Os ratios greater than 0.120 and TRD ages younger than 1 Ga. Thus, despite the complications caused by the isotopic heterogeneity of the convecting mantle, consideration of Os-isotopic data from alkali-basalt xenoliths in their tectonic context can yield important constraints on the evolution of the subcontinental lithosphere and the driving geodynamic processes. For example, based on the juxtaposition of deep peridotites with young whole-rock TRD ages with shallow mantle samples with older ages, Lee et al. (2000) argued for Mesozoic removal and replacement of the mantle lithosphere beneath the Sierra Nevada region of California. Partial lithospheric replacement was also proposed to explain the observed trend between 187Os/188Os and spinel Cr# found among samples from the Rio Grande rift (Byerly and Lassiter 2012). A very large number of Os isotope studies have also been undertaken in the aim of understanding the timing, extent and mechanisms of lithospheric removal beneath eastern and central China. A summary view of the geodynamic processes that affected the North China Craton, based on Os results from a large number of localities, can be found in Liu et al. (2011). On the basis of temporal relationships with the overlying crust, these authors argue that in the southern part of the craton, both the crust and the mantle lithosphere preserve late Archean ages representing the time of cratonization, while in the northern part of the craton, represented for example by the Hannuoba and Yuangyuan xenolith localities, density foundering (i.e., delamination) and replacement of the mantle lithosphere occurred at around 1.8 Ga. These authors further suggest that the profound lithospheric thinning and replacement that affected eastern China more recently, which they also attribute to density foundering, started in the Triassic beneath the Korean peninsula and migrated gradually towards the west. Other authors, while recognizing the profound nature of the lithospheric replacement process in the region, argue for different mechanisms. Zhang et al. (2008) proposed that magmatic refertilization of ancient harzburgitic lithosphere could produce mantle with the observed asthenospheric characteristics, including young Os model ages. In contrast, based in part on in situ BMS Os-isotopic data, Zheng et al. (2007) emphasized the importance of upwelling of asthenospheric material along major translithospheric features, which would then spread out and rejuvenate portions of the lithospheric mantle through melt-peridotite reactions. Consideration of these studies shows that while Os-isotopic data from peridotite xenoliths carried by alkali basalts do not provide unequivocal answers concerning the large-scale processes controlling lithospheric evolution, they do contribute greatly to the debate.
Highly Siderophile Elements and Os-isotopic compositions in non-cratonic spinel-bearing peridotites brought to the surface by alkali-basaltic volcanism reflect a complex petrological history consisting of partial melting and reaction with percolating melts and fluids en route to the surface. Our understanding of the effects of these multistage processes must be based on comprehensive and detailed studies including the full spectrum of available tools, both geochemical (HSE, 187Os/188Os, S, Se) and petrographical (peridotite texture, mineralogy, BMS–PGM petrography). Investigations of BMS are particularly critical as their specific mineralogy and textural relationships with the silicate or non-silicate minerals provide crucial clues concerning their origin (residual vs. metasomatic) and underpin our understanding of HSE and 187Os/188Os signatures. A concerted effort is needed to link these fine-scale observations with the broader scale relationships observed among whole rock samples. Additional studies of the chalcogen elements (Se, Te), which have received little attention up to now, are sorely needed to provide additional constraints on the behavior of mantle BMS, and by extension, on the HSE that they host.
The multistage petrological history of xenoliths can be summarized in a very simplistic manner as follows (see Fig. 18):
Partial melting triggers a progressive depletion in Pt, Pd and Re due to the preferential partitioning of Pd and Pt in low melting point Cu–Ni-rich sulfides and the partial affinity of Re for silicate phases. High degrees of partial melting will result in high Os, Ir, Ru contents in the whole rock due to their extreme compatibility in the residual BMS or subsequently exsolved high-temperature PGM (pattern 2 on Fig. 18). This initial partial melting episode should probably not be viewed as a simple batch or fractional melting process, but instead as a more complex series of events including aspects similar to refertilization. If left undisturbed, the variable Re/Os ratios left after this ancient melt extraction episode would with time produce correlations between 187Os/188Os and indices of extent of melt extraction.
Once accreted to the lithospheric mantle, the peridotitic residue will experience percolation of silicate melts, which will vary in terms of volume and composition, thus generating a large assortment of signatures. Percolation of large volumes of S-undersaturated OIB-like melt, perhaps related in some cases to the impingement of a plume at the base of the lithosphere, will produce peridotites with poikilitic or granular textures and dissolve the residual BMS (e.g., MSS) originally left after the partial melting event. Dissolution of BMS would result in a dramatic dilution of the HSE concentrations. The extremely low HSE contents resulting from such processes, affecting entire xenolith suites from certain localities, are purely a feature of non-cratonic peridotite xenoliths and are not observed at all in orogenic peridotite massifs, which have higher HSE contents similar to the average values seen in cratonic peridotite xenoliths. This may suggest that such S-undersaturated melt-percolation episodes are broadly related to the tectonic processes culminating in the alkali-basalt volcanism that brings the xenoliths to the surface. Extreme re-equilibration with the percolating melt will generate HSE patterns resembling those of the melts and may also modify the 187Os signatures towards more radiogenic values (pattern 3 on Fig. 18).
The volatile-rich evolved melts derived from the consumption of the S-undersaturated silicate melts as well as primary S-saturated silicate melts generated by low degrees of partial melting will in contrast likely precipitate Cu–Ni-rich BMS, along with clinopyroxene and possibly carbonatitic melt pockets. Because of the pronounced Pd enrichment in the Cu–Ni-rich sulfides, such metasomatism will be evidenced by HSE patterns characterized by Pd enrichments compared to Pt for example, which are generally inconsistent with the degree of major element depletion of the peridotites (pattern 4 on Fig. 18). This metasomatic process can lead to PUM-like HSE patterns and 187Os compositions, though the hypothesis that this can produce the observed whole-rock correlations between 187Os/188Os and indices of melt extraction remains subject to debate.
Fluids, ubiquitous metasomatizing agents of mantle rocks particularly prevalent in regions affected by subduction, have the ability to fractionate HSE in manners different from those of the magmatic metasomatizing agents. For example, such fluids may mobilise Os and Pd preferentially over the other HSE, or provoke sulfidation reactions with magnesian silicates triggering precipitation of solid MSS (pattern 5 on Fig. 18). Comparison of HSE signatures from subduction-related mantle-wedge xenoliths demonstrates the complexity of these overprinting processes, involving either combined fluids and melts or possibly dehydration fluids derived from different subducted materials (oceanic crust ± sediments).
Finally, volatilization or remobilization of Os upon entrainment by the host lavas or after emplacement due to the supergene weathering of their BMS host phases has been proposed to explain the sub-chondritic OsN/IrN ratios of many non-cratonic xenoliths, which similarly to their often low IPGE concentrations are absent from tectonically emplaced peridotites and occur only sporadically in cratonic peridotites (pattern 6 on Fig. 18). However the absence of expected low RuN/IrN ratios and the lack of correlation between OsN/IrN and extent of BMS alteration pose serious challenges to the syn-eruptive Os devolatilization and post-eruptive Os mobility hypotheses respectively. Thus the sub-chondritic OsN/IrN systematics of non-cratonic peridotite xenoliths remain difficult to explain.
Of course, all of these successive processes and their effects on HSE fractionation are presented here in a very simplistic manner. Such a summary should not be viewed as an absolute recipe for explaining every suite and every sample from a given suite, as each may have experienced a specific petrogenetic history. Furthermore, it should be remembered that mantle peridotites, regardless of their tectonic setting, always show the overprints of several petrogenetic processes, as revealed by their HSE patterns. Nothing can ultimately replace detailed study of the peridotite xenoliths, as well as other types of mantle samples, if our aim is to understand the chemical evolution of the different terrestrial silicate reservoirs and to reconstruct the composition of the primitive upper mantle left after extraction of the core and the late veneer bombardment.
We would like to thank James Day and Jason Harvey for inviting us to contribute to this volume of “Reviews in Mineralogy and Geochemistry” on High Temperature HSE geochemistry. Masako Yoshikawa is gratefully acknowledged for her review, as well as James Day and Jason Harvey for their thorough editorial handling and their suggestions to improve this chapter. Katharina Schweitzer (Steinmann Institut, Universität Bonn, Germany) is deeply thanked for her extensive help with the data compilation.