Mantle sulfides (Fe–Ni–Cu-rich base metal sulfides or BMS; Fig. 1) play a crucial role in the distribution of Re, Os, and Pb in mantle rocks and are thus fundamental to obtaining absolute ages by direct geochronology using the Re–Os and Pb–Pb isotope systems on mantle samples. Mantle samples exist as hundreds of exposures of peridotites, pyroxenites and diamonds, either brought to the surface as accidental xenoliths and xenocrysts during kimberlitic or alkali basaltic volcanism (for comprehensive reviews, see Pearson et al. 2014; Aulbach et al. 2016, this volume; Luguet and Reisberg 2016, this volume), or as orogenic, ophiolitic and abyssal peridotite obducted at convergent margins and drilled / dredged from oceanic basins (e.g., Bodinier and Godard 2014; Becker and Dale 2016, this volume).

This chapter reviews the occurrence of BMS in mantle samples and the role that they play in controlling the Re–Os and Pb isotope systematics of the mantle. Included in this review is a discussion of the role BMS plays in recording the multiple depletion / enrichment / metasomatic events that the mantle has undergone and the preservation of chemical heterogeneities that are inherently created by these processes. Along with discussions of the utility of Re–Os and Pb isotope measurements, this review will also consider the potential pitfalls and some of the surprises that can arise when analyzing these BMS micro-phases. Specifically excluded from this review is the extensive literature on Re–Os and Pb for the geochronology of sulfide systems in magmatic ores. This study is another field entirely from the study of sulfides in their native mantle hosts because of the complicated magmatic concentration processes occurring at crustal levels.

This review represents the first time that BMS in mantle rocks and diamonds are discussed together for the purposes of age determination. The interested reader should also consult other articles that have discussed these topics individually (e.g., Shirey and Walker 1998; Burton et al. 1999, 2012; Pearson and Shirey 1999; Luguet et al. 2001, 2003, 2004, 2007, 2008; Richardson et al. 2001, 2004, 2009; Shirey et al. 2001, 2002, 2004a,b, 2013; Alard et al. 2002, 2005; Aulbach et al. 2004a,b, 2009a,b,c, 2011; Harvey et al. 2006, 2010, 2011; Pearson and Wittig 2008; Gurney et al. 2010; Shirey and Richardson 2011; Warren and Shirey 2012).

Through the 1980s, much of the information obtained regarding mantle composition and its inherent heterogeneity was derived through indirect evidence, i.e., the study of basaltic volcanism and the interpretation of its isotopic signatures and trace element compositions (e.g., Zindler and Hart 1986). This method of studying inaccessible regions of the mantle relies heavily on the fundamental assumption that isotope ratios of a source peridotite are faithfully transferred to the resultant melt. However, isotopic studies of mantle rocks have shown that melts average out some of the variability and extreme depletions present in the mantle, especially when the melting region is larger than the scale of heterogeneity (e.g., Saal et al. 1998; Cipriani et al,. 2004; Harvey et al. 2006; Liu et al. 2008; Maclennan 2008; Warren et al. 2009; Stracke et al. 2011; Day 2013; Lassiter et al. 2014; Gannoun et al. 2016, this volume).

The focus on basalts is not only because of a lack of mantle rocks exposed at the Earth’s surface. Problems arise in analyzing mantle samples because of the mobility during melting of the incompatible trace elements upon which early geochronological measurements relied (e.g., Rb–Sr, Sm–Nd, U–Pb, Th–Pb, Lu–Hf isotope systems).

Mantle rocks exposed at the Earth’s surface typically undergo melting (i) as a result of adiabatic decompression, such as beneath a mid ocean ridge, continental rift, or in a mantle plume, and/or (ii) by lowering the solidus with CO2 or water such as when fluid fluxes into the mantle wedge at convergent margins, or carbonated peridotite generates carbonatite and kimberlite. In peridotites, melting is frequently overprinted by melt–rock interaction at a variety of scales and intensities (Warren 2015). Observations from seismic investigations, convection models, and basalt geochemistry also indicate that the mantle is heterogeneous due to prior episodes of melt depletion and enrichment (e.g., Stixrude and Lithgow-Bertelloni 2012; Dalton et al. 2014). The overprint of recent melt addition makes it difficult to “see through” metasomatic processes and determine the nature and, critically, the timing of melt-depletion in the mantle using lithophile elements. In contrast, the highly siderophile elements (HSE), which comprise the platinum group elements (PGE: Ir, Os, Ru, Rh, Pt, Pd) and Re, are compatible at a bulk-rock scale during melting (Re excepted) and have increasingly been used to study mantle composition and processes.

Compared to the Earth’s core, HSE in the mantle are highly depleted (around 10 mg·g−1 versus 30 ng·g−1, respectively; Palme and O’Neill 2003; Lorand et al. 2008). However, these concentrations in the mantle are still unexpectedly high compared to those predicted by core–mantle separation models (e.g., Borisov et al. 1994). The most likely explanation for the elevated HSE concentrations in the Earth’s mantle is the “late veneer” hypothesis - an influx of meteorites impacting the Earth after core–mantle differentiation (Kimura et al. 1974; Chou 1978; Holzheid et al. 2000; Palme and O’Neill 2003; Lorand et al. 2008; Walker 2009).

Two of the HSE, rhenium and osmium, comprise the Re–Os isotope system. Osmium remains in the mantle residue during bulk peridotite melting, while Re behaves incompatibly, so basaltic melts contain relatively little Os. Consequently, metasomatic and refertilization processes involving silicate melts can have little overall effect on the Os isotope ratio of residual mantle, which is fixed by its high Os content. As such, the Re–Os system in ultramafic rocks is generally regarded as the most favorable isotopic system for recovering model ages of mantle melting events (different types of model ages based upon Re–Os isotope systematics are discussed later in this chapter), even though they may contain inherent uncertainties of up to 300 Ma caused by heterogeneity or poor characterization of reference reservoirs such as the mantle (Carlson 2005; Rudnick and Walker 2009). As a result, over the past 26 years Os isotopes have been used routinely to date the melting events that an individual portion of the mantle has experienced, i.e., the transformation of fertile asthenospheric material into a depleted, buoyant lithosphere (e.g., Walker et al. 1989; Snow and Reisberg 1995; Pearson et al. 1995a; Reisberg and Lorand 1995; see also Luguet and Reisberg 2016, this volume, and Aulbach et al. 2016, this volume, for comprehensive reviews).

Partitioning experiments (see Brenan et al. 2016, this volume, for details) have consistently demonstrated that PGE, including Os, on the whole behave as a group of strongly chalcophile and siderophile elements in mantle peridotites, being hosted—in the case of moderately melt-depleted ultramafic lithologies—entirely in accessory Fe–Ni–Cu BMS. As will be discussed in more detail later in this chapter, BMS therefore exert the main control on Re–Os isotope systematics in most ultramafic samples. The high KdOssulfide/silicate of ~ 105–106 therefore ensures that bulk-rock 187Os/188Os is tracked by the behavior and abundance of BMS in the mantle. However, in highly depleted and/or altered peridotites, HSE are often hosted in refractory platinum-group minerals (PGM) and alloys (e.g., Luguet et al. 2007; Lorand et al. 2013) that have formed from BMS as a consequence of reactions involving melts or fluids. Thus, in strongly melt-depleted and/or altered peridotites, the influence of PGM on Os mass balance should not be ignored. Platinum-group minerals are discussed in detail in O’Driscoll and González-Jiménez (2016, this volume).

In the mantle, BMS grains (Fig. 1) are usually present as trace or ultra-trace phases (≪ 0.1 modal %; e.g., Luguet et al. 2001). They have been observed as intergranular grains and as inclusions in silicate minerals in spinel lherzolite and spinel harzburgite xenoliths (Fig. 2; White 1966; Vakhrutshev and Prokoptsev 1972; MacRae 1979; Pasteris 1981; Amundsen 1987), garnet peridotite (e.g., Bishop et al. 1975; Meyer and Tsai 1979) and pyroxenite xenoliths (e.g., O’Reilly and Griffin 1987; Wilshire et al. 1988). Similar BMS grains are also found in abyssal peridotites (Fig. 3; e.g., Luguet et al. 2001, 2003; Alard et al. 2005; Harvey et al. 2006; Warren and Shirey 2012; Marchesi et al. 2013), orogenic peridotities (e.g., Reisberg et al. 1991; Luguet et al. 2007; Lorand et al. 2008; van Acken et al. 2010a), and ophiolites (e.g., Luguet et al. 2004). In diamonds they are recognized as a common group of mineral inclusions, although 99% of gem-quality macroscopic diamonds are inclusion-free (Stachel and Harris 2008).

In the last 15 years, the accurate and precise measurement of Re–Os isotope systematics in individual BMS grains has become possible (e.g., Pearson et al. 1998; Burton et al. 1999; Richardson et al. 2001; Shirey et al. 2001; Griffin et al. 2002). The utility of Re–Os measurements for individual BMS grains has become increasingly apparent in light of growing evidence for the disturbance of bulk-rock HSE systematics by C–O–H-bearing fluids (e.g., Alard et al. 2011) and sulfur-saturated silicate melts (e.g., Harvey et al. 2015, and references therein). Distinct generations of PGM and BMS grains have different Re–Os isotopic signatures, thus leading to the observation that the mantle has a heterogeneous isotopic composition at various length scales (Rehkämper et al. 1999; Griffin et al. 2002; Alard et al. 2005; Beyer et al. 2006; Frei et al. 2006; Harvey et al. 2006). Moreover, in the last few years, combined Pb–Pb and Re–Os isotope studies of peridotite-hosted sulfides have shed new light on mantle geochronology (Burton et al. 2012; Warren and Shirey 2012).

Whole-rock Re–Os analyses of mantle-derived peridotites almost inevitably reflect the mixing of different generations of BMS because they have experienced multiple generations of processes. The movement of Os during melting, recrystallization, or metasomatism may occur independently of Re and there can be separate addition of Re (e.g., Alard et al. 2002; Griffin et al. 2002, 2004), making the interpretation of bulk-rock Re–Os isotope data ambiguous in some samples, in the context of the timing of melt depletion (see reviews in Reisberg et al. 2004; Carlson 2005; Rudnick and Walker 2009). For example, in the study of BMS inclusions in the Western Gneiss region of Norway, Beyer et al. (2004) identified at least two generations of BMS with model ages that ranged from 4.0 to < 0 Ga. Similarly, in the study of Marchesi et al. (2010) of BMS grains from the Ronda peridotite, Spain, multiple generations of BMS were detected at the thin section scale, implying that the whole-rock Os isotope composition and model ages of these samples are controlled by the modal abundance and Os contents of distinct generations of BMS. Such variability at the mineral scale overprints the inherent heterogeneity known to exist in mantle reservoirs fostering concerns that the mantle is adequately homogeneous in terms of 187Os/188Os to be described by large-scale geochemical reservoir systematics (Meibom et al. 2002). Although unradiogenic 187Os/188Os in peridotitic residues often can be adequately explained by ancient melt depletion, using bulk-rock Re–Os isotope systematics indiscriminately can have its drawbacks (Reisberg et al. 2004; Rudnick and Walker 2009). Like the standard caveats in other radiogenic isotopic systems, the Re–Os geochronology of bulk-rock peridotites has two built-in weaknesses: (i) mixing ages can be generated that are not geologically meaningful model ages, and (ii) potential parent (Re)–daughter (Os) mobility either directly or in sulfide liquids at deep-mantle temperatures lead to problems in interpreting Re–Os isotopic composition of sulfides as an absolute age.

The utility of the Re–Os isotope system as a geochronometer thus lies in the ability of BMS to retain an Os isotope ratio that may have been generated billions of years ago. Despite the low closure temperature for Re–Os (e.g., ≤ 300 °C for pyrrhotite; Brenan et al. 2000), the low diffusivity of Os in silicate materials (Behrens et al. 1990; Cherniak 1995; Ganguly et al. 1998a,b; Burton et al. 1999) means that once Os is contained in BMS, it is difficult to equilibrate the Os isotope ratio of a BMS grain with surrounding silicates. For example, for a BMS inclusion in olivine, if the KdOssulfide/olivine is 1, then at 1200 °C DFe–Mg in olivine is about 10−17 m2s−1 (Chakraborty 1997) and the diffusion distance in 106 years is about 1.7 m (using the approximation x2 ~ Dt). However, using a Kd of 106 for sulfide/olivine; i.e., a more realistic partition coefficient as derived from experimental data and measurement of natural samples, the diffusional flux drops by 4 orders of magnitude about 10−21 m2 s−1, and the diffusion distance is reduced to less than 0.5 mm. This makes the retention of ancient 187Os/188Os in a BMS grain trapped within a silicate much more plausible, irrespective of re-equilibration of silicates with metasomatic events subsequent to BMS grain isolation.

Over the last 15 years or so, two main methods have been developed for the analysis of Re–Os isotopes in BMS grains; (i) in situ laser ablation multi-collector inductively coupled plasma mass spectrometry (LA MC ICPMS) and (ii) sulfide extraction, chemical purification, and thermal ionization mass spectrometry (TIMS). Lead isotopes have also been measured using the TIMS technique, as well as by in situ secondary ion mass spectrometry (SIMS). There are distinct advantages and disadvantages associated with each analytical method, particularly with respect to sample preparation and interpretation of the information obtained.

Early work on Pb, Re, Os, and other PGE concentrations in BMS grains was performed by in situ proton microprobe analysis coupled with a MC ICPMS (Bulanova et al. 1996; Guo et al. 1999). Due to limited availability and because detection limits were higher and standard reproducibility for this method was less precise, in situ determination of Re and PGE concentrations are now routinely performed using LA ICPMS (e.g., Alard et al. 2000, 2002,2005; Lorand and Alard 2001; Luguet et al. 2001; Pearson et al. 2002). For Os isotope ratios, in situ LA MC ICPMS typically employs a Nd:YAG 266-nm or 213-nm laser microprobe coupled to the mass spectrometer (e.g., Pearson et al. 2002; Nowell et al. 2008). The new generation of Excimer lasers are also eminently suitable for this purpose and are routinely employed for PGE analysis in metals, BMS grains and PGM (e.g., van Acken et al. 2012).

The spatial resolution achievable by LA MC ICPMS and SIMS, with analytical spot sizes that are typically as small as 50 μm, means that multiple analyses can be made on large (> 100 μm) BMS grains, allowing within-grain heterogeneity to be assessed and any contributions from exsolved phases to be examined. Also, it is possible to raster across large BMS grains to search for variability across a larger area. For smaller grains, the entire grain can be analyzed by increasing the beam diameter such that the whole grain and even part of the surrounding host phase are incorporated in the analysis, reducing the likelihood of skewing of data by omitting sub-grain exsolution products. As silicate or oxide phases are unlikely to contain significant amounts of Re or Os, letting the laser overlap small amounts of surrounding silicates in order to capture the entire volume of the sulfide is unlikely to result in any significant interference on the analysis of the sulfide itself. The ability to sample grains of a wide range of sizes is only limited by the amount of Re and / or Os contained in the grain and the resulting analytical precision that will be achieved. Moreover, after ablation, samples can be polished down further to reveal more material for subsequent ablations, as long as the sulfide is large enough. Hence, a benefit of in situ analysis is that internal textures of BMS grains can be assessed prior to ablation and that profiles through individual grains are possible.

For the analysis of Re–Os in BMS grains by LA MC ICPMS, a mass bias correction is made by bleeding a dry aerosol of Ir into the gas line between the ablation cell and the ICPMS. Although the values obtained by plasma machines may drift over the course of a day, this can easily be corrected for by analyzing a sulfide standard at regular intervals during the analytical session. Typically, instrument induced drift is limited to 1–2 % over a 24-hour period (Pearson et al. 2002).

However, there are several limitations that have been highlighted regarding the analysis of BMS grains using LA MC ICPMS. The overlap of 187Re on 187Os needs to be corrected for by measuring 185Re and using the natural ratio of 187Re/185Re of 1.6742. This isobaric interference can only be corrected for accurately when the 187Re/188Os is below a certain level. There has been some debate as to exactly what the limit should be for precise 187Os/188Os determination. The early study of Pearson et al. (2002) defined the limit of 187Re/188Os as < 1.2 but this was later refined by Nowell et al. (2008), who suggested that the much lower 187Re/188Os value of < 0.5 is more appropriate to ensure an accurate correction of the isobaric overlap. Pragmatically, this 187Re/188Os upper limit rules out the analysis of eclogitic sulfides that have been in equilibrium with basaltic and more evolved liquids—leaving peridotitic sulfides to be tractable. LA MC ICPMS only allows the determination of mass ratios, and not Re and Os concentrations, as internal normalization is not possible. Semi-quantitative Os contents in BMS grains can be obtained indirectly by comparing the signal intensity of an unknown to that obtained on the standard that is analyzed throughout the analytical session. There are, however, several potential limitations on the accuracy of these data. For example, it is extremely difficult to manufacture completely homogeneous sulfide standards for LA MC ICPMS. Nanoclusters of PGE alloy may form when the bead is not quenched instantly, resulting in the possibility of tiny, but significant, heterogeneities on an analysis-to-analysis basis. Ablation conditions may also vary. This method cannot make use of an internal standard and optimum signals may not always be obtained when analyzing particularly small unknowns. In BMS grains of sufficient size, it may be possible to ablate a second hole, next to that ablated for isotopic analysis, for concentration determinations. However, this is only advisable when a large enough area of homogeneous sulfide is available for both measurements.

Where in situ methods have a distinct advantages over TIMS analyses is the speed of analysis. A typical analysis of a BMS grain, including laser warm up time, the ablation of the sulfide itself and then wash out to return to the instrumental baseline often takes less than 15 minutes. Sample preparation time is, in general, low compared to the amount of chemical purification necessary for a TIMS analysis, and is limited to the amount of time necessary to cut and polish a thick (ca. 250 μm) section, inspect the section under microscope, and to locate the BMS grains. Significant sample preparation time is also saved as a single polished section may contain several 10s of grains that would otherwise have to be prepared separately for analysis by TIMS. However, care must be taken in the preparation of sulfides for laser ablation analysis as it is possible to inadvertently pluck some phases from the BMS grains. For example, removal of exsolved chalcopyrite will skew measured Re/Os ratios to lower values, as chalcopyrite may preferentially partition Re (Richardson et al. 2001). Nonetheless, several hundred sulfides can be analyzed in the same time it takes to prepare and analyze a single BMS grain by TIMS.

Textural and mineralogical information relating to a BMS grain and its surroundings is preserved using in situ measurements. This can be particularly useful in immediately identifying BMS that are enclosed versus interstitial relative to the silicate phases. However, the work of Alard et al. (2002) demonstrated that the simplistic distinction between “enclosed” versus “interstitial” as corresponding to “primary” versus “metasomatic” BMS grains is not appropriate in many cases.

The large datasets obtainable by LA MC ICPMS often have to be heavily filtered for sufficiently low Re/Os ratios and for high enough Os contents to give precise isotopic analyses, meaning that, in some extreme examples, up to 90% of a dataset needs to be cut because the 187Re/187Os isobaric interferences cannot be resolved. If bulk-rock Re–Os ratios are also available, then it is advisable to assess the remaining BMS data points in the context of the data on the bulk rock. One of the benefits of being able to rapidly generate large datasets is that probability density plots of Os isotope ratios and model ages (see section on the utility of sulfide Re–Os and Pb isotope geochronology below) can be generated. The large number of analyses required to produce meaningful probability density plots often, but not always, precludes this type of study using TIMS analyses alone (cf. Pearson and Wittig 2008).

The TIMS method is conceptually much simpler than LA MC ICPMS—simply physically remove the sulfide grain by picking or drilling it out from surrounding silicate minerals. There also are some distinct advantages of analyzing individual BMS grains by TIMS. The level of precision and accuracy attainable for Re–Os–Pb isotope analysis is substantially higher, allowing precise measurements at the picogram level for Os and Re and at the tens of picogram level for Pb. This level of sensitivity makes TIMS the more desirable method for the analysis of Re–Os isotope systematics of diamond-hosted BMS, which tend to be very small. By adding enriched spike solutions (185Re 190Os 205Pb) to samples during their dissolution, precise elemental abundances can be obtained in addition to isotope ratios. Without prior knowledge of the paragenesis of a sulfide grain, it can be difficult to accurately spike a sulfide and only an estimate can be made in advance. In practice, the accuracy of a TIMS measurement compensates for under- or over-spiking, except when either excess is catastrophically large. Weighing errors can artificially influence the measurement, particularly for smaller sulfide grains that may have a mass of only a few micrograms, but the use of mixed spike solutions minimizes uncertainty in the analysis.

Sample preparation is by far the greatest drawback of TIMS analysis. Extraction of the sulfide from the host rock, dissolution and purification are time consuming. If the sulfide is being extracted from a polished block, then this represents an additional step in the sample preparation compared to LA MC ICPMS. As with the polishing of samples for laser ablation, the problem exists of incomplete extraction of grains. When a sulfide is extracted from a host diamond or a thick section, the hole can be inspected to check that no sulfide phase remains. However, if sulfides are handpicked from a peridotite crushate or magnetically separated from an aggregate, then this check is not possible and the possibility exists of fractionating Re from Os in an individual BMS grain and producing an artificially low Re/Os ratio (e.g., Harvey et al. 2010, 2011). The advantage of hand picking or magnetically separating sulfide grains is that no material is lost compared to polishing down through a sulfide to identify a large enough grain to analyse. However, individual grains could become broken and grains can be fractured and entire fragments of a sulfide lost. Handpicking of sulfides from a crushate may also result in the preferential selection of interstitial grains that are easier to liberate from a bulk rock crushate.

For the analysis of diamond-hosted sulfides (Fig. 4), the distinction between an eclogitic or peridotitic origin for the sulfide is important because of the large Os concentration differences that can be up to 1000 times less in the eclogitic sulfides. This distinction is typically made by placing the sulfide on a conducting carbon sticky disk in a scanning electron microscope and analyzing the grain for Fe, Ni, Cu, Co, and S. Exsolution features and grain surfaces can also be examined to determine if multiple BMS phases are present (Fig. 4a–b) and to check if the grain has been fractured. This technique could also be applied to BMS grains separated from bulk rock by magnetic separation in order to determine BMS paragenesis, but the textural context that is retained using in situ methods is still lost through the disaggregation of the bulk rock.

Two recent studies (Burton et al. 2012; Warren and Shirey 2012) have presented Pb isotope and concentration data for individual BMS grains from abyssal peridotites, measured using TIMS analytical methods. Burton et al. (2012) measured Pb compositions in sulfides from the same section of peridotite drill core that was previously analyzed by Harvey et al. (2006) for Re–Os compositions. Warren and Shirey (2012) measured Pb, Re, and Os isotopic compositions and concentrations in the same grain by extracting all three elements during the chemical purification procedure and employing spikes. Lead isotopic compositions have also been measured in situ by SIMS, using either a Sensitive High-Resolution Ion Microprobe (Eldridge et al. 1991; Rudnick et al. 1993b) or a Cameca IMS 1280 ion microprobe (Blusztajn et al. 2014). The main drawbacks of this technique are that 204Pb concentrations are too low to be detected and Pb concentration data have large uncertainties due to the lack of low concentration sulfide standards.

For nearly half a century, BMS have been identified in a number of mantle-derived materials, including: peridotite and pyroxenite xenoliths entrained in basaltic and kimberlitic rocks (Fig. 1; e.g., Meyer and Brookins 1971; Desborough and Czamanske 1973; Frick 1973; Vakhrushev and Sobolev 1973; Bishop et al. 1975; De Waal and Calk 1975; Meyer and Boctor 1975; Mitchell and Keays 1981; Lorand and Conquéré 1983; Dromgoole and Pasteris 1987; Fleet and Stone 1990; Szabó and Bodnar 1995), disaggregated clinopyroxene, garnet and diamonds preserved as megacrysts (Figs. 1 and 4; e.g., Peterson and Francis 1977; Gurney et al. 1984; Andersen et al. 1987; Fleet and Stone 1990; Rudnick et al. 1993b; Deines and Harris 1995; Bulanova et al. 1996); and tectonically exposed ultramafic sequences derived from the upper, mostly oceanic, mantle (Fig. 3; e.g., Lorand 1985, 1987; Luguet et al. 2004; Alard et al. 2005).

The interest in sub-mm BMS in ultramafic material stems from early, broad scale investigations into the behavior of BMS under magmatic conditions and sulfide liquid immiscibility in mafic and ultramafic rocks (Figs. 5 and 6; Bell et al. 1964; Kullerud et al. 1969; Naldrett 1969; Skinner and Peck 1969; Craig 1973; Misra and Fleet 1973; Haughton et al 1974; Usselman 1975; Buchanan and Nolan 1979; Fleet and Pan 1994; Karup-Møller and Makovicky 1995). More recently, experimental investigations into the behavior of BMS and, in particular, the HSE and strongly chalcophile elements that they contain, have provided a more comprehensive overview over a wide range of mantle P–T–X conditions (e.g., Bockrath et al. 2004; Ballhaus et al 2006; Holzheid 2010; and, in particular, see the review of experimental methods of Brenan et al. 2016, this volume). The most impressive, and economically interesting examples of the redistribution of BMS through magmatic processes are the large Fe–Ni–Cu-sulfide ore deposits such as Sudbury in Canada, Norilsk in Russia, and the Duluth Complex in the USA. Highly siderophile and chalcophile element-bearing ore deposits are reviewed in this volume by Barnes and Ripley (2016).

The major sulfide phases observed in mantle rocks are generally pentlandite ((Fe,Ni)9S8), chalcopyrite (CuFeS2) ± pyrrhotite (Fe1−xS) and, in isolated instances, bornite (Cu5FeS4), shown on phase diagrams for the systems Ni-Fe-S (Fig. 6) and Cu-Fe-S (Fig. 5). These phases are interpreted as representing the low-temperature (< 300 °C) subsolidus assemblages formed first from monosulfide solid solution (MSS = Fe1−x,Ni1−x S; Fig. 6) and later from intermediate solid solution (ISS) of the Cu–Fe–S system (Fig. 5), and heazlewoodite (Ni3S2) solid solution (HzSS) of the Ni–Fe–S system (e.g., Cabri 1973; Lorand and Conquéré 1983; Dromgoole and Pasteris 1987; Szabó and Bodnar 1995; Guo et al. 1999; Fleet 2006; Lorand and Grégoire 2006). Bornite, if present, likely crystallizes from bornite solid solution (BoSS), which is shifted toward a higher Cu content (and higher metal/sulfur ratio) than chalcopyrite. At the ambient, high pressure and high temperature conditions of the mantle, BMS forms either a homogeneous sulfide melt (known as a matte if it is high abundance) or is only partially molten (Bockrath et al. 2004). Experimental work suggests that MSS is the main stable BMS phase in the uppermost convecting mantle (Bockrath et al. 2004). Independent of their tectonic origin, the BMS found in peridotites, i.e., MSS/pyrrhotite–pentlandite–chalcopyrite, are the result of fractional crystallization from a high temperature sulfide melt followed by sub-solidus re-equilibration. At 1192 ºC, MSS with a composition close to FeS crystallizes, leaving a Cu–Ni-rich residual sulfide liquid at 1192 °C. As temperature decreases towards 1000 ºC, the MSS field extends toward Cu and Ni-richer compositions. With continued cooling (between 800 and 900 ºC, the ISS and HzSS fractionally crystallize from the Cu–Ni-rich residual sulfide liquid. This is often manifested as an outer rim on BMS grains or sometimes as veinlets escaping intergranular BMS. Subsequently, at ca. 600 ºC, HzSS and the MSS may react and crystallize pentlandite, while chalcopyrite crystallizes from the ISS shortly afterwards at ca. 560 ºC. Below 300 ºC further exsolution into a Ni-poor MSS, Ni-rich MSS (> 30 wt% Ni) and pentlandite takes place. However, when MSS–pentlandite–chalcopyrite melts, it does so incongruently, producing a Ni–Cu-rich sulfide melt, leaving behind a Ni–Cu-poor sulfide.

Peridotitic BMS represent ca. 0.1 vol% of the upper mantle, which corresponds to a S concentration of 250 ± 50 μg·g−1 (Lorand 1990; O’Neill 1991; Palme and O’Neill 2003), and therefore, typically, sulfides are exhausted from the mantle after 12–30% partial melting (Luguet et al. 2003; Lorand and Grégoire 2006; Alt et al. 2007). The extent of sulfide exhaustion depends on local sulfur abundance, pressure of melting (Mavrogenes and O’Neill 1999) and melt composition. These reactions with melting result in the formation of PGM that become the major carriers of the PGE, as discussed in detail in this volume by O’Driscoll and González-Jiménez (2016). During melting, Ir and Os are strongly partitioned into the MSS, whereas Cu, Pt and Pd are partitioned into the coexisting melt (Brenan et al. 2016 this volume). Monosulfide solid solution at 900–1100 ºC can accommodate > 104 μg·g−1 of Os, Ir, Ru, and Rh (Li et al. 1996; Alard et al. 2000; Brenan 2002; Mungall et al. 2005; Ballhaus et al. 2006), while pentlandite, chalcopyrite, and pyrrhotite can accommodate from 102–103 μg·g−1 of each PGE (e.g., Peach et al. 1990; Fleet et al. 1996; Fonseca et al. 2011).

Alteration of primary BMS occurs by supergene weathering (Fig. 7), in the case of continentally derived non-cratonic peridotites and by hydrothermal alteration / seawater interaction, in the case of oceanic peridotites. These processes typically increase the number of sulfide minerals observed in ultramafic assemblages. For example, during serpentinization at low temperature (< 250 ºC) and reducing conditions, BMS can experience desulfurization or, conversely, metal enrichment, to S-poor BMS such as heazlewoodite (Ni3S2), jaipurite (CoS), violarite (FeNi2S4), millerite (NiS, covellite (CuS), digenite (Cu9S5) and ultimately native metals or alloys such as copper or awaruite (Ni2Fe–Ni3Fe; Lorand and Conquéré 1983; Lorand 1987; Luguet and Lorand 1998; Alt and Shanks 2003; Bach et al. 2004; Marchesi et al. 2013; Schwarzenbach et al. 2014; Foustoukos et al. 2015). Continental supergene weathering also results in a loss of sulfur but, under oxidizing weathering conditions, it tends to produce Fe oxy-hydroxides at the expense of primary sulfides (Fig. 7; Dromgoole and Pasteris 1987; Lorand et al. 2003a).

Peridotite-hosted sulfides

Mantle-derived BMS can be divided into those that are peridotite-hosted, pyroxenite-hosted, and diamond-hosted. Most primary BMS grains found in mantle peridotites have a diameter of 10–50 μm (Figs. 2 and 3; e.g., Guo et al. 1999), although in exceptional cases this may extend to 200 μm for sulfides enclosed in silicate grains (e.g., Beyer et al. 2004) and up to 500 μm for sulfides in intergranular interstices (e.g., Gonzáles-Jiménez et al. 2014). Sulfide commonly resides as a discrete inclusion within a silicate grain, but rarely within spinel (Ferraris and Lorand 2008). Alard et al. (2000, 2002, 2011) and Lorand and Alard (2001) define Type-1 sulfides as sulfides enclosed in olivine, orthopyroxene and garnet that have MSS compositions with minor amounts of pentlandite and chalcopyrite. The Type-2 BMS of Alard et al. (2002) have compositions dominated by pentlandite and chalcopyrite, with little or no pyrrhotite or MSS. These BMS tend to be intergranular and restricted to grain boundaries, fractures and pockets interstitial to the silicate minerals, as well as metasomatic BMS encapsulated within secondary (i.e., melt-crystallized) clinopyroxene. Type-2 sulfides may be derived by crystallization from a melt or liquid that was trapped during partial melting (Frick 1973; Mitchell and Keays 1981; Hamlyn and Keays 1986; Dromgoole and Pasteris 1987; Szabó and Bodnar 1995; Fig. 1) or introduced after melt depletion as a metasomatic melt or volatile-rich fluid (e.g., Alard et al. 2002, 2011; Luguet et al. 2008; Harvey et al. 2011, 2015; Lorand and Luguet 2016, this volume). These sulfides are the most common BMS population in peridotites, whereas Type-1 sulfides are rare and generally restricted to peridotite xenoliths (e.g., Alard et al. 2002).

The original Type-1 and Type-2 definitions of Alard et al. (2002) encompass not only the geochemical, but also the petrological and textural contexts of BMS grains. Historically, the sub-division into Type-1 versus Type-2 BMS has served the BMS community well, yet the last decade of investigations has highlighted increasing difficulties with dividing mantle BMS into only two categories. In some instances, a third population of BMS may be present where limited mixing of Types 1 and 2 has occurred, resulting in a hybrid sulfide population that shares characteristics of both of its precursors (e.g., Griffin et al. 2002; Wang et al. 2009; Harvey et al. 2010). As will be discussed in the following sections, simple discrimination between “primary” versus “metasomatic” and “enclosed” versus “interstitial” can be difficult. In off-craton xenoliths, sulfide precipitation has also been attributed to sulfidation reactions between S-rich fluids and olivine, or metals dissolved in highly alkaline, volatile-enriched melts. Sulfides related to sulfidation have been identified in xenolith suites from the Kerguelen archipelago (Lorand et al. 2004), the Western United States (Lee 2002), and the Massif Central, France (Lorand et al. 2003b) and Languedoc, France (Alard et al. 2011).

A first order approximation of the nature of the BMS hosted in a particular peridotite, i.e., the balance of BMS composition resulting from melt depletion versus metasomatism, can be made based upon sulfide modal abundance. Sulfur, like Al, is moderately incompatible during partial melting (e.g., Burnham et al. 1998). A sulfide mode of 0.1 % is the approximate theoretical maximum that could be expected from primitive mantle, i.e., a portion of mantle that has not undergone any melt extraction. This is based upon a fertile mantle S abundance of 250 ± 50 μg·g−1 S (Lorand 1990; O’Neill 1991; Palme and O’Neill 2003) and assumes a metal:sulfur ratio of 2:1. Modal abundances of BMS can be determined by careful point counting (e.g., Luguet et al. 2001, 2003, 2004), which is often made difficult by the low abundance and heterogeneous distribution of BMS in peridotite. The preparation of several thin sections is often required in order to ensure a realistic estimate of BMS abundance has been achieved. Although this is not an infallible means for detecting peridotites that have experienced extensive metasomatism, sulfide modal abundances in excess of 0.1 % are unlikely to be the result of melt depletion alone and may signify melt addition. For example, Aulbach et al. (2004a) reported unusually abundant BMS in garnet spinel lherzolites from the Slave craton, which could only be attributed to a secondary influx of sulfur accompanying a silicate melt. This conclusion was also supported by silicate melt patches in nearby spinel lherzolites, half of which comprised aggregates of polygonal pentlandite. Wang et al. (2009) also reported high BMS modal abundances in Taiwanese spinel lherzolite xenoliths as being the result of extensive metasomatism. The modal BMS contents of their lherzolites from Penghu Island range from 0.1–1 modal % sulfide, far in excess of that preserved in peridotite that has only experienced melt depletion.

Peridotite-hosted BMS are predominantly derived from a MSS precursor which was in equilibrium with a sulfide melt (e.g., Craig 1973). Its occurrence is commonly interpreted as representing a residuum after partial melting (Helz 1977; Lorand 1987). Lherzolite-hosted BMS are typically polyphase assemblages of pentlandite ± chalcopyrite ± Ni-poor MSS ± cubanite ± heazlewoodite ± millerite ± bornite. The experiments of Kullerud et al. (1969) in the Cu–Fe–Ni–S system demonstrated that Fe-rich MSS, with a composition close to FeS, crystallizes at 1192 ºC, i.e., at the temperatures at which most non-cratonic peridotite xenoliths equilibrated (<1200 ºC; e.g., Galer and O’Nions 1989; Zangana et al. 1997; Alard et al. 2002). Therefore MSS is a solid sulfide phase at equilibration depths of non-cratonic sub-continental lithospheric mantle (SCLM), and probably for those cratonic xenoliths with the highest equilibration temperature (e.g., ca. 1400 ºC; Rhyzenko and Kennedy 1973). Below 1192 ºC, Ni and Cu partition into FeS to form MSS, which at 1100 ºC accommodates up to 14 wt% Ni, but < 5 wt% Cu. Continued cooling results in an expansion of the MSS field toward Cu, with MSS containing up to 7.5 wt% Cu at 935 ºC (Fig. 5; Cabri 1973). Nickel, on the other hand, may be present at concentrations up to 46 wt% at 1000 ºC, increasing to a maximum of 60 wt % in MSS at ca. 900 ºC, which can be retained during cooling to 700 ºC (Fig. 6; Kullerud et al. 1969; Guo et al. 1999).

Irrespective of the equilibration temperature and cooling rate of the peridotite host, sulfides do not preserve the high-temperature phase relations observed under experimental conditions. Instead, they recrystallize to low-temperature assemblages such as pyrrhotite ± pentlandite ± chalcopyrite (Ballhaus et al. 2001, and references therein). This means that it can prove difficult to determine the high-temperature composition of an individual BMS grain as exsolution requires reconstruction of the original composition. Using rounded or polygonal BMS grains trapped within silicate phases, Aulbach et al (2004a) reconstructed the high-temperature composition of multi-phase BMS inclusions that exsolved from high-temperature solid solutions, despite the presence of exsolved flames of pentlandite in pyrrhotite and thin rims of chalcopyrite (cf. Guo et al. 1999). Precursor MSS was identified by its microstructure and bulk composition, which plots in the MSS-field of Craig (1973) at 300 ºC and a cut-off of 5 wt% Ni was applied to distinguish MSS from pyrrhotite (the maximum solubility of Ni in the pyrrhotite at this temperature; Cabri 1973).

Another example of how high temperature phases can be reconstructed from their exsolved and / or recrystallized products is demonstrated by Wang et al. (2009). These authors reported several peridotite-hosted sulfides as having Cu contents > 7.5 wt% and up to 41 wt%, clearly in excess of the Cu solubility limit in residual MSS, indicating that these BMS grains are not residual MSS but, in fact, ISS. The cut-off of 7.5 wt% Cu serves as a convenient geochemical indicator for the distinction between ISS and MSS. In this instance, it is probable that the ISS exsolved from Cu-rich sulfide melts (e.g., Lorand et al. 2003b; Wang et al. 2009). Ni-rich (> 46 wt% Ni) BMS in these same samples exceed the solubility limit of Ni in residual MSS at 1000 ºC. One possibility suggested by Wang et al. (2009) is that these high-Ni BMS crystallized from sulfide melts at up to 1000 ºC (Ni < 54 wt%).

The interpretation of BMS produced metasomatically through sulfidation reactions (Lorand et al. 2004; Delpech et al. 2012) is based on their morphology (high dihedral angle and sharp-straight BMS grain contours), textural relationships (associated with carbonates), and mineralogy (low Cu/Cu + Ni (< 0.1) and low-Ni, pyrrhotite-dominated assemblages; Lorand et al. 2013). These BMS are always derived from small fractions of alkali melts that have experienced extensive reaction with the host peridotite in the uppermost SCLM (Menzies and Hawkesworth 1987; Bedini et al. 1997; Alard et al. 2011). Such melts have the potential to be particularly effective metasomatic agents as they are enriched in C–O–H–S and remain liquid, and therefore mobile, to low temperatures (Fig. 8; e.g., Menzies and Dupuy 1991; Moine et al. 2004).

In addition to the compositional heterogeneity preserved in BMS as a result of magmatic or metasomatic processes, many intergranular BMS grains preserve varying intensities of dendritic veinlets of Fe-oxyhydroxides (Fig. 7), or can be partially or completely replaced by magnetite. Both of these features are commonly interpreted as resulting from the oxidation of sulfide at sub-magmatic temperatures (e.g., Luguet and Lorand 1998). Sulfide grains in mantle xenoliths may have been partially oxidized by fluids released from the host melt during degassing associated with eruption (e.g., Ryabchikov et al. 1995). Supergene weathering is another cause of oxidation of BMS to oxy-hydroxides (Fig. 7), which results in the selective removal of S in its more soluble sulfate form (Dreibus et al. 1995; Lorand et al. 2003a). This also provides a mechanism for at least partial oxidation of Os and its mobilization, resulting in sub-chondritic Os/Ir in many peridotite xenoliths (Handler and Bennett 1999; Lorand et al. 2003a; Pearson et al. 2004). Such oxidation may even extend to those BMS grains that resided at depth armored within a silicate host grain. For example, Lorand (1990) reported the loss of sulfur and Fe en route to the surface due to the fracturing of host olivine grains during decompression. Finally, metasomatic BMS grains that contain high amounts (> 4 wt%) of Si have been attributed to silicate melt inclusions (Aulbach et al. 2004b). Some sulfides also have unusually high-O abundances (up to 16 wt%), which has also been ascribed to oxidation and alteration of the sulfides (Aulbach et al. 2004b).

Despite a narrow range of starting compositions, the large range of processes that can effect peridotite-hosted BMS results in a great potential for morphological and compositional heterogeneity. This variability is also expressed in the HSE and Re–Os isotope systematics of BMS, which may be a function of primary composition, cooling history, metasomatism and mixing of different populations of sulfides. Alard et al. (2002) described peridotite-hosted BMS as being either Type-1 (MMS enclosed within a silicate host), or Type-2 (dominated by Cu-rich pentlandite, commonly interstitial, but also occurring enclosed within metasomatic minerals such as secondary clinopyroxene) and resulting from either precipitation from a sulfur-saturated silicate melt or the introduction of a metasomatic sulfide melt. It is within this context that the Re–Os and Pb isotope systematics of peridotite-hosted BMS are now reviewed, using primarily the nomenclature of Alard et al. (2002). However, this nomenclature is modified as necessary in the examples that follow, where this division fails to adequately describe the petrogenesis of the BMS in question.

The origin and nature of “Type-1” sulfides

Many mantle peridotites contain at least two generations of BMS (Alard et al. 2000; Fig. 1). As discussed in the previous section, Type-1 BMS are mixtures of MSS (with minor pentlandite and chalcopyrite), whereas Type-2 BMS comprise pentlandite + chalcopyrite ± pyrrhotite that exsolved from the higher temperature MSS or due to reaction between MSS and HzSS. The two sulfide types have very different Re and Os contents, and consequently Re/Os and 187Os/188Os, which in turn have implications for their utility in Re–Os geochronology.

The underlying assumption regarding Type-1 BMS is that they formed during a melt depletion event. This produced a sulfide melt which was lost from the original peridotite and a residual Ni-rich MSS (Type-1 BMS), which was retained within the residual peridotite (e.g., Holzheid 2010 and references therein). Enclosed BMS can exhibit a wide range of shapes that range from polygonal cross-sections with straight contacts to spheroids, but all have habits consistent with entrapment. In harzburgites and dunites, the increase in olivine modal abundance that occurred during melt depletion (70–90 %) could have trapped pre-existing BMS (Lorand 1987).

Osmium concentrations of Type-1 BMS typically range from 1–10s of μg·g−1 in non-cratonic peridotites (e.g., Burton et al. 1999; Alard et al. 2000; Wang et al. 2009; Harvey et al. 2010, 2011) but this extends to 1000s of μg·g−1 in cratonic peridotites (e.g., Griffin et al. 2002; Aulbach et al. 2016, this volume), as summarized in Fig. 9; see also Supplementary Information. Type-1 BMS are Os-rich because their mineralogy is dominated by MSS. This is a function of the greater number of octahedral sites in Ni-rich MSS (Mackovicky et al. 1986; Cabri 1992; Ballhaus and Sylvester 2000), which are capable of accommodating Os, Ir, and Ru. Osmium isotope ratios (Fig. 10; see also Supplementary Information.) in undisturbed Type-1 BMS are typically less radiogenic than the upper limit estimated for Primitive Upper Mantle (PUM, 187Os/188Os ≤ 0.1296 ± 8, Meisel et al. 2001) or mean chondrite meterotites (187Os/188Os ≤ 0.127, Shirey and Walker 1998).

Rhenium concentrations, on the other hand, tend to be much lower (< 1 μg·g−1 in many instances; Fig. 9) with typical 187Re/188Os for Type-1 BMS being equal to or lower than the upper value for chondrite meterotites (187Re/188Oschon ≤ 0.40186; Shirey and Walker 1998). The host phase for Type-1 BMS is most frequently olivine (e.g., Griffin et al. 2002; Aulbach et al. 2004a; Harvey et al. 2006), but occasionally orthopyroxene or garnet (e.g., Beyer et al. 2004). Base metal sulfide included in clinopyroxene with unradiogenic Os isotope ratios (e.g., Burton et al. 1999; Harvey et al. 2011) are likely to be Type-1 BMS, whereas BMS in clinopyroxene with radiogenic Os ratios (187Os/188Os > 0.1270, e.g., Warren and Shirey 2012) are probably Type-2 BMS. Base metal sulfides contained in clinopyroxene megacrysts and in pyroxenitic samples have very different Re/Os and 187Os/188Os systematics, which are a consequence of a different petrogenesis. These are discussed in the following section along with the potential pitfalls of associating unradiogenic Os isotope ratios exclusively with Type-1 BMS enclosed within primary silicates.

The identification of unequivocally wholly enclosed BMS can be problematic. Minute (<10 μm) inclusions can often be seen along healed fractures associated with many Type-1 BMS (Fig. 2), for example in patterns that radiate out along former fractures in the host grain. Other sulfide trails intersect larger BMS inclusions in the host mineral (Fig. 2). These trails may originally have been sulfide droplets controlled by cleavage planes, in the case of pyroxenes, or trapped along olivine sub-grain boundaries (Frick 1973). Once these fractures, or sub-grain boundaries annealed, the sulfide droplets become discrete inclusions in the silicate (Andersen et al. 1987). An alternative hypothesis for the trails of minute inclusions that often surround larger (50–100 μm) inclusions is that, just like silicate melt and fluid inclusions (e.g., Bodnar 2003), BMS inclusions hosted within peridotite xenoliths may experience a degree of decrepitation. This could be induced by the rapid drop in pressure associated with entrainment of a xenolith and transport to the surface within the host basalt (e.g., Andersen et al. 1987). Griffin et al. (2002) noted that many BMS grains in Siberian craton peridotite xenoliths, especially the larger ones, show radiating expansion cracks lined with films of sulfide. Critically, sulfide exposed in a thin section or grain mount that may appear to be distant from any grain boundary in the two observable dimensions may in fact intersect a grain boundary in the third dimension.

Following melt depletion, any residual BMS that becomes trapped within a silicate host grain has the potential to evolve in a low-Re environment indefinitely. The radiogenic decay of 187Re to 187Os (λ = 1.666 × 10−11 year−1; Shen et al 1996; Smoliar et al 1996; Selby et al. 2007) is the primary control on the long term evolution of 187Os/188Os in the mantle. Therefore, undisturbed Type-1 BMS may preserve important geochronological information pertaining to ancient melt depletion events, as the lack of subsequent 187Os ingrowth due to the near absence of Re that was partitioned into the melt preserves the original 187Os/188Os of the BMS in presumed isotopic equilibrium with the silicate melt at the time of melt depletion.

Burton et al. (1999) discussed the significance of the closed-system behavior of enclosed BMS versus the open system behavior of the silicate phases and interstitial BMS. They noted that Os isotope disequilibrium between an enclosed BMS in a silicate grain and its surrounding phases does not occur simply as a function of the enclosure of the sulfide. Osmium isotope ratios that differ between the enclosed BMS and the other peridotite phases can only be produced by (i) partial re-equilibration of the interstitial BMS and silicates with the host basalt during xenolith eruption or (ii) by diffusional equilibration of the xenolith matrix phases (silicates, spinel and interstitial BMS) with a high-pressure mantle melt originally in equilibrium with a peridotite assemblage. In the absence of any strong petrographic or geochemical evidence for host basalt infiltration during transport and eruption of a xenolith (preserved in either the sulfide or silicate phases), interaction and re-equilibration with a subsequent high pressure melt seems the most likely source of Os isotope disequilibrium. Basaltic melt can form an interconnected network even at low melt fractions (e.g., Kohlstedt 1992; Faul 2001) through which diffusion is rapid (e.g., Ganguly et al. 1998a). This facilitates the re-equilibration of silicate and sulfide phases, i.e., the overprinting upon the silicate phases of the osmium isotope signature of the metasomatic melt. If such interactions are responsible for the re-equilibration of the silicate phases and Type-2 BMS, then bulk-rock Re–Os isotope systematics will not necessarily preserve the age of a single event.

These interpretations are supported by the wide range of Os isotope ratios and Re–Os isotope systematics that can be preserved in a single peridotite xenolith. For example, in SCLM xenoliths from non-cratonic settings (e.g., Massif Central, France; Kilbourne Hole, New Mexico, USA), Harvey et al. (2010, 2011 respectively) demonstrated that in several samples, bulk-rock peridotite 187Os/188Os was significantly more radiogenic than enclosed BMS recovered from the same xenolith. This suggests that the bulk-rock only preserves mixtures of primary and re-equilibrated Os isotope ratios and Re–Os isotope systematics. In contrast, BMS that are armored within a silicate grain and protected from open-system behavior preserve the 187Os/188Os associated with the original melt depletion event. The different model ages derived from single enclosed BMS at these localities can be as much as 0.4 Ga older than their corresponding bulk-rock model ages. This phenomenon has also been reported in oceanic peridotites, where an individual Type-1 BMS preserved an age 0.5 Ga older than the abyssal peridotite that hosted it (Harvey et al. 2006). The potential offset between individual BMS model ages and that of their corresponding bulk-rock has been observed, to even greater effect, in peridotite xenoliths from the Kaapvaal craton (e.g., Griffin et al. 2004), where the age of an individual enclosed BMS was shown to be up to 1.5 Ga older than its host peridotite.

The origin and nature of Type-2 BMS

Depending upon the degree of sulfur saturation, a predominantly silicate metasomatic melt may dissolve or precipitate BMS in interstices between and as inclusions within silicate minerals (e.g., Alard et al. 2002). The S-saturation concentration of lavas increases with decreasing pressure/depth (Mavrogenes and O’Neill 1999), allowing S-undersaturated intra-plate basaltic lava to pass through the SCLM and thus remove any sulfide not enclosed in silicates (e.g., Lorand and Alard 2001; Lorand et al. 2003a,b, 2004; Reisberg et al. 2004, 2005). This effect may be enhanced by ongoing open-system melting associated with reactive porous flow at increasing melt/rock ratios, as this has the effect of diluting S in the percolating melt. On the other hand, an upward decrease of porosity and temperature in the SCLM can result in sulfide precipitation in the uppermost SCLM (Bedini et al. 1997). This occurs when metasomatic silicate melts reach S-saturation after/during precipitation of metasomatic clinopyroxene and spinel, leading to the formation of a sulfide melt. If a percolating melt undergoes fractional crystallization, MSS will crystallize first, concentrating incompatible chalcophile/chalcogenide elements (critically Re, but also Pd, Cu, Se, Te, Bi, Sb) in a still mobile Cu–Ni-rich sulfide melt (Alard et al. 2000; Luguet et al. 2003; Bockrath et al. 2004; Ballhaus et al. 2006), while retaining Os in the MSS. Where this MSS crystallization occurs, it is generally a smaller amount than observed in Type-1 BMS, and it would react (possibly completely) with HzSS to produce coarse grains of pentlandite often described as Type-2 BMS. The Cu–Ni-rich sulfide melt remains fluid above 900 °C, but ultimately crystallizes as pentlandite ± chalcopyrite (Lorand and Alard 2001; Powell and O’Reilly 2007). A percolating, S-saturated silicate melt will therefore deposit BMS within the host peridotite (Fig. 1), which corresponds to the Type-2 BMS of Alard et al. (2002).

The occurrence of Type-2 BMS is not restricted to intra-plate continental tectonic settings, as the textural association of clusters of secondary clinopyroxene, spinel and metasomatic BMS has also been described in abyssal peridotite dredged or drilled from mid-ocean ridges (e.g., Luguet et al. 2001, 2003; Seyler et al. 2004; Alard et al. 2005; Harvey et al. 2006; Warren and Shirey 2012; Fig. 3). In extreme examples, i.e., at high melt–rock ratios, the distinction between primary versus secondary and enclosed versus interstial BMS becomes difficult to sustain. If the volume of melt is large, it may be possible to completely replace the original silicate phases and crystallize new ones (e.g., modification of harzburgite into dunite; Büchl et al. 2002, 2004; Lorand et al. 2004; Delpech et al. 2012; González-Jiménez et al. 2014), which may lead to a significant, if not complete, mobilisation and dissolution of the enclosed BMS. Type-2 BMS are often associated with a silicate glass from which the sulfide melt exsolved, as well as with (and sometimes enclosed within) the secondary clinopyroxene and secondary spinel whose fractional crystallization may have triggered S-saturation. Aulbach et al. (2004b) also reported the co-existence of pentlandite (which contained silicate inclusions) with secondary spinels contained in patches of metasomatic glass in cratonic peridotite xenoliths recovered from Buffalo Hills kimberlites, Alberta, Canada.

An alternative hypothesis for the presence of Type-2 BMS in peridotite xenoliths, consistent with their mineralogical association with metasomatic glass, secondary clinopyroxene and spinel, is that they are related to contamination by the melt (basaltic or kimberlitic) that brought the xenolith to the surface. However, several lines of evidence suggest that this is a rare occurrence: (1) BMS derived from host lavas are present as stringers or veinlets that are only ever interstitial. This contrasts with the presence of some Type-2 BMS that are located in the secondary silicates with which they co-precipitate. (2) The major element composition of the interstitial BMS (< 27 wt% Ni) is distinct from the BMS found in host basalts (< 1.0 wt% Ni), as described in xenoliths from Kilbourne Hole (Dromgoole and Pasteris 1987) and Mt. Gambier, Australia (Alard et al. 2002). (3) Bulk-rock and mineral major and trace element abundances are not consistent with the infiltration of REE-enriched basalt. Several early investigations into the relationship between peridotite xenoliths and their host basalts demonstrated that the silicate glass found inside the xenoliths is not derived from the lava. The glass typically has a chemical signature that requires the participation of a hydrous accessory phase (e.g., phlogopite or amphibole), suggesting that the source of the metasomatic melt may be hydrated SCLM (Green et al. 1968; Cooper and Green 1969; Kleeman et al. 1969; Frey and Green 1974; Dasch and Green 1975; Yaxley and Kamenetsky 1999). (4) Textural and mineralogical evidence for pyrometamorphism is absent in the majority of peridotites that contain Type-2 BMS (e.g., Burton et al. 1999).

The preferential partitioning of Re into sulfide melt compared to Os (Brenan 2002, 2008; Brenan et al. 2016, this volume and references therein), combined with Re behavior that is both chalcophile and lithophile, results in a much wider and ubiquitously higher range of Re/Os than those seen in undisturbed Type-1 BMS (Handler et al. 1997; Burton et al. 1999; Harvey et al. 2010, 2011; Fig. 9). With time, high Re/Os will evolve into 187Os/188Os that is more radiogenic than Type-1 BMS in the same peridotites (Fig. 10). Type-2 BMS are S-poor (relative to MSS), and have Ni and/or Cu-rich bulk compositions—like the sulfide melt from which they crystallize. These features, taken together, demonstrate that they are produced by mantle processes that are distinct from the large-scale melt depletion events associated with the formation of Type-1 BMS.

Deficiencies in the “Type-1” versus “Type-2” nomenclature

The nomenclature used in the previous two sections works very well for peridotites that have experienced a simple petrogenesis. However, when peridotites have experienced multiple events, both chemical and thermo-tectonic following an original melt depletion episode, the simple distinction between “melt depleted” and “metasomatic” BMS becomes somewhat inadequate and simple categorization into “primary” versus “secondary” and “enclosed” versus “interstitial” becomes unwieldy. For example, based upon Re/Os, Os/Pt, 187Re/188Os and 187Os/188Os, Griffin et al. (2002) further subdivided peridotite-hosted BMS populations included in olivines from the Siberian craton into five groups (1, 2, 3A–3C). Like the Type-1 BMS of Alard et al. (2002), the Re–Os isotope systematics of Group 1 BMS were consistent with derivation after small percentages of melt extraction from a fertile mantle source under sulfur-saturated conditions. Group 2 BMS were described, based on modelled compositions derived from the experimental literature (see Brenan et al. 2016, this volume), as mixtures of MSS and PGE-rich alloy phases, also formed at low degrees of melting, but under sulfur-undersaturated conditions (Griffin et al. 2002). This was supported by the observation of Pt-rich micronuggets that were apparent only during laser ablation ICPMS. Group 3C BMS were described as broadly analogous to the Type-2 BMS of Alard et al. (2002), appearing to have crystallized from evolved sulfide liquids, whereas Group 3A and 3B BMS were interpreted by Griffin et al. (2002) as reaction products between the MSS found in Groups 1 and 2, and liquids of Group 3C.

Similarly, distinctions based on “enclosed” versus “interstitial” morphologies has already been demonstrated to not uniquely account for where residual versus metasomatic BMS are found. Griffin et al. (2002) demonstrated that inclusions of different BMS groups may occur within an individual olivine grain. This can only occur if the host peridotite has experienced the repeated introduction of sulfide melts and subsequent recrystallization of the surrounding silicates. Group 1 and 2 BMS of Griffin et al. (2002), and some Group 3A BMS yield ages that appear geologically reasonable, i.e., they are at least consistent with the age of the Earth (≤ 4.6 Ga). However, the majority of the Group 3 BMS had particularly high 187Os/188Os, similar to those observed in the Type-2 BMS of Alard et al. (2002), and consistent with a long-term evolution in a high Re–Os environment. The interpretation of the Re–Os isotope systematics of an individual sulfide and its geochronological implications can often be convoluted, requiring a sequence of several events to account for the composition and “age” that it yields (see section below on pitfalls in Re–Os geochronology).

Since 2002, several studies investigating the mobility of sulfide phases during grain boundary migration and metasomatism have been performed, with contrasting conclusions being reached. In some instances, the reaction of primary silicates with interstitial metasomatic melt/fluids may expose BMS grains that were formerly enclosed (Type-1) to percolating metasomatic fluids, disturbing their Re–Os isotope systematics (Aulbach et al. 2004a; Griffin et al. 2004; Wang et al. 2009; Harvey et al. 2010; Alard et al. 2011). In contrast, Marchesi et al. (2010) observed that, irrespective of subsequent metasomatism, BMS from the Ronda peridotite massif preserve ancient magmatic signals. Powell and O’Reilly (2007) and Greau et al. (2013) have suggested that sulfides can still retain their original Os isotope signature if they are only transported over relatively short distances during metasomatism. Because of the high concentration of Os, particularly in Type-1 BMS, they may be able to buffer against the effects of metasomatism. This contrasts with the interpretation of Harvey et al. (2010), where enclosed BMS share some of the Re–Os isotope characteristics of interstitial BMS (and vice versa) in a similar manner to that reported by Griffin et al. (2002).

In a recent study of peridotite xenoliths from Calatrava (Spain) by González-Jiménez et al. (2014), ancient 187Os/188Os model ages were derived from BMS that bore no consistent relationship to their textural location in the peridotite host. This was interpreted, like Griffin et al. (2002), Wang et al. (2009), and Harvey et al. (2010), as the result of fluid/melt interaction with Type-1 and Type-2 BMS. The ancient model ages are interpreted to result from a Type-1 BMS (which have high Os concentrations and unradiogenic 187Os/188Os) with a smaller amount of radiogenic Os derived from the subsequent interaction with metasomatic Type-2 BMS. González-Jiménez et al. (2014) went on to question the validity of the original Type-1 versus Type-2 categorization. Similar to the examples from Calatrava (González-Jiménez et al. 2014), Penghu Islands, Taiwan (Wang et al. 2009), Mont Briançon, French Massif Central (Harvey et al. 2010), and the Siberian craton (Griffin et al. 2002), the Type-1 and Type-2 categories do not adequately describe the populations of BMS present.

Additional complications associated with BMS from oceanic peridotite

In addition to the effects of melt depletion and subsequent metasomatism discussed in the previous sections, oceanic peridotites, i.e., abyssal peridotites and ophiolitic ultramafic assemblages, may also be affected by seafloor weathering (e.g., Bischoff and Seyfried 1978; Snow and Dick 1995), serpentinization (Miyashiro et al. 1969; Wicks and Whittaker 1977; Komor et al. 1985; Janecky and Seyfried 1986; O’Hanley 1996), and other hydrothermal processes (Boudier and Nicholas 1991; Mével and Stamoudi 1996; Alt et al. 2007). Each of these processes may affect the preservation or modification of Re–Os isotope systematics retained in oceanic BMS.

Serpentinization involves the reaction of water with primary silicate minerals (principally olivine) at low to moderate temperatures (generally < 400 °C) to form hydrous phyllosilicates such as lizardite, chrysotile and antigorite, along with brucite and magnetite (e.g., Bach et al. 2004). In the shallow oceanic mantle, the replacement of primary silicates by hydrous minerals can extend up to 100 %, i.e., no olivine or orthopyroxene remains. For example, the degree of serpentinization in oceanic peridotite exposed at the seafloor in the 15° 20′N Fracture Zone of the Mid Atlantic Ridge ranged from 40 % at the surface to 100 % at 100 m depth in the drill hole (Bach et al. 2004). The consequences for the pre-existing populations of sulfide can be profound under these circumstances. Osmium abundance in seawater is extremely low (1.1 × 10−14 g·g−1 Os; Levasseur et al. 1998), but also extremely radiogenic (187Os/188Os ~ 1; Levasseur et al. 1998) compared with PUM (0.1296, Meisel et al. 2001). Rhenium, on the other hand, is enriched in seawater by approximately 740 times compared to Os (Anbar et al. 1992). The high Re/Os of seawater means that peridotite that interacts with large volumes of seawater during serpentinization has the potential to rapidly evolve to radiogenic 187Os/188Os. For example, bulk-rock abyssal peridotites from the Southwest Indian Ridge (SWIR) and American-Antarctic Ridge, analyzed by Snow and Reisberg (1995), possessed elevated 187Os/188Os (187Os/188Os = 0.1305 to 0.1321) for the weathered sample rim compared to the interior of the sample (187Os/188Os = 0.1224 to 0.1253). While absolute Re concentrations were not reported in this study, Harvey et al. (2006) reported a rapid increase in Re abundance with proximity to the surface of the seafloor in drilled abyssal peridotites from the 15° 20′ N Fracture Zone (up to 4.9 ng·g−1, i.e., enriched by up to 3 orders of magnitude compared to similarly serpentinized samples deeper in the drilled core).

The ultimate fate of Type-1 BMS in serpentinized abyssal peridotite depends upon a number of factors. Their MSS-dominated mineralogy has the potential to limit the amount of Re-addition through seawater/sulfide interaction due to the negligible uptake of Re in typically enclosed BMS grains. In addition, the absolute amount of extremely radiogenic Os contained in seawater is unlikely to have any significant effect at seawater/sulfide ratios < 106. Under the reducing conditions of serpentinization, sulfide is generally stable (Snow and Schmidt 1998; Rehkämper et al. 1999) so even highly serpentinized abyssal peridotites have the potential to preserve highly unradiogenic Os isotope ratios (e.g., Parkinson and Pearce 1998; Alard et al. 2005; Harvey et al. 2006; Liu et al. 2008; Warren and Shirey 2012).

Highly siderophile element mobility during serpentinization is controlled by several factors, including oxygen and sulfur fugacities (Bell et al. 2009), pH and temperature (Wood 1987), and fluid chlorinity (Fleet and Wu 1987). Marchesi et al. (2013) recently examined the T, pH, fO2, and fS2 conditions of serpentinization at Hole 1274A near the 15° 20′ N Fracture Zone in the context of HSE mobility. They concluded that HSE mobilization may be minimal overall, being limited to some mobilization of Pt and Pd, in agreement with experimental studies of Gammons and Bloom (1993) and Pan and Wood (1994). Under the reducing conditions of serpentinization (Seyfried and Dibble 1980), it is unlikely that interstitial BMS will be dissolved or physically advected from their host peridotite. It is more likely that, during serpentinization and late stage seawater interaction, interstitial BMS will accumulate Re. If this occurs recently the Os isotopic composition would not be affected. But over time, this grain would evolve to a more radiogenic 187Os/188Os. For example, Harvey et al. (2006) observed skeletal interstitial BMS with unradiogenic Os isotope ratios (187Os/188Os ~ 0.117), but high Re/Os (187Re/188Os ≤ 10). These sulfides must have recently acquired their high Re/Os signature as otherwise their high Re/Os would have resulted in radiogenic 187Os/188Os ratios.

Seafloor peridotites may undergo mutliple episodes of alteration at a range of conditions. For example, some abyssal peridotites from the 15° 20′ N Fracture Zone contain cross-cutting serpentine veinlets, some of which contain valleriite (4(Fe,Cu)S3(Mg,Al)(OH)2), indicating late-stage alteration at higher fO2 conditions (Alt et al. 2007). Less reducing conditions result in higher Pt and Pd mobilization (Gammons and Bloom 1993), while reducing conditions have the potential to break down Ni-rich MSS to awaruite, heazlewoodite, and magnetite (Klein and Bach 2009; Foustoukos et al. 2015). In highly altered peridotites, BMS grains dispersed throughout a serpentine matrix are often partially replaced by veinlets of magnetite, minor awaruite, and occasional native copper. This assemblage is common in partly serpentinized peridotites from the Ronda massif (e.g., Marchesi et al. 2010). While the increase in metal/sulfur ratio of BMS has obvious consequences for the sulfur budget of serpentinites, the effect is also significant for the HSE in BMS (e.g., Foustoukos et al. 2015). These BMS are ultimately transformed into alloys, such as awaruite, with the greater mobility of Re and Pt in seawater resulting in these elements becoming fractionated from Os, although the extent to which this will affect Os isotope heterogeneity in the mantle remains uncertain. However, the preferential removal of Re has the potential to preserve the original Os isotope ratio present in the BMS in the newly formed alloy.

In addition to serpentinization, higher temperature hydrothermal alteration of pyroxenes to talc and amphibole-rich assemblages has the potential to moderate the BMS and HSE budgets of abyssal peridotite. A pervasive, high temperature (ca. 350 °C) hydrothermal overprint superimposed onto a low temperature (100–200 °C) serpentinite assemblage has been documented in some Mid-Atlantic Ridge (MAR) peridotites (Paulick et al. 2006; Boschi et al. 2008; Harvey et al. 2014). The fluids involved are silica rich, having partially equilibrated with local gabbro intrusions, and have Sr and B isotope ratios consistent with a strong seawater influence. While these rocks contain up to 2 wt% S (Paulick et al. 2006), the only BMS identified in these serpentinites are secondary pyrites (Bach et al. 2004), although pyrrhotite, chalcopyrite, pentlandite, and marcasite may also be hydrothermal phases in similar rocks (e.g., Luguet et al. 2004; Alard et al. 2005). Neither bulk-rock nor BMS Re–Os data are available for these samples, but the absence of primary BMS and the presence of hydrothermal BMS makes it likely that these rocks are very radiogenic due to the high 187Os/188Os ratio of seawater. The spatial extent of Si-metasomatized rocks along mid-ocean ridges is not well constrained, but the eventual subduction of oceanic lithosphere containing highly radiogenic BMS has implications for the Re–Os and 187Os/188Os heterogeneity of the convecting mantle and for the mantle wedge underlying subduction zones, where radiogenic Os isotope ratios are common (Brandon et al. 1996; Becker and Dale 2016, this volume).

Overall, studies of oceanic peridotite-hosted BMS indicate that serpentinization, hydrothermal alteration and seafloor weathering can add a number of unwanted variables to the interpretation of Re–Os geochronological data for BMS. However, these studies also indicate that the high-temperature mantle history of abyssal peridotites can be recovered from BMS, provided the degree of alteration is not too extreme, as constrained by careful analysis of sulfide composition and petrography.

Pyroxenite-hosted sulfides

Base metal sulfides have been observed in pyroxenites in several studies over the last 40 years from continental localities as diverse as Australia, the French Massif Central, SW USA and Taiwan (e.g., Irving 1974; Peterson and Francis 1977; Andersen et al. 1987; Dromgoole and Pasteris 1987; Luguet et al. 2008; Wang et al. 2009). Several groups of workers have attributed the formation of Al-augite pyroxenite-hosted BMS, and BMS hosted within clinopyroxene megacrysts and clinopyroxenites, to sulfur-saturation resulting from fractional crystallization of mafic magmas at mantle depths (e.g., Andersen et al. 1987; Dromgoole and Pasteris 1987; Guo et al. 1999). However, Re–Os isotope data for pyroxenite-hosted BMS are scarce and are restricted to a handful of studies (Luguet et al. 2008; Wang et al. 2009; Ackerman et al. 2013).

Pyroxenite-hosted BMS grains are usually pyrrhotite with subsidiary pentlandite ± chalcopyrite, which are the low-T (≤ 300 °C) re-equilibrated products of high-T MSS (Guo et al. 1999; van Acken et al. 2010a). In addition, chalcosite (Cu2S) and Ni- and Co-rich BMS coexisting with MSS and pentlandite have been reported in pyroxenites from the Canadian Shield (Aulbach et al. 2004a). Base metal sulfides measured in pyroxenites from the Penghu Islands, Taiwan (Wang et al. 2009) were reported as having Fe-rich MSS cores, surrounded by a Ni-rich MSS mantle and a Co-rich rim, while a small number of BMS grains contained Cu-rich phases such as chalcopyrite, chalcopyrrhotite and bornite, similar to the range of BMS described from the Bohemian Massif by Ackerman et al. (2013).

The shape of pyroxenite-hosted BMS and their relations to growth planes in pyroxenes suggest that some of them were trapped as drops of immiscible sulfide melt which nucleated on the surfaces of pyroxene crystals growing synchronously from a silicate melt. For example, an x-ray crystallographic study of a clinopyroxene megacryst from Nunivak Island, Alaska demonstrated that BMS grains are aligned along the growth planes (111) and (221), and are not associated with cleavage or exsolution planes (Peterson and Francis 1977). Sulfide shapes include spheroidal, elongated rods and squares, and multi-facetted polygons, suggesting that their shape is controlled epitaxially by the host pyroxene (Dromgoole and Pasteris 1987; Andersen et al. 1987; Guo et al. 1999). While individual pyroxenite-hosted BMS can have sizes up to 500 μm diameter (Guo et al. 1999), more typically they range from 20 to 80 μm (Dromgoole and Pasteris 1987; Wang et al. 2009). However, many enclosed BMS are of a questionable origin. Their association with fractures and with trails of fluid and sulfide inclusions is more consistent with either their introduction sometime after the initial crystallization of the pyroxene host, or by remobilization of pre-existing sulfide (Andersen et al. 1987; Dromgoole and Pasteris 1987). As typically seen for large BMS inclusions in mantle peridotites, it is common to find BMS-bearing inclusion arrays that radiate from sulfides enclosed in magmatic pyroxenes. These sulfides, in extreme examples, have an “exploded” appearance (Fig. 2), which may have resulted from melting and expansion of the sulfide—similar to fluid inclusions that have undergone rapid decompression during their ascent to the surface following entrainment of their host xenolith (Andersen et al. 1987; Dromgoole and Pasteris 1987).

The modal abundance of BMS in pyroxenites tends to be greater than in peridotites (Frick 1973; DeWaal and Calk 1975; Irving 1980; Garuti et al. 1984). Of the pyroxenite suites studied by Irving (1980; BVSP 1981), samples from Kilbourne Hole contained as much as 0.5 modal % BMS, based on bulk-rock sulfur analyses. This is likely to represent a minimum given the susceptibility of BMS to partial or complete oxidation and removal by supergene weathering (Lorand et al. 2003a). Peterson and Francis (1977) described an extreme example of 1 vol% BMS in a clinopyroxene megacryst from Nunivak Island, Alaska. Alternatively, the high abundance of BMS in pyroxene megacrysts may be attributed to the volatile-rich nature of the parent liquid. For example, Harvey et al. (2010) identified compound CO2–silicate–sulfide inclusions in Massif Central peridotite xenoliths (Fig. 8).

The dearth of direct Re/Os elemental and 187Os/188Os isotope measurements of pyroxenite-hosted BMS stems from the fact that (i) pyroxenites make up only around 10% of ultramafic lithologies exposed at the surface (e.g., Frey and Prinz 1978), and (ii) these BMS grains tend to be small (< 100 μm) and possess very little Os compared to peridotite-hosted BMS, making precise isotopic measurement difficult. For example, Luguet et al. (2008) reported Re and Os abundances (Fig. 9), 187Os/188Os ratios, and 186Os/188Os modelled from Pt/Os ratios in 3 pyroxenite-hosted BMS from Beni Bousera, Morocco. Osmium concentrations ranged from 262 ng·g−1 to 1.1 μg·g−1, with supra-chondritic Os isotope ratios (187Os/188Os = 0.130 to 0.145) and significantly lower Re abundances (65 ng·g−1 to 101 ng·g−1). For BMS grains of < 100 μm in diameter, this corresponds to Os amounts of 10s–100s pg. As a result, the Re–Os isotope systematics of pyroxenite-hosted BMS are more typically inferred from the measurement of bulk-rock pyroxenite in which abundant BMS grains have been observed and petrographically described.

The Re–Os isotopic composition of bulk pyroxenite may not always reflects the source melt composition but rather that of the mantle into which the pyroxenite melt has infiltrated. In this situation, the former Re–Os elemental and isotopic fingerprint of the peridotite may not have been entirely overprinted. Distinguishing the 187Os/188Os of a pyroxenite magmatic source from that of Type-2 peridotite-hosted BMS that may have been scavenged during the transit and emplacement of the host pyroxenite is practically impossible. For example, BMS grains with radiogenic 187Os/188Os may be derived from interstitial BMS in a peridotite body prior to the infiltration of large volumes of silicate melt or be derived from the silicate melt itself during the formation of pyroxenite. If pyroxenite represents the interaction between melt generated by a mafic precursor and peridotite, then the melt derived from the mafic component will already have a high Re/Os elemental ratio and time integrated 187Os/188Os ratio compared to the peridotite that surrounds it. If the generation of silicate melt at asthenospheric depths results in a S-undersaturated melt, this melt has the capacity to dissolve interstitial peridotite-hosted BMS as it ascends. These Ni-Cu-rich sulfides have significantly lower melting temperatures than residual Fe–Ni-rich MSS and, as long as the ascending melt remains S-undersaturated, interstitial peridotite-hosted BMS will become incorporated into the melt, mixing their high Re/Os and radiogenic 187Os/188Os with that of a mafic-derived melt with similar Re–Os characteristics (e.g., Sen et al. 2011). However, what seems more likely is that melting of, for example, eclogite at asthensospheric depths will generate a sulfur-saturated melt that is capable of precipitating sulfide with a high, time-integrated Re/Os, leading to the deposition of radiogenic sulfide.

The model for pyroxenite formation favoured by Sobolev et al. (2005, 2007) invokes the reaction of eclogite-derived melt with peridotite, where eclogite would begin melting at a greater depth than peridotite (Hirschmann and Stolper 1996). As eclogites are the metamorphosed equivalent of MORB, they should have high S contents (few hundreds to >1000 μg·g−1 S), thus melting at depths where the S solubility of basaltic melt is relatively low. This melt would therefore be S-saturated, with the potential for precipitation of eclogite-derived BMS when the pyroxenites crystallize. This would result in pyroxenite-hosted BMS with similar Re–Os isotope characteristics to MORB itself, notwithstanding any interaction with peridotite-hosted BMS during pyroxenite formation.

The complexity of progressive melt infiltration, increasing melt/peridotite ratios, pyroxenite formation and sulfide compositional evolution was investigated in two suites of ultramafic rocks from the South China craton by Wang et al. (2009). They demonstrated that changes in HSE distribution and Re–Os isotope systematics of peridotite- and pyroxenite-hosted BMS could be accounted for by the progressive percolation of melts. In porphyroclastic and partly metasomatized peridotites, Wang et al. (2009) reported subchondritic to suprachondritic 187Re/188Os and 187Os/188Os in BMS grains, which likely reflected fluid/melt interaction at low melt/rock ratios with residual MSS and/or crystallization of fractionated sulfide melts. This resulted in elevated Cu, Ru, Rh and Pd, and high Re/Os. At intermediate melt/rock ratios, peridotites displayed equigranular textures, were extensively metasomatized and contained sulfur-rich Ni–(Co)-rich MSS, with subchondritic to chondritic 187Os/188Os and subchondritic 187Re/188Os. At the highest degrees of melt/rock interaction, i.e., where spinel-pyroxenite formed, pyroxenite-hosted BMS possessed the lowest HSE contents, had subchondritic to suprachondritic 187Os/188Os and subchondritic 187Re/188Os, and were interpreted as having precipitated from sulfide melts that segregated from basaltic melts under S-saturated conditions. With the exception of the subchondritic 187Re/188Os, this process is similar to that observed during the formation of dunites at high melt / rock ratios in the Troodos ophiolite by Büchl et al. (2002, 2004) and in Shetland, Scotland (O’Driscoll et al. 2012).

In the Totalp Massif, Switzerland, in garnet–spinel-clinopyroxenites and spinel-websterites contained within melt-infiltrated Jurassic ocean-floor peridotites, van Acken et al. (2008, 2010a,b) identified pyroxenite-hosted pentlandite and godlevskite ((Ni,Fe)8S9) as being the major hosts of HSE. In the host lherzolites, up to 20% of the observed BMS grains had been derived from a mafic melt, resulting in substantial enrichments of Re, Pd and Pt, whereas Ir, Os, and Ru remained largely unaffected. Heterogeneous HSE abundances within individual pyroxenite-hosted BMS were attributed to subsolidus processes, such as exsolution, not resolvable using LA ICPMS, due to small sub-grain sizes. Large grain-to-grain variations were likely caused by multiple events of melt–rock interaction and sulfide precipitation, which demonstrate the potentially complex Re–Os isotope systematics that may occur in pyroxenites and that bulk-rock Re–Os isotope systematics are unlikely to preserve geochronological information pertaining to a single incidence of melt–rock interaction. van Acken et al. (2010a) reported cm-scale heterogeneity amongst pyroxenite-hosted BMS, corresponding to both residual-peridotite and melt-like HSE signatures, suggesting that the prior Re–Os signature of the original peridotite may not have been completely erased. Over time, the enrichment in Pt and Re and depletion in Os of the Totalp pyroxenites would be expected to develop radiogenic 187Os and 186Os, compared to ambient mantle.

In a recent study of Horní Bory, Bohemian Massif, Ackerman et al. (2013) demonstrated that the relationship between bulk-rock Fe-wehrlites, wehrlite-clinopyroxenites, websterites, clinopyroxenite and olivine clinopyroxenite, and the BMS grains they contain are also convoluted and result from the interaction between a melt-depleted peridotite residue and a sulfur-undersaturated melt generated in a subduction setting. Here, the bulk-rock pyroxenites are depleted in Os, Ir, and Ru, enriched in Pd, Pt, and S, and have radiogenic 187Os/188Os. They interpret the origin of these pyroxenites as melt–rock reaction between peridotite and subduction-related melt at high melt/rock ratios, which resulted in primary BMS grains being replaced by radiogenic (high 187Os/188Os) but HSE-poor BMS. Bulk-rock Os abundances are highly variable (0.08–9.3 ng·g−1) but generally low (< 1 ng·g−1), while Re abundances are less variable (1–4 ng·g−1), corresponding to high 187Re/188Os (up to 198) and radiogenic 187Os/188Os (0.139–1.23). However, most of the pyroxenite-hosted BMS have exceptionally low Os and Re, typically below detection limits (< 0.01 μg·g−1), with the result that Ackerman et al. (2013) calculated that the PGE mass balance was not wholly dependent upon the BMS present. Only 28% of the pyroxenite HSE budget was hosted by BMS, with the remainder contained in micron-scale PGM. Despite the small-scale heterogeneity, wehrlite, dunite and pyroxenite rocks yielded a Re–Os errorchon age of 334 ± 19 Ma (MSWD = 1318), which is consistent with the age of other high temperature/pressure rocks from the Bohemian Massif, and likely represents the time of melt–rock reaction and subduction. However, it should be remembered that this “age” is derived from a complex mixture of PGM, secondary BMS, and pyroxenite melt, all assuming that the Re–Os characteristics of the replaced peridotite were completely erased.

The three case studies just described represent the current state of the art with regard to what has been achieved with pyroxenite-hosted BMS. While the geochronology of these rocks must be interpreted with care, detailed grain-scale analyses have the potential to provide unique constraints on the timing and nature of melt–rock reaction events.

Diamond-hosted sulfides

Around 1% of macrocrystalline, gem-quality diamonds contain inclusions large enough to identify with a hand lens or optical microscope. Of such inclusions, BMS grains are not uncommon (Harris 1992; Deines and Harris 1995). The association between diamond occurrences and cratonic mantle has been known since the 1960s (e.g., Clifford 1966; Gurney and Switzer 1973; Boyd and Gurney 1986; Janse 1992). As a result of this association, coupled with (i) cold geotherms derived from geothermobarometry on peridotite xenoliths that pass through the diamond stability field, (ii) the actual presence of diamonds in xenoliths, and (iii) inclusions in diamonds that yield Archean ages, the hypothesis that diamonds reside in ancient lithospheric keels beneath continental interiors has gained wide acceptance (e.g., Richardson et al. 1984; Boyd et al. 1985; Boyd and Gurney 1986; Haggerty 1986). Compared to their low abundance in the mantle in general, BMS inclusions in diamonds from some kimberlites are unusually common (Gurney 1989), making them of particular interest to those who are interested in diamond paragenesis (Meyer 1987) and absolute ages of diamond formation (Pearson and Shirey 1999; Gurney et al. 2010; Shirey and Shigley 2013; Shirey et al. 2013).

Once enclosed in their chemically inert host, inclusions in diamonds are shielded from interaction with transiting melts and fluids, meaning that their chemical composition and the age of diamond formation can be preserved for > 3 Ga (Richardson et al. 1984, 1993, 2001; Smith et al. 1991). Just like the peridotite-hosted BMS discussed above, diamond-hosted sulfide mineralogy is dominated by pyrrhotite (Fe1−xS) and/ or pentlandite ((Fe,Ni)9S8), with lesser amounts of Cu- or Co-rich sulfides such as chalcopyrite (Deines and Harris 1995; Fig. 4). Diamond-hosted BMS have been interpreted as precipitating from a C-O-S-H-rich fluid associated with Archean mantle melting (Deines and Harris 1995) and suggested to be syngenetic with the host diamonds (Bulanova et al. 1996). However, recent work (e.g., Howarth et al. 2014a,b) has shown that diamonds can also be found in metasomatic veins in eclogite and bear no particular spatial association with BMS grains. Furthermore, some diamonds host multiple BMS inclusions that are not in sulfur isotope equilibrium (e.g., Thomassot et al. 2009). In these cases the sulfides must be protogenetic (growing before the diamond). Protogenetic versus syngenetic distinctions are mineralogical in nature and may not compromise the use of BMS for age determination if the age difference between protogenetic inclusion growth and host diamond growth is within the resolution of the age dating method. Nonetheless, the similarity of BMS inclusions to peridotite-hosted BMS is demonstrated by their identical experimentally determined paragenesis i.e., derivation from sulfide/silicate immiscibility in the FeS–FeO–Fe3O4–SiO2 system at around 1100 °C (Maclean 1969).

Base metal sulfide inclusions are distinguished from silicate and oxide inclusions by the presence of large rosette fractures surrounding each sulfide (Fig. 4). The rosette fractures result from rapid decompression during transport of the host diamond to the surface, commonly in a kimberlitic eruption. Silicate inclusions in diamond can be subdivided into eclogitic (E-type diamonds) and peridotitic (P-type diamonds; see Meyer 1987; Stachel and Harris 2008), More recent work has added lherzolitic and wehrlitic parageneses to the classification scheme. Diamond-hosted BMS can be also be categorized by the presence or absence of peridotitic or eclogitic silicate and / or oxide mineral inclusions in the same diamond (e.g., Bulanova 1995), but the co-occurence of BMS and silicate inclusions in the same diamond is so rare that it is hardly a workable classification method for BMS inclusions. Sulfide composition can itself be used, but this classification can only be made after the inclusion is broken out of the diamond as the reflectivity and color differences between different BMS are hard to observe through the diamond.

Major element discriminators for peridotitic versus eclogitic BMS inclusions, such as Ni wt%, were used early in the 1980s in Siberian diamonds (Yefimova et al. 1983). However, with increasing numbers of analyses, the distinction between these two groups of Siberian diamonds became less clear (Deines and Harris 1995; Bulanova et al. 1996). Some studies (e.g., Smit et al. 2010) have observed a pyroxenitic BMS paragenesis compositionally in between eclogitic and peridotitic, which perhaps may explain the less clear distinction for some BMS inclusions. Subsequent studies have found that a combination of Ni and Os abundances is a more robust discriminator (Pearson et al. 1998). Osmium abundances range by a factor of 103 between eclogitic and peridotitic BMS inclusions and provide a more sensitive means of discrimination. Although 12% Ni content was the original identifier for peridotitic BMS, defined by Yefimova et al. (1983) and further refined by Deines and Harris (1995) and Pearson et al. (1999a,b), this scheme of classification does not work as well for BMS inclusions in Siberian diamonds (Sobolev 1974; Gurney 1989). As discussed in the following section, eclogitic (as opposed to peridotitic) BMS were originally defined as containing < 8 wt% Ni (Yefimova et al. 1983), but the two-fold classification was found not to be appropriate for all sulfide inclusions. For example, BMS inclusions containing 11–18 wt % Ni in several Yakutian diamonds (Bulanova et al. 1996) left the provenance of Siberian diamonds in doubt. Consequently, a third sulfide-based paragenesis was proposed for BMS inclusions with intermediate Ni contents between 8 and 12%; a pyroxenitic paragenesis (Deines and Harris 1995; Pearson et al. 1998; Smit et al. 2010).

As with peridotite-hosted BMS, the relatively high Re and Os concentrations in diamond-hosted BMS mean that the Re–Os isotope system is ideal for the analysis of individual inclusions (Pearson et al. 1998; Pearson and Shirey 1999; Richardson et al. 2001). Single grain analysis of BMS has an advantage over the use of lithophile element-based isotope systems, e.g., Sm–Nd, in the analysis of silicate inclusions, as in these instances several inclusions may need to be pooled in order to provide enough material for an adequately precise measurement. However, the advent of a new generation of stable, high-ohmic input resistors on TIMS instruments is likely to change this story in the near future and allow single silicate grains to be analyzed. The strength of single BMS grain Re–Os measurements comes from the observation that not only can a single BMS grain give a model age for its formation, but also multiple non-touching sulfides with a range of Re/Os elemental ratios, closed to diffusive exchange between each other by the intervening inert diamond, have the potential to yield an isochron age if they grew during the same generation of diamond growth. Combined with the Sm–Nd system in lherzolitic and eclogitic garnet and clinopyroxene, for example (e.g., Richardson et al. 1990) or perhaps Pb–Pb if there is enough in BMS, it may be possible to generate complimentary age results by using two or more independent chronometers on silicate and BMS inclusions.

Peridotitic sulfides

Peridotitic diamonds are identified mainly by the peridotite mineral inclusions that they contain. These include diopside, olivine, enstatite, chromite and Cr-pyrope, (Meyer and Boyd 1972; Gurney and Switzer 1973; Harris and Gurney 1979; Fig. 4). Peridotitic diamond paragenesis can also be subdivided into harzburgitic or lherzolitic on the basis of the melt depletion signature of the inclusions. For example, this may be determined by olivine and / or orthopyroxene composition, the presence or absence of diopside or the CaO and Cr2O3 content of included garnet (Sobolev et al. 1973; Harris and Gurney 1979). As discussed in the previous section, unequivocal diamond paragenesis becomes more problematic in the absence of silicate inclusions.

Peridotitic BMS are likely to be residual after low to moderate degrees of partial melting (< 25%; Keays 1995), where the basalts produced will almost certainly be sulfur saturated at the source of melting. In contrast, at the higher degrees of partial melting (>25%) necessary for the production of picrites or komatiites, the exhaustion of sulfur at the source of melting is much more likely to occur, leaving little or no residual sulfur for the formation of BMS that could ultimately become enclosed in diamond (e.g., Keays 1995; Mavrogenes and O’Neill 1999). Therefore, highly depleted portions of the SCLM that are associated with komatiite formation seem to be unlikely hosts for peridotitic BMS inclusions (e.g., Aulbach et al. 2009a). Nickel concentrations of diamond-hosted BMS grains provide some insights into diamond paragenesis, in the absence of co-genetic silicate inclusions. A minimum, cut-off value of 12 wt% Ni for BMS was suggested to be appropriate for peridotitic diamond formation (Yefimova et al. 1983; Deines and Harris 1995; Pearson et al. 1999a,b). More recently, Aulbach et al. (2009a) used (Ni + Co)/Fe ratios for BMS, rather than absolute abundances, in order to determine diamond paragenesis. Rather than analyzing the sulfide major element abundances using in situ methods—and the possible difficulties in reconstructing sulfide major element geochemistry from a number of phases that exsolved from a high temperature MSS (cf. Griffin et al. 2002)—Richardson et al. (2001), Westerlund et al. (2006), and Aulbach et al. (2009b) obtained the sulfide major element abundances from the eluent produced during the purification procedures for Re–Os isotope analysis. As the major metals are removed in a single aliquot, their relative abundance can be determined by mixing the eluent with appropriate gravimetric standards and then analyzing by ICPMS.

Peridotitic BMS share geochemical and isotopic similarities with Type-1 peridotite-hosted BMS, and in some respects can be thought of as extreme versions of those BMS found in non-cratonic SCLM. Osmium concentrations tend to be high (up to μg·g−1 levels) compared to BMS derived from the other diamond parageneses (Figs. 9 and 11). Rhenium concentrations are low, and with high Os concentrations lead to extremely unradiogenic 187Os/188Os ratios in BMS older than 2 Gyr (see below). Although the Re–Os isotope systematics of cratonic diamonds will be discussed in greater detail below (see section titled “The utility of Re–Os and Pb isotope geochronology”), typical Os abundances range from 10s–100s of μg·g−1 in peridotitic BMS from the Siberian craton, Russia (Griffin et al. 2002), and the Slave Craton, Canada (Westerlund et al. 2006; Aulbach et al. 2011, respectively). Exceptionally, several thousand μg·g−1 and even up to 2 × 104 μg·g−1 have been reported in one Siberian BMS grain (Griffin et al. 2002). The abundance of Re in Slave Craton peridotitc BMS grains tend to be < 1 μg·g−1 (e.g., Aulbach et al. 2011) and, while studies using LA MC ICPMS to obtain Os isotope ratios do not report Re abundances because of the isobaric interference on 187Os (e.g., Griffin et al. 2002, 2004; Beyer et al. 2004; González-Jiménez et al. 2014), 187Re/188Os ratios in peridotitic BMS are very low, i.e., significantly less than the mean for chondrites (187Re/188Os = 0.40186; Shirey and Walker 1998; see also Day et al. 2016, this volume), and often much lower, e.g., 187Re/188Os = 0.002 in sulfide LT98/10-5 from the Kaapvaal craton (Griffin et al. 2004) and DA36 peridotitic sulfide from Diavik, Slave craton, Canada (Aulbach et al. 2011). In general, undisturbed peridotitic BMS grains possess 187Os/188Os ratios that are less radiogenic than mean chondrite and PUM (i.e., 187Os/188Os ≤ 0.1270 and 0.1296; Shirey and Walker 1998 and Meisel et al. 2001, respectively). A consequence of these sulfides evolving for several billion years, trapped as inclusions in diamond, and hence in a low-Re environment, is that peridotitic BMS tend to have very unradiogenic Os isotope ratios; typically 187Os/188Os < 0.110 in sulfides with the lowest 187Re/188Os.

Eclogitic sulfides

Unlike the high degree of melt depletion associated with komatiite and picrite formation, subduction of oceanic crust provides more favorable conditions for growth of diamonds with BMS inclusions. Not only may there be an adequate supply of carbon, but also sulfur, as oceanic crust formation occurs under S-saturated conditions. In addition, the low Ni content of most eclogitic BMS leads to low Ni/Fe ratios (0.1–0.2), which are too low to be in equilibrium with mantle olivine and suggest a basaltic precursor (Deines and Harris 1995; Richardson et al 2001). Although a subduction origin for the carbon from which some diamonds form may be controversial (cf. Cartigny et al. 2001), a subduction-related origin for diamond formation and for the supply of material to Archean SCLM roots is independently supported by (i) both S and Pb isotope systematics of BMS grains in eclogitic diamonds (Eldridge et al. 1991; Rudnick et al. 1993b; Farquhar et al. 2002; Thomassot et al. 2009), (ii) by diamond-bearing eclogite xenoliths whose oxygen isotopic systematics are not consistent with a mantle origin (e.g., Jacob 1994; Shirey et al. 2001, 2004a, 2008), and (iii) by the correlation between isotopically light C in diamond and the isotopically heavy O in eclogitic silicate inclusions (Ickert et al. 2013, 2015; Schulze et al. 2013). The S-saturated nature of the eclogitic protolith is less controversial and is based on the occurrence of BMS globules in MORB (e.g., Roy-Barman et al. 1998) and the absence of evidence for wholesale melting of the eclogitic protolith that would have removed the sulfur during subduction. A consequence of the mafic nature of the subducted material is that ultimately this will experience eclogite facies metamorphism, form eclogite and hence supply material that may become encapsulated within diamonds as eclogitic inclusions.

Diamonds are classified as eclogitic when they contain pyrope–almandine garnet ± coesite ± kyanite ± omphacite or sulfide with low Ni contents. In contrast to the high Ni content of peridotitic BMS, eclogitic BMS grains are defined as containing predominantly Fe-sulfide, with less than 8 wt% Ni. Eclogitic BMS mineralogy is dominated by a low-Ni, pyrrhotite–pentlandite–chalcopyrite assemblage derived from cooling of MSS and ISS precursors. Pyrrhotite is usually the dominant phase, but the proportions of chalcopyrite and pentlandite may vary considerably within an individual sulfide. Chalcopyrite is concentrated on exterior surfaces of inclusions (Richardson et al. 2001), as a result of exsolution during kimberlite emplacement. Because Re is preferentially concentrated in chalcopyrite relative to pyrrhotite and/or pentlandite (Richardson et al. 2001; Brenan 2002; Brenan et al. 2016, this volume), incomplete recovery of a diamond enclosed inclusion may preferentially skew the Re–Os elemental ratio to lower or higher values than the true ratio. If no significant exsolution of chalcopyrite occurred, for example if the sulfide was molten during extrusion into the surrounding fractures, then this is unlikely to be a problem. However, the observation that chalcopyrite is preferentially concentrated around the edges of a wide range of sulfides suggests that great care should be taken to exclude exsolution as a potential source of bias from Re–Os geochronological measurements. This exsolution problem and the high Re content make these grains unsuitable for LA MC ICPMS.

Rhenium–osmium isotope systematics of eclogitic BMS differ from those of peridotitic BMS. Osmium concentrations are often several orders of magnitude lower than peridotitic BMS, e.g., < 1.5 μg·g−1 (e.g., Pearson et al. 1998; Pearson and Shirey 1999; Richardson et al. 2001, 2004; Aulbach et al. 2009a), while rhenium contents are only moderately lower in eclogitic BMS than peridotitic BMS (≤ 2.2 μg·g−1; Aulbach et al. 2009a), as shown in Figure 9. Consequently Re/Os are also higher (eclogitic mostly > 1 versus peridotitic < 1; Pearson et al. 1998) and 187Re/188Os are also high (often > 100; e.g., Pearson et al. 1998, Aulbach et al. 2009a) and sometimes exceptionally high (> 1500; Richardson et al. 2004). Again, like pyroxenitic BMS, 187Os/188Os of eclogitic samples are all suprachondritic and often very high in old BMS grains, which contrasts with the dominantly unradiogenic Os isotope signature of undisturbed peridotitic BMS. This is predominantly because of the high 187Re/188Os and, especially in the case of diamond-hosted eclogitic BMS recovered from cratons, time integrated decay of 187Re to 187Os. Interestingly, in a recent study of eclogitic BMS inclusions recovered from Yakutian diamonds in the Siberian craton (Wiggers de Vries et al. 2013), Re–Os isotope dating demonstrated that individual diamonds may have protracted timescales of growth, with vastly different ages for cores and rims. For example, eclogitic BMS recovered from individual growth zones in diamonds from Mir and 23rd Party Congress kimberlite pipes were demonstrated to have crystallized over intervals lasting < 200 Ma, but separated by up to ca. 1 Ga.

Osmium mass balance

The preferential partitioning of HSE into sulfide versus silicate was demonstrated experimentally in the early 1990s (e.g., Peach et al. 1990) and was well-known in ore-forming sulfide systems (see Shirey and Walker 1998 for an early review). But the investigation of HSE concentrations in natural sulfides in non-ore systems has lagged behind by several years. It was not until the investigation of Os abundances in handpicked mineral separates from a picrite from Mauna Loa, Hawaii, an ankaramite from Pico, Azores, and a Kilbourne Hole lherzolite (Hart and Ravizza 1996) that an Os mass balance was estimated for sulfides in equilibrium with silicate melt. This early study demonstrated that in the Kilbourne Hole peridotite, a mineral separate of BMS grains contained 3 orders of magnitude more Os than the bulk rock, and 4–5 orders of magnitude more Os than silicate and spinel mineral separates. Moreover, leaching an aliquot of the powdered lherzolite (in 2.5 M HCl for 1 hour) removed almost 75% of the Os (and all of the S) compared to an unleached aliquot, suggesting that the majority of the Os was hosted by sulfide. Leaching of a finely powdered bulk-rock lherzolite clearly does not provide direct constraints on sulfide Os compositions in the context of their Type-1 or Type-2 assemblage, but can serve as a rough guide to where the phases may be in the rock. In this study, the distribution of Os amongst the other constituent phases of the lherzolite was determined, resulting in calculated KdOssulfide/silicate (≤ 105) that were comparable to those determined experimentally.

Working on a different Kilbourne Hole lherzolite, Burton et al. (1999) analyzed the Re–Os isotopic and elemental composition of inclusion-free silicates, spinel and BMS. By careful handpicking and sulfide selection, they were able to separate both enclosed and interstitial BMS and analyze them separately. In addition, they separated and analyzed clinopyroxene and orthopyroxene fractions that were visibly contaminated by sulfide inclusions. The inclusion-free silicate phases contained very small amounts of Os (104–136 pg·g−1), whereas spinel contained an order of magnitude more Os (1.2 ng·g−1). However, Type-1 and Type-2 BMS contained 3.9 and 3.7 μg·g−1 Os, respectively, unlike several later studies (Alard et al. 2002; Harvey et al. 2010, 2011) where Type-2 BMS were described as containing significantly less Os. Critically, Burton et al. (1999) reported that the sulfide-contaminated pyroxenes had significantly higher Os concentrations than the uncontaminated equivalent, with both orthopyroxene and clinopyroxene containing 1.9 ng·g−1 Os compared to ca. 130 pg·g−1 in the clean mineral separates. Moreover, the pyroxenes that were contaminated by sulfide inclusions yielded 187Os/188Os that were remarkably close to that of the enclosed BMS (included sulfide and contaminated orthopyroxene 187Os/188Os = 0.11719 ± 11; contaminated clinopyroxene 187Os/188Os = 0.11790 ± 28), which were all significantly less radiogenic than the bulk-rock (187Os/188Os = 0.12076 ± 62 to 0.12126 ± 64), the interstitial BMS (187Os/188Os = 0.12086 ± 20), and the uncontaminated silicates (187Os/188Os = 0.12066 ± 11 to 0.12131 ± 22). For this particular Kilbourne Hole lherzolite, this not only confirmed that the Os elemental (and therefore 187Os/188Os) budget is controlled by BMS, but also suggested that the bulk-rock budget was likely dominated by Type-2 BMS in equilibrium with the silicates (Fig. 12) that were slightly more radiogenic than the Type-1 BMS. However, the bulk-rock 187Os/188Os was still significantly less radiogenic than PUM (Meisel et al. 2001) and mean chondrite (Shirey and Walker 1998; Day et al. 2016, this volume).

More recently, Harvey et al. (2010, 2011) revisited the issue of Os mass balance in SCLM lherzolites (Fig. 13), examining Os elemental and 187Os/188Os systematics in four lherzolites from Mont Briançon, French Massif Central (Harvey et al. 2010) and two harzburgites and a further two lherzolites from Kilbourne Hole, NM USA (Harvey et al. 2011). The harzburgites tended to have silicates that were more radiogenic than the bulk rock (187Os/188Os = 0.11600 to 0.12131) with low Os concentrations (≤ 142 pg·g−1). Enclosed BMS grains were unradiogenic (187Os/188Os = 0.1185 to 0.1235) with high Os concentrations (6–25 μg·g−1), while interstitial BMS grains were radiogenic (187Os/188Os = 0.1304 to 0.2163) with generally low Os concentrations (0.001–6 μg·g−1, but with one exceptional grain containing 21 μg·g−1). This suggests that enclosed BMS dominates the Os mass balance of these harzburgites, as confirmed by a quantitative Os elemental and isotopic mass balance (Fig. 13). In total, less than 5% of bulk-rock Os was hosted by the silicates, while the contribution from interstitial BMS was no more than 17% (and as low as 3%) and the remainder (up to 95%) of Os was hosted by Type-1 BMS (Harvey et al. 2011).

One of the two Kilbourne Hole lherzolites in the Harvey et al. (2011) study was similar to the harzburgites, with the same Os elemental and isotope distributions, but requiring an even smaller contribution of Os from interstitial BMS in order to balance the bulk-rock budget (Fig. 13). In the other Kilbourne Hole lherzolite, both the bulk-rock and silicate phases were supra-chondritic, as were most of the interstitial BMS (187Os/188Os = 0.1291 to 0.1694), even though some sub-chondritic BMS grains were recovered. In this example, the Os mass balance was controlled more by the interstitial BMS, although Harvey et al. (2011) were unable to calculate an unequivocal Os mass balance in this sample. The lherzolites from Mont Briançon (Harvey et al. 2010) shared many of the Os mass balance characteristics of the Kilbourne Hole peridotites dominated by enclosed BMS (Fig. 13) but, as discussed below, the 187Re/188Os and Re/Os signatures of these lherzolites are more difficult to interpret than a simple enclosed versus interstitial BMS division.

Overall, the studies of Hart and Ravizza (1996), Burton et al. (1999) and Harvey et al. (2010, 2011) demonstrate that: (i) it is possible to quantify the contributions to bulk-rock peridotite Os mass balance from its constituent phases; (ii) in peridotites where enclosed BMS dominates the Os mass balance, melt depletion is sometimes ancient; and (iii) where interstitial BMS dominates the Os mass balance, bulk-rock Os isotope ratios are not likely to reveal any significant information pertaining to melt depletion. However, this contrasts with the information recorded in interstitial PGM, thought to have formed from extreme melt depletion in spinel lherzolites from Lherz (Luguet et al. 2007). But, as was demonstrated by the radiogenic Kilbourne Hole lherzolite of Harvey et al. (2011), analysis of enclosed BMS from otherwise metasomatized peridotites has the potential to “see through” the late events that precipitated radiogenic interstitial BMS and / or PGM and alloys.

Rhenium mass balance

Rhenium has high Kd,sulfide/silicate, but estimated values are significantly lower than the equivalent partition coefficient for Os. For example, the results of experiments designed to measure partitioning of Re and Os between sulfide and silicate melt (Brenan 2008; Brenan et al. 2016 this volume) produced values of KdOssulfide/silicate/ KdResulfide/silicate >150, which is the minimum required to produce the Re/Os fractionation observed in mantle derived magmas. However, the absolute value for KdResulfide/silicate varies over a wide range, from > 20,000 to ~ 20, depending on fO2fS2 conditions imposed during an experiment (Fonseca et al. 2007; Brenan et al. 2016 this volume). This suggests that Re mass balance in peridotites should also be strongly controlled by BMS and, given the strong partitioning of Re into chalcopyrite and hence interstitial BMS, this phase should dominate the Re bulk-rock mass balance. However, a quantitative Re elemental mass balance that is consistent with observed 187Re/188Os in peridotites is still lacking. Achievement of this mass balance is difficult due to the low abundance of Re reported in silicates (typically < 100 pg·g−1) combined with difficulties in obtaining low Re analytical blanks. In addition, the partial lithophile character of Re compared to the other HSE (Handler and Bennett 1999; Burton et al. 1999; Harvey et al. 2010, 2011) and associated implications for Re–Os fractionation during mantle melting (e.g., Righter and Hauri 1998; Mallmann and O’Neill 2007) make Re behavior with melting more difficult to predict, especially in the presence of residual sulfide. Rhenium is easily removed from a peridotite source during partial melting as it has a low bulk Kdperidotite/basaltic melt (Brenan et al. 2003). The bulk-rock incompatibility of Re in the mantle leads to its high concentration in melt (Alard et al. 2002; Harvey et al. 2010, 2011; Warren and Shirey 2012), especially as BMS dissolve into undersaturated silicate melt during partial melting.

This first order interpretation of Re mass balance is not sufficient to interpret the Re–Os isotope systematics of peridotites with complicated histories such as those from Mont Briançon (Harvey et al. 2010), as the two distinct populations of BMS are difficult to distinguish based on 187Re/188Os, Re and Os concentrations. Similar to the Griffin et al. (2002) study of Siberian xenoliths, Mont Briançon has different populations of BMS that were interpreted as having interacted with each other (see the section on peridotite-hosted BMS above). In both studies, unradiogenic Os isotope ratios have often been preserved in sulfides that at least originated as enclosed BMS, but 187Re/188Os ratios were elevated and high Re concentrations often accompanied high Os concentrations. The original low Re/Os elemental ratio of enclosed BMS may have become elevated through melting and mixing with interstitial BMS melt, but the immediate effect on 187Os/188Os was negligible because of the low Os concentration in interstitial BMS. It is only with the addition of time-integrated 187Os from 187Re decay that the original 187Os/188Os of the hybrid BMS becomes progressively obscured.

Lead mass balance

Although the Pb isotope composition of the mantle has been of interest to geochemists for over half a century (e.g., Holmes 1946; Patterson 1956; Allègre 1969), the role that BMS play in the Pb mass balance of the mantle has been less well explored. Recent interest in the Pb elemental and isotopic composition of mantle sulfides stems from a potential solution to the first Pb paradox (i.e., the position of oceanic Pb isotopic compositions in the “future” Pb position, to the right of the geochron; Allègre 1969), as discussed in the final section of this chapter. A quantitative Pb mass balance (elemental and isotopic) based upon individual peridotites has yet to be accomplished, despite studies that have suggested that mantle Pb resides predominantly in BMS (Meijer et al. 1990; Hart and Gaetani 2006). Hart and Gaetani (2006) predicted that if BMS are the main mantle Pb reservoir they should represent c. 0.05% of the mantle to be consistent with PUM sulfur concentration estimates (Lorand 1990; O’Neill 1991; Palme and O’Neill 2003) and contain ca. 75 μg·g−1 Pb. More recently, Pb concentrations in individual BMS grains extracted from abyssal peridotites from the MAR, Gakkel and SWIR were measured by Burton et al. (2012) and Warren and Shirey (2012). Warren and Shirey (2012) found concentrations that ranged from 0.12–12 μg·g−1 (mean 4 μg·g−1) whereas those in Burton et al. (2012) ranged from 1.5–52 μg·g−1 (mean 19 μg·g−1). These values possibly over-estimate the average Pb concentration as data becomes increasingly difficult to obtain at abundances below about 1 μg·g−1, due to the minute quantities involved in an individual BMS grain analysis.

Using these estimates for BMS Pb concentrations yields a significant deficit of Pb in many abyssal peridotites (Warren and Shirey 2012). For example, using the average Pb concentration of 4 μg·g−1 Pb found in Gakkel and SWIR BMS (Warren and Shirey 2012) and the average abyssal peridotite BMS modal abundance of 0.02% (Luguet et al. 2001,2003) gives a bulk peridotite Pb concentration of 0.8 ng·g−1, significantly lower than the estimate of 150 ng·g−1 for PUM (McDonough and Sun 1995) and also lower than the estimate of 21 ng·g−1 for depleted MORB mantle (DMM; Salters and Stracke 2004; Workman and Hart 2005). Using the maximum modal estimate of 0.06% for mantle BMS in abyssal peridotite (Luguet et al. 2003) and the maximum BMS Pb concentration observed by Warren and Shirey (2012) of 12 μg·g−1, only about 40% of Pb in DMM can be accounted for. Using the mean value of Pb in BMS analyzed by laser ablation (Burton et al. 2012) from the 15° 20′ Fracture Zone, MAR (19 μg·g−1) can account for approximately 60% of DMM Pb. Using the maximum value of 52 μg·g−1 and the maximum modal abundance of 0.06 % can account for the mean DMM Pb abundance of 21 μg·g−1, but would still only represent ca. 25% of the estimated PUM Pb abundance. However, these maximum estimates of BMS Pb concentration and abundance are not representative of the oceanic mantle.

Observations of xenoliths suggest that Pb is present in the silicate phases at concentrations < 1 μg·g−1 (e.g., Meijer et al. 1990; Hauri et al. 1994; Carignan et al. 1996; Eggins et al. 1998; Norman 1998; Ionov et al. 2002), but given their modal abundance this can account for a significant proportion of the bulk Pb. For example, Carignan et al. (1996) measured the Pb concentrations of the main silicate phases (olivine 11 ng·g−1; orthopyroxene 20 ng·g−1; clinopyroxene 470 ng·g−1) in two peridotite xenoliths and calculated their bulk-rock Pb abundances to be 21 and 29 ng·g−1, without the inclusion of peridotite-hosted BMS. These values were remarkably close to the measured bulk-rock Pb concentrations of 16 and 40 ng·g−1, respectively. In this instance the three main silicate phases in peridotites can account for 73–100% of mantle Pb concentrations. Overall, it would appear that mantle BMS contain much less Pb than that predicted by Hart and Gaetani (2006).

The major difficulty with a Pb elemental and isotopic mass balance for the oceanic mantle is a lack of data on Pb in the silicate phases and bulk rock. Warren and Shirey (2012) provided an elemental mass balance for a single SWIR sample based on Pb concentration data for pyroxenes and bulk peridotite. They found reasonable agreement between the measured bulk-rock concentration of 31 μg·g−1 and the reconstructed concentration of 22 μg·g−1 (excluding sulfides), as olivine would only need to contain 13 μg·g−1 to fully match the measured value. However, the measured bulk-rock concentration of 31 μg·g−1 is higher than the current estimate of 21 μg·g−1 for Pb in DMM, indicating the need for more systematic analysis of Pb concentrations in abyssal peridotite, particularly as studies of peridotite xenoliths have shown that bulk Pb concentrations can be highly variable (e.g., Carignan et al. 1996). This will also require improved constraints on the effect of alteration on peridotite Pb concentrations. Clearly several fundamental unknowns remain to be investigated with respect to the Pb isotopic and elemental mass balance in the mantle, most importantly, the contribution that mantle BMS make to solving the first Pb paradox.

Sulfide Re–Os isochrons, TMA, TRD, and γOs

An isochron is the “gold standard” in geochronology, notwithstanding the possiblility that an isochron may in fact represent a mixing line. For sulfides, Re–Os isochrons (Fig. 14) can be generated when a selection of co-genetic BMS in chemical equilibrium at the time of system closure, but with a range of Re–Os, evolve over time, under closed-system conditions. This, in theory at least, allows their isochronous relationship to be used as a precise measure of the time elapsed since they were isolated from open-system conditions. Isochrons also have the advantage of providing both age and initial isotopic composition. By projecting the slope of the isochron back to the intercept on the y-axis, i.e., at zero 187Re/188Os, this allows the 187Os/188Os at the time the system closed to be calculated, and hence inferences about the source of the sulfides and petrogenetic history can be made. Figure 14 shows examples of true isochrons among diamond-hosted BMS populations from Aulbach et al. (2004a) and Westerlund et al. (2006). Harvey et al. (2006) came close to a true isochron with a population of enclosed serpentinite-hosted BMS from the 15° 20′ N Fracture Zone on the MAR (mean square weighted deviation or MSWD of 3.3), as shown in Figure 15. For this same location, Burton et al. (2012), in a plot of 207Pb/204Pb versus 206Pb/204Pb (Fig. 16), determined an isochron age of 1.83 ± 0.23 Ga (MSWD = 0.18) for enclosed serpentinite-hosted BMS. This value agrees, within analytical uncertainty, with the Re–Os age of 2.06 ± 0.26 Ga from Harvey et al. (2006); i.e., compare Figs. 15 and 16.

Unfortunately, many arrays of data derived for non-diamond related sulfides have MSWD values that are > 2.4, indicating poor statistics for the data regression (Faure 2001), and cannot meet the conditions for a precise Re–Os geochronometer. Instead, isotopic data for these BMS can be compared to reference isochrons as an alternative to actual isochron regressions. In many instances, this is the closest to an isochron age that many studies are able to achieve. In BMS from diamonds, such data arrays can give a useful indication of the general age for the diamonds in question. That such BMS grains might not fall on a strict isochron is understandable given that diamonds are xenocrysts whose spatial relationship to each other in the lithosphere is not known and thus may be too far apart to have achieved strict chemical equilibrium.

The complexity of melt depletion and subsequent metasomatism, tectonism and metamorphism often obscures any meaningful information that could have been extracted from a population of BMS grains prior to their disturbance. Diamond-hosted BMS are less susceptible to these complexities because they are encapsulated and protected from secondary melts and fluids, but they do have xenocrystic-related uncertainties. As a consequence of these difficulties in producing Re–Os isochron ages, model ages are often employed instead. A model age represents a time when the isotopic growth curve for a mineral coincides with the isotopic evolution of a reference reservoir such as the mantle. A model age is an attempt to derive meaningful geochronological information with imperfect knowledge of the sequence of events that have affected the samples, or when there is simply an insufficient number of samples with which to construct an isochron. For example, they allow an “age” to be derived from a single data point.

In the case of the Re–Os isotope system, two types of model age calculations are in common use (Fig. 17). Both calculations compare 187Os/188Os in an unknown to the 187Os/188Os evolution curve for mantle with chondritic Re/Os. The evolution of 187Os/188Os from the initial osmium isotope ratio of IIIA iron meterorites (the most primitive Solar system initial value: 187Os/188Osi = 0.09531; Smoliar et al. 1996) is used to estimate the average composition of chondritic meteorites for the present day (Fig. 17a). A 187Re decay constant (λ) of 1.666 × 10−11 yr−1 is used for this and all other age-dependent calculations:


The measured composition for chondrites includes early measurements that gave 0.1262 ± 0.0005 (Allègre and Luck 1980; Luck et al. 1980; Walker and Morgan 1989). Meisel et al (1996) reported a narrow range of present-day 187Os/188Os ratios for 11 ordinary chondrites of 0.1289 ± 0.0022 (2σ) and a similar present-day 187Os/188Os ratio of 0.1283 ± 0.0008 (2σ) for 8 enstatite chondrites. The latter was later refined by Walker et al. (2002a) and termed the enstatite chondritic reservoir (ECR), with present-day 187Os/188Os = 0.1281 and 187Re/188Os = 0.421. In contrast, a suite of 4 carbonaceous chondrites defined a 1–2% lower 187Os/188Os ratio of 0.1263 ± 0.0008. A detailed discussion of the variability of chondrite 187Os/188Os can be found in this volume in the chapter on HSE in planetary bodies (Day et al. 2016, this volume).

The two commonly employed model age calculations (Walker et al. 1989; Shirey and Walker 1998) are TMA (Eqn. 2) and TRD (Eqn. 3), where both model ages use Equation (1) to calculate a mantle evolution curve. For TMA, both the 187Re/188Os and 187Os/188Os ratios in the unknown are used to calculate the time of separation from asthenospheric mantle:


This model age can be used both for BMS with low Re/Os (e.g., peridotite-hosted Type-1 BMS and diamond-hosted peridotitic BMS) and for BMS with much higher Re/Os (e.g., pyroxenite-hosted BMS, peridotite-hosted Type-2 BMS, and diamond-hosted eclogitic BMS). It assumes that the Re/Os measured today in a sample is representative of its long-term history in the mantle. If this assumption is correct, this implies that the geochronometer started at the time that the sulfide cystallized and no subsequent gain or loss of Re occurred. As such, TMA ages can provide robust geochronological information regarding the age of an individual BMS grain (Fig. 17b). It is not always safe to assume that undisturbed Re/Os systematics have been preserved. For example, where an elevated Re/Os elemental ratio accompanies an unradiogenic 187Os/188Os, or the converse, where an unsupported radiogenic 187Os/188Os (i.e., high 187Os/188Os with low Re/Os) is observed, a TMA calculation will yield an erroneous number that likely bears no resemblance to any actual event in the history of the sulfide (Fig. 17c). Such a situation can occur when Re from the host lava (such as a kimberlite) is added to a xenocryst or xenolith during transport to the surface. Under these circumstances a TMA age might yield a nonsense or “future” age and in this case a TRD, or rhenium depletion age, is likely to be the most appropriate model age calculation.

TRD is calculated as an intersection time with chondritic mantle growth, assuming a melt-depletion event previously removed all Re from the sample. It is calculated using the following equation:


where 187Os/188Ossample (EA) is the ratio at the time of separation from the mantle:


For a xenolith-hosted BMS (Fig. 17c), 187Os/188Ossample (EA) represents the osmium isotope ratio at the time the host xenolith was entrained and brought to the surface in a lava. However, the correction in the sample’s isotopic composition defined by 187Os/188Ossample(EA) is only necessary if the eruption of the host lava is far enough in the past or enough Re has been added by the host lava to dramatically change the 187Re/188Os such that sufficient ingrowth of 187Os could be anticipated. With TRD, the assumption is made that Re was completely removed from the sulfide during the event that isolated it from the convecting mantle. Therefore, in the absence of any ingrowth of 187Os from 187Re, the measured 187Os/188Os of the sulfide has remained unchanged since the sulfide crystallized. This is clearly an oversimplification, as even undisturbed diamond-hosted peridotitic BMS and Type-1 peridotite-hosted BMS contain some Re and some 187Os ingrowth is inevitable. Hence the measured 187Os/188Os used in a TRD age calculation will only yield a minimum age (Fig. 17) but one that could nonetheless be useful if the Re concentration is low.

A quick examination of equation (3) reveals that relying entirely upon a measured 187Os/188Os ratio for a model age may not yield the most accurate geochronological information and that, in examples where the measured 187Os/188Os exceeds that of the chondritic reference value, it is not possible to obtain any useful geochronological information. Meaningless TRD ages in the future and TMA ages older than 4.5 Ga reflect derivation of Os from a source more radiogenic than PUM and / or the local disturbance of the Re–Os system. Fortunately, most host lavas have very low Os content and thus their ability to modify the Os isotopic composition of a mantle peridotite or a BMS grain is very limited (e.g., Walker et al. 1989; Shirey and Walker 1998).

An additional method has been developed to handle datasets produced by LA MC ICPMS as a consequence of the relative ease with which large numbers of 187Re/188Os and 187Os/188Os ratios can be generated (e.g., Griffin et al. 2002, 2004; Marchesi et al. 2010; González-Jiménez et al. 2014). Calculating model ages for these sulfides can give a high density of “ages” for a given region, allowing model ages to be plotted on histograms. This in turn allows peaks in the frequency of ages to be determined. More recently, large datasets of model ages have been presented as relative probability plots. For example, the model ages generated from BMS inclusions in olivines from Udachnya, Siberia (Griffin et al. 2002, using the models of Ludwig 2000) were incorporated into a relative density plot and used to suggest that most of the lithospheric mantle beneath the Daldyn kimberlite field formed during the period 3–3.5 Ga.

Bulk-rock data and BMS data generated by TIMS could also be used for this type of analysis, but the time taken to acquire sufficient data using TIMS is considerably longer. A case in point are the relative probability plots for the Kaapvaal and Slave cratons presented in Pearson and Wittig (2008), which are each comprised of TIMS data compiled from measurements acquired over several years (Kaapvaal: Pearson et al. 1995a; Carlson et al. 1999; Menzies et al. 1999; Irvine et al. 2001; Carlson and Moore 2004; Simon et al. 2008; Slave craton: Irvine et al. 2003; Aulbach et al. 2004a; Westerlund et al. 2006). In addition to a large peak in BMS Re–Os model ages at 2.5–3.0 Ga for the Kaapvaal craton (Fig. 18), the relative probability plots of Pearson and Wittig (2008) also demonstrated a main peak in BMS model ages around 2.8–2.9 Ga (see also Aulbach et al. 2004a) and subsidiary peaks at 1.3 Ga, 2.2 Ga, and 3.9 Ga for the Slave Craton (Fig. 18). These examples demonstrate the power of large numbers of Re–Os model ages to constrain the timing of regional scale events and the link, or lack thereof, between ages preserved in the SCLM and crust forming events (i.e., large-scale partial melting). Aulbach et al. (2016, this volume) discuss the significance of relative probability plots in more detail, including the use of HSE in dating cratons.

Finally, often quoted alongside Re–Os model ages, although not a model age itself, is the concept of γOs (gamma osmium), where the Os isotope ratio of an unknown is compared to that of a chondritic reference, with the difference at a specific time between the unknown and the chondritic reference (γOs) reported as either a positive or negative percentage (Fig. 17a). The underlying concept is exactly the same as the delta notation (δ) commonly used for many stable isotope systems (per mil deviations from a known standard) and the epsilon notation (ɛ) for Nd and Hf isotope ratios (parts per 10,000 deviation from a known standard):


Samples with positive γOs often are described as enriched, radiogenic or suprachondritic and imply long-term elevated 187Re/188Os. Samples with negative γOs often are described as depleted or unradiogenic and imply long-term lowered 187Re/188Os.

Potential pitfalls with sulfide geochronology

The fractionation of Re and Os during mantle melting preserves distinct Re/Os and 187Os/188Os systematics in primary BMS grains enclosed within either a silicate or diamond host (i.e., Type-1). These contrast with Type-2 BMS that are associated with metasomatism, which makes the Re–Os isotope system an ideal tool for dating melt depletion in the former and metasomatism in the latter. However, as with all isotope systems, there are limits to the utility of the Re–Os isotope system.

Ages calculated from sulfide isotopic compositions are subject to a number of potential pitfalls, including uncertainties in model age calculations, modifications induced during sample preparation or analysis, and changes to original Re–Os isotope systematics that derive from mantle processes. All of these modifications may obscure or even render useless model age information preserved in mantle BMS.

Uncertainties in model age calculations

Although the sulfides which are the subject of this chapter are mantle-derived, to obtain geologically meaningful TRD and TMA ages a choice of the reference reservoir for 187Os/188Os at time in the past must be made. The range and heterogeneity of published mantle values and, in particular, the wide disparity between mean 187Os/188Os for abyssal peridotite versus that of PUM (ca. 4% difference in 187Os/188Os between these two reservoirs, discussed in detail in Rudnick and Walker 2009; cf. Gannoun et al. 2016, this volume) means that a wide range of model ages for a single BMS grain could be derived depending upon which mantle value was chosen for normalization. Using abyssal peridotite or PUM as a basis for calculating TMA and TRD ages yields younger and older model ages, respectively, than using mean chondrite. The main problem with PUM estimates is that PUM is an artificial construct designed to predict the composition of a notional pristine mantle by projecting back from samples that have been modified over the course of Earth’s history. The disparity in the evolution models for the mantle impacts model ages by ca. 300 Ma (e.g., Pearson and Wittig 2008; Rudnick and Walker 2009). These differences are relatively unimportant for ancient systems but can be significant when attempting to date Phanerozoic events using the Re–Os geochronometer. Therefore model age calculations relative to PUM are probably not appropriate (Rudnick and Walker 2009). The carbonaceous chondrite average 187Os/188Os of 0.1262 ± 0.0005 (see Day et al. 2016, this volume) is probably still the best, and is the most commonly used reference for calculating model ages. In general, however, model ages rarely can carry a precision better than several hundred million years and this imprecision will offset some of the need for an accurate knowledge of the compositional evolution of the reference reservoir (e.g., the mantle).

Sample preparation issues

Some diamond-hosted BMS break up during the process of removing them which can result in skewed Re/Os ratios (Richardson et al. 2009). This is because Re is preferentially partitioned into chalcopyrite relative to pyrrhotite and/or pentlandite whereas the reverse is true for Os (e.g., Craig and Kullerud 1969; Li et al. 1996; Piña et al. 2012). The separation of chalcopyrite from pyrrhotite or former MSS will typically occur at a much later time (e.g kimberlite eruption) relative to the time averaged Re/Os that was established for the whole sulphide grain. This can introduce a different measured 187Re/188Os if chalcopyrite or pyrrhotite makes up a disproportionate part of the analysis. For example, Richardson et al. (2001) reported up to 30% difference in 187Re/188Os amongst fragments of chalcopyrite, pentlandite and pyrrhotite recovered from eclogitic diamonds. An artificial lowering of 187Re/188Os will skew the calculated TMA of an affected sulfide to an artificially young model age (see Eqn. (2); Fig. 19). Consequently, Richardson et al. (2001) attributed scatter on Re–Os isochrons to the possible incomplete recovery of sulfide inclusions, in particular chalcopyrite-rich rims. The absence of well correlated linear arrays for BMS Re–Os isotope data was also attributed by Beyer et al. (2004) to the potential loss of Re-rich chalcopyrite rims during the preparation of polished sections for in situ analysis by LA MC ICPMS. With careful sample preparation, these artificially induced Re/Os fractionations and potential corruptions of model age calculations may be avoided, or at least minimized.

Mantle-derived changes to Re–Os systematics

Negative model ages, often meaninglessly referred to as “future ages”, along with model age calculations that return “ages” that are greater than the age of the Earth are the result of a departure away from the idealized Re–Os isotope systematics of Type-1 peridotite-hosted (or diamond-hosted peridotitic) BMS grains. Examination of Equations (2) and (3) reveals that if melts with Re or Os interact with a population of BMS grains, the original geochronological information that was contained within that population may be obscured. Hence, BMS model ages should not be used indiscriminately, as their compositions may reflect cumulative effects of several events. Of particularly importance are the addition of Re (both recent and ancient) and radiogenic Os, and the implications of these on TMA and TRD are summarized in Figure 19.

Osmium enrichment

Unsupported radiogenic Os (i.e., insufficient 187Re to account for the 187Os present in a geologically reasonable time, also referred to as suprachondritic; Fig. 19) can be attributed to subduction-related processes. Subduction can involve sulfides equilibrated with the slab (the crustal portion of which is basaltic with a high Re/Os) and fluids in the wedge which can move Os. Highly oxidising subduction-related fluid liberates radiogenic Os, but not Re, from the downgoing slab (Brandon et al. 1996), resulting in a higher 187Os/188Os that suggests derivation from a time-integrated high-Re/Os reservoir. This elevated Os isotopic composition appears to be a general feature of arc peridotite whole rocks (e.g., Widom et al 2003). For this to be observed in an individual BMS grain requires a multistage history for that grain, which may have crystallized as an unradiogenic MSS with a low Re/Os ratio, followed by interaction with a melt or fluid that contains radiogenic Os, but relatively little Re (e.g., Wang et al. 2009). This process has the effect of raising the 187Os/188Os, but suppressing still further the originally low Re/Os ratio of the sulfide. These features were evident in the peridotitic BMS inclusion suite studied from the Ekati Mine, in the Panda kimberlite pipe, Slave craton, where low Re/Os BMS grains were enriched in radiogenic Os associated with specific zones of diamond growth (Westerlund et al. 2006). Complicated, multistage scenarios are not as common in sulfide inclusions in diamond as they are in sulfides from mantle peridotites (see below), because the encapsulation in diamond gives the sulfide an impervious host. While a detailed appraisal of subduction-related HSE behavior is beyond the scope of this chapter, some selected examples that are specific to BMS Re–Os isotope systematics are highlighted here.

Gonzáles-Jiménez et al. (2014) reported suprachondritic 187Os/188Os in a population of BMS grains analyzed in situ in a suite of peridotite xenoliths from Calatrava, Spain (Fig. 20). As a consequence of elevated 187Os/188Os, none of the disturbed BMS Re–Os isotopic ratios yielded geologically reasonable TMA or TRD model ages. They attributed these perturbed Re–Os isotope systematics to interaction with fluids originating in a supra-subduction zone environment where dehydration produces a fluid enriched in radiogenic Os (e.g., Brandon et al. 1996; Reisberg et al. 2004; Walker et al. 2002b). Beyer et al. (2004) also reported radiogenic initial 187Os/188Os (0.1282 ± 0.0061) in BMS grains from the Western Gneiss Region of Norway and implicated the introduction of Os derived from sources with higher long-term Re/Os than the peridotites, likely local garnet pyroxenite and eclogite dykes. Similarly, in a study of diamond-hosted BMS grains from the Slave Craton, Canada by Westerlund et al. (2006), sulfides with unsupported radiogenic 187Os/188Os, which yielded future model ages, incorporated evolved radiogenic Os, thought to have been derived from fluids in an Eoarchean subduction setting.

Some independent evidence for a subduction zone origin is preserved in the metal/sulfur ratios of BMS that possess unsupported radiogenic Os isotope ratios. Wang et al. (2009) reported a negative correlation between metal/sulfur ratios and 187Os/188Os ratios of BMS that bore the signature of unsupported radiogenic Os. They suggested that as oxygen is added to a sulfide melt in the oxidizing environment of a subduction zone, magnetite appears as an early crystallizing phase and subsequent MSS will therefore be metal-deficient (Ballhaus et al. 2001). Thus, sulfides that have interacted with subduction-like oxidizing conditions are more likely to have low metal/sulfur ratios than those that do not display evidence for unsupported radiogenic Os. However, an alternative to subduction-related transport of radiogenic Os was recently postulated in the study of Gonzáles-Jiménez et al. (2014), who ascribed suprachrondritic 187Os/188Os and low Re/Os to a carbonate-like metasomatic overprint where volatile-rich fluids with elevated CO2 were the likely source of the radiogenic 187Os/188Os signature (see also Reisberg et al. 2004; Lorand et al. 2004; Powell and O’Reilly 2007; Alard et al. 2011).

Rhenium enrichment

Rhenium enrichment results in elevated Re/Os elemental ratios and, consequently, raised 187Re/188Os (Fig. 19). Where this has occurred recently, i.e., there has been little opportunity for 187Re to produce ingrown 187Os in significant amounts, this may be accompanied by unradiogenic 187Os/188Os. However, the effect on TMA model age calculations will be to decrease the calculated age—potentially producing a “future” age (Fig. 17c)—whereas TRD model ages will be unaffected. Gonzáles-Jiménez et al. (2014) attributed a horizontal trend in one of their Re–Os isotope plots to Re addition. Similarly, Archean model ages observed by Marchesi et al. (2010) in BMS from the Ronda peridotite, southern Spain, are only evident in the TMA age calculations and not in TRD model ages; a feature that does not represent any real magmatic event and indicates the secondary uptake of Re.

In a study of peridotite- and pyroxenite-hosted sulfides from the Penghu Islands, Taiwan, Wang et al. (2009) reported Re–Os isotope systematics of BMS grains that originally formed as residual MSS or by crystallization of sulfide melts after an ancient melting event, which later reacted with percolating metasomatic melts. Wang et al. (2009) suggested that these BMS grains may have been variably affected by one or more of up to three separate metasomatic episodes, which modified the Re–Os isotopic compositions of the sulfides. They attributed positive co-variations on Re–Os isochron style plots to interaction of BMS with an asthenosphere-derived silicate melt, whereas negative co-variations on the same plot were attributed to interaction with a fluid or melt that carried a radiogenic Os signature, but no significant Re (cf. Westerlund et al. 2006 and González-Jiménez et al. 2014, outlined above). In addition, several of the BMS grains analyzed by Wang et al. (2009) showed elevated 187Re/188Os with little or no concomitant increase in 187Os/188Os, which are characteristics of recent Re-addition. In the case of those Penghu Island sulfides where positive correlation trends are apparent in 187Os/188Os–187Re/188Os plots, the apparent “isochron” ages are likely to be the result of mixing of pre-existing BMS with percolating basaltic melts and therefore have no age significance.

Griffin et al. (2002) also commented extensively on the necessity for an addition of unsupported radiogenic Os for many of their cratonic Group 3 BMS grains (broadly analogous to the Type-2 BMS of Alard et al. 2002) from Udachnya, Siberia, and a requirement for a multistage evolution. Critically, it led them to scrutinize the other sulfides that they had analyzed and conclude that these also may have experienced multistage, but not as intensive, metasomatic histories. This resulted in model ages that may at first sight seem plausible but are erroneous and simply mixing trends (cf. Wang et al. 2009). For example, a comparison of TRD and TMA model ages revealed that some of the Group 3 BMS grains yield TMA > 0, but TRD < 0, which implies recent Re addition.

Mixing between primary and metasomatic populations of BMS, or pre-existing BMS grains and sulfide-bearing melts and / or fluids, can occur in mantle much younger than cratonic mantle. Harvey et al. (2010) reported similar, although less complex interactions between BMS associated with melt depletion and secondary, metasomatic BMS in a suite of harzburgites and lherzolites from the French Massif Central. While many of these BMS grains had sub-chondritic Os isotope ratios, their Re and Os concentrations were not consistent with a simple sub-division between Type-1 and Type-2 peridotite-hosted BMS (cf. Alard et al. 2002). For example, most of the BMS grains with sub-chondritic 187Os/188Os ratios had Re abundances (up to 412 μg·g−1) far in excess of their Os concentrations (maximum 41 μg·g−1), consistent with Re addition. For these BMS grains to retain sub-chondritic 187Os/188Os, Re addition must have occurred shortly before or during the entrainment of the host xenolith in the lava that brought it to the surface. Moreover, as in earlier studies of cratonic peridotites (Griffin et al. 2002; Beyer et al. 2004; Wang et al. 2009), not all of the BMS grains with supra-chondritic Os isotope ratios have a correspondingly high Re abundance. BMS grains that have Re abundances as low as 0.01 μg·g−1, but retain suprachondritic Os isotope ratios, can neither be accounted for by melt depletion nor ingrowth of Os. While the radiogenic Os for the Massif Central lherzolites and harzburgites could have been sourced during Variscan subduction, the addition of radiogenic Os from subduction-related fluxing of the Variscan mantle wedge was largely discounted by Harvey et al. (2010) on the basis of insufficient S and modal BMS abundances to accommodate a significant influx of “new” BMS. Harvey et al. (2010) concluded that overall, the Re–Os isotope systematics of these BMS were most likely the result of limited physical mixing, when interstitial melts and melt inclusions were added. Mixing has therefore occurred between originally unradiogenic BMS grains with high Os concentrations and metasomatic BMS grains with high Re/Os and low Os concentrations. An alternative suggestion made by Harvey et al. (2010) was a scenario involving the partitioning of Re into pre-existing BMS from a transient silicate melt, but that would have required a mechanism whereby a formerly enclosed BMS was exposed to a Re-rich percolating melt and then subsequently re-enclosed in a silicate grain.

Given the role that BMS clearly plays in the distribution (and redistribution) of the HSE, it seems likely that information hard won from the analysis of individual BMS grains will pay dividends in the interpretation of bulk-rock Re–Os isotope systematics that range from straightforward (e.g., Alard et al. 2002) to otherwise unfathomable. In several case studies, the Re–Os isotope systematics of BMS have revealed critical information regarding mantle evolution, geodynamic processes and the relationship between the upper mantle and the crust that it is inferred to have produced. Although these mixing and metasomatic scenarios may appear convoluted, there are a finite number of reservoirs from which melts and fluids can interact with pre-existing BMS grains. However, just like the silicate-hosted isotope systems, where early melt depletion events are easily obscured by subsequent metasomatism, a limit must exist in the number of metasomatic events the Re–Os isotope system will be able to “see through” before becoming of little use in determining the history of melt depletion in suites of mantle-derived rocks.

Dating the formation of diamonds

Diamonds are rare minerals in the mantle that crystallize in the deeper portions of the lithosphere and in the convecting mantle as deep as the top of the lower mantle, at pressures as high as ~ 25 GPa (Harte 2010; Shirey et al. 2013; Stachel and Luth 2015). Until the 1980s and the publication of the first accurate diamond ages (Kramers 1979; Richardson et al. 1984), the crystallization ages of diamonds was one of the least understood aspects of diamond petrogenesis. Direct dating of diamond itself using radiogenic isotope systems is not practical currently, and still relies on mineral inclusions because diamond does not contain enough atoms of radiogenically decaying nuclides (e.g., 87Rb, 147Sm, 187Re, 235U, 238U,232Th) to produce enough daughter nuclides to be analyzed at current mass spectrometer sensitivity.

Diamonds can include silicate and / or sulfide phases that either grew synchronously with the diamonds themselves (syngenetic; Sobolev 1977; Meyer 1987; Stachel and Harris 2008) or mineralogically predate diamond growth (protogenetic; e.g., Thomassot et al. 2009). Many early studies involving the dating of inclusions stated that inclusions in diamonds can be clearly identified as being syngenetic with diamond growth based upon (i) their relationship to the internal growth structures of the host diamond, (ii) the relationship between inclusions and the nitrogen aggregation state of the host diamond, and (iii) the morphology of the inclusions themselves (Richardson et al. 1984, 1993; Richardson 1986; Smith et al. 1991; Bulanova 1995; Pearson and Shirey 1999). However, with new crystallographic CT-scans and stable isotopic studies, other workers maintain that most inclusions are pre-existing and become encapsulated by a growing diamond (Anand et al. 2004; Thomassot et al. 2009). In either case, if the mineral inclusion grew at a time close to diamond growth, these inclusions provide an opportunity to date diamonds using radiogenic isotope systems that are present in greater abundance in these inclusions than in the diamonds themselves.

In detail, silicate minerals that were included within the diamond as the diamonds grew (either syngenetic or protogenetic diamond inclusions) can be dated by Pb–Pb (Kramers 1979) and Sm–Nd isotopes (Richardson et al. 1984, 1993; Richardson 1986; Richardson and Harris 1997; see Pearson and Shirey 1999 and Gurney et al. 2010 for summaries) and have yielded Archean and Proterozoic ages, although some recent ages that are similar to kimberlite eruption ages have also been reported (Pearson et al. 1998; Shirey et al. 2001). The low abundance of Sr and Nd in the garnets and pyroxenes used by Richardson and coworkers for age dating, combined with their small size, meant that multiple inclusions were often combined to make a single age estimate. Often, grains were combined not just from a single diamond, but from several tens to hundreds of diamonds (Pearson and Shirey 1999). This necessitates the assumption that all of the silicate growth and encapsulating diamond genesis happened synchronously at a given locality.

Secondary ion mass spectrometry of Pb-isotopes in BMS from discrete diamond growth zones in a diamond from Yakutia, Siberia, suggested that diamond may grow sporadically with intervals of no growth that may extend for 1 Ga (Rudnick et al. 1993b). Inclusions in the centre of the Yakutian diamond generated Pb–Pb model ages of > 2 Ga, whereas inclusions in the outer portions of the diamond yielded much younger ages that approached the age of the kimberlite eruption that brought the diamonds to the surface. A consequence of the uncertainties associated with the Pb isotope evolution of the mantle was that early Pb–Pb model ages (Rudnick et al. 1993b) were associated with large uncertainties. In Siberian diamonds, small BMS grains are frequently found in the innermost zones of diamonds and it has been suggested that these inclusions acted as seeds for the nucleation of the diamonds (Bulanova et al. 1996), although how long these BMS grains exist before the actual growth of the diamond is uncertain. Pearson et al. (1998) use this line of evidence to argue for syngenetic growth of diamond and BMS, which implies that BMS growth was effectively isochronous with the host diamond. Shirey and Richardson (2011) suggested that although there is evidence for silicate inclusions often being 500 to 1500 million years older than the diamond crystallization age, BMS inclusions tend to preserve ages that are only 10–100 million years older than the diamond that hosts them, making BMS inclusions a much more reliable measure of the age of diamond formation.

To improve upon Sm–Nd isochron ages derived from composites of silicate inclusions (e.g., Richardson and Harris 1997), a means of accurately and precisely dating individual BMS inclusions was sought in order to better elucidate the timing of diamond growth. Bulanova et al. (1996), using a proton probe, determined that many diamond-hosted BMS contained substantial Os that, in retrospect, would have been high enough to date with the Re–Os isotope system, either to produce model ages on individual grains or to generate isochron style Re–Os plots with multiple cogenetic BMS grains. A modified version of the Os micro-distillation method of Roy-Barman and Allègre (1995) was used to obtain total procedural blanks approaching 10−15 g (Pearson et al. 1998; Pearson and Shirey 1999). This level of blank, which had previously been precluded by the conventional Carius tube digestion method (Shirey and Walker 1995), permitted the analysis of sulfides containing sub-μg·g−1 to μg·g−1 concentrations of Os (corresponding to an absolute amount of Os ranging from 0.1 to 10 pg). This technique produced the first successful date on a suite of diamonds from Koffiefontein, South Africa (Pearson et al. 1998). Since then this has been the preferred method for the dating of diamond formation (e.g., Pearson et al. 1999a,b; Pearson and Shirey 1999; Richardson et al. 2001, 2004, 2009; Shirey et al. 2001, 2002; Aulbach et al. 2004a,b, 2009a,b,c, 2011; Westerlund et al. 2004, 2006; Smit et al. 2010).

Diamond formation through time

Peridotitic diamond protoliths mostly date from the Paleoarchean and silicate inclusions in these diamonds have the depleted major element composition expected for refractory continental mantle (Richardson et al. 1984; Westerlund et al. 2006). These paleoarchean BMS grains have low Re/Os because they formed in equilibrium with mantle peridotite that was residual after melt extraction, consistent with their observed low Re concentrations (< 10 pg·g−1 Re). Eclogitic BMS, in general, forms from protoliths that have a systematically higher initial Os isotopic composition than peridotitic BMS, implying that some degree of prior Re/Os enrichment must have occurred, i.e., they are derived from metamorphosed basaltic or komatiitic lavas.

Isochron and mantle model ages for peridotitic BMS grains extend to older ages (> 3.2 Ga; Gurney et al. 2010; Shirey and Richardson 2011) than those of eclogitic BMS. To date, no Paleoarchean BMS inclusions with high, eclogitic Re/Os ratios have been found (Shirey and Richardson 2011). Consequently, eclogitic diamonds are only commonplace after 3.0 Ga, which Shirey and Richardson (2011) used to suggest represents the age of the earliest slab subduction and thus the establishment of modern plate tectonics. In their model, eclogitic diamonds usually occur where lithospheric mantle has been subject to continental collision. For example, the collision of two continental blocks of the Kaapvaal craton, the Witwatersrand and Kimberley blocks at 3.1–2.9 Ga (Moser et al. 2001; Schmitz et al. 2004), is suggested to have resulted in underthrusting of oceanic lithosphere and eclogite capture (Shirey and Richardson 2011).

The absence of Paleoarchean eclogite xenoliths and eclogitic BMS in diamonds from cratonic lithospheric mantle suggests that the earliest continental nuclei may have formed by processes that were different to what today are recognized as the Wilson cycle. The absence of a true Wilson cycle in the Palaeoarchean does not rule out the presence of some form of recycling or even shallow plate subduction, but the style of subduction observable today had not been initiated at this point. The earliest detected geochemical signatures consistent with some form of recyling of hydrated oceanic lithosphere date back as far as 3.9 Ga and perhaps even 4.2 Ga, based upon the changing Nb/Th (Jochum et al. 1991) and Th/U ratios in basaltic and ultramafic rocks (Collerson and Kamber 1999; Zartman and Richardson 2005), and their trace and major element systematics (O’Neil et al. 2011). Crustal rocks with a signature consistent with derivation from depleted mantle (ɛNd : ɛHf = 1:2) became apparent after 3.6 Ga (Shirey et al. 2008). The appearance of subduction-related imprints on the stable keels of cratons supports the hypothesis that the generation of continental crust was progressively depleting the mantle from 3.6 Ga onwards. By the Neoarchean (after ca. 2.8 Ga), recycling processes more strongly resembled the style of subduction consistent with the Wilson cycle (Shirey et al. 2008).

The age of the continental lithospheric mantle and the assembly of its domains

The timing of diamond formation and the conditions of their formation are invaluable in unravelling the series of events that together resulted in the assembly of the cratons. Moreover, the geochronological information preserved in BMS grains that become isolated within the SCLM as a result of melt depletion can help to constrain the timings of melt depletion and the inherent link between the formation of continental crust and the stabilization of the SCLM. With specific regard to mantle BMS geochronology, three cratons in particular have been studied in detail: the Kaapvaal craton, South Africa, the Siberian craton, Russia, and the Slave Craton, Canada. A detailed discussion of the HSE systematics of cratonic mantle is covered in depth in Aulbach et al. (2016, this volume) and only the role of BMS in the determination of cratonic structure and history will be briefly discussed here.

The history of the Kaapvaal craton, from the perspective of BMS inclusions in diamonds, was summarized by Shirey et al. (2004) and more recently by Aulbach et al. (2016, this volume). Briefly, the southern African cratonic keel, comprising the Kaapvaal–Zimbabwe cratons, separated by the Limpopo mobile belt, is underlain by Archean mantle peridotite and subordinate quantities of eclogite, which host multiple generations of diamonds of Archean to Proterozoic age (Kramers 1979; Richardson et al. 1984, 1993, 2001; Navon 1999; Pearson and Shirey 1999; Shirey et al. 2002). Some parts of the craton are underlain by portions of the mantle keel that probably relate to the generation of the oldest crust observed in the Kaapvaal. This hypothesis in entirely consistent with the numerous bulk-rock Re–Os model ages derived from xenolith studies of the region (e.g., Walker et al. 1989; Carlson et al. 1999, 2000; Janney at al. 2010) and more recently in situ Re–Os model ages on silicate-hosted BMS grains derived by LA MC ICPMS (e.g., Griffin et al. 2003a,b).

Of the suites of eclogitic diamonds associated with the Kaapvaal craton, four (De Beers Pool, Jwaneng, Koffiefontein, and Orapa) preserve Archean ages that overlap the average Neoarchean Re–Os mantle model ages of cratonic peridotite (Carlson et al. 1999; Irvine et al. 2001; Shirey et al. 2001). The eclogitic BMS grains form an array in 187Re/188Os–187Os/188Os space that highlights two general groups of ages. A large number of inclusions from the De Beers Pool, Jwaneng and Orapa plot on a 2.9 Ga isochron (Richardson et al. 2001), with the remainder displaying more scatter (Shirey et al. 2001). At Koffiefontein, only two inclusions correspond to a ca. 2.9 Ga age, while the majority yield much younger ages of 1 Ga (Pearson et al. 1998). The younger Proterozoic diamond ages vary from locality to locality within the Kaapvaal (Pearson et al. 1998; Richardson et al. 2001; Shirey et al. 2004).

The sequence of events proposed for the formation of the Kaapvaal craton begins with formation of the nucleus of the craton at ca. 3.7–3.3 Ga (Aulbach et al. 2009a). This would have been accompanied by hot, large melting intervals, such as those inferred to have been responsible for the 3.5 Ga Barberton komatiites. This process would have left a highly depleted residue with refractory olivine and garnet compositions (Stachel et al. 2003; Stachel and Harris 2008) and a sulfide-free residuum (e.g., Keays 1995). In this hot, volatile-poor peridotitic environment, diamond formation would be difficult. Percolation of S-undersaturated melts during secondary processes at this time may also have worked to strip the SCLM of its sulfide content (Reisberg et al. 2005). Those diamonds that could form may have been able to trap refractory silicate inclusions, but the sulfur-undersaturated nature of the protolith may have impeded the encapsulation of large numbers of peridotitic-BMS grains despite the abundance of peridotitic silicate inclusions (Aulbach et al. 2009a). Continental collision of the eastern and western blocks of the Kaapvaal craton at 2.9 Ga allowed for incorporation of volatile bearing and sulfur rich oceanic lithosphere into the mantle keel peridotite. BMS added at this time were eclogitic and preferentially added in the west, explaining both the timing and distribution of 2.9 Ga eclogitic diamond growth (Shirey et al. 2004a, 2013; Aulbach et al. 2009a). Subsequent repeated episodes of sulfide re-introduction into the lithosphere accompanying metasomatism and marginal subduction are documented by peridotitic BMS grains in mantle xenoliths and xenocrysts (Griffin et al. 2004) and multiple episodes of Proterozoic diamonds with eclogitic sulphides (e.g., Pearson et al. 1998; Richardson et al. 2004; Aulbach et al. 2009b). The few examples of peridotitic BMS-containing diamonds in the Kaapvaal craton have proved to be young. (Pearson et al. 1998; Aulbach et al. 2009b).

Archean peridotitic BMS inclusion-bearing diamonds are more abundant in the central Slave craton, where diamond formation appears to have been coeval with lithosphere formation (Aulbach et al. 2009a, 2011). The Slave craton comprises a juvenile eastern and an ancient central to western domain that may have become amalgamated as a result of east-oriented subduction at ca. 2.7 Ga ago (Kusky 1989; Bleeker et al. 1999a,b). The formation of shallow lithosphere beneath the Slave craton is thought to have commenced through accretionary processes as early as ca. 3.5 Ga (Aulbach et al. 2011). Several peridotitic diamond-hosted BMS grains derived from Diavik plot along a 3.3 Ga isochron, whereas 11 similar inclusions yield an isochron age of 3.52 Ga i.e., from the Palaeoarchean (Westerlund et al. 2006). This early generation of peridotitic BMS that is predominant in the Slave craton is completely missing in the Kaapvaal craton. In contrast to the Kaapvaal craton, the less depleted deep lithospheric mantle beneath the central Slave craton is proposed to have formed as a result of plume-related underplating, but with a lower overall degree of melt depletion because of the presence of a pre-existing lithospheric mantle lid. This smaller degree of melt depletion did not result in a complete exhaustion of the SCLM sulfur supply and hence permitted the formation of peridotitic BMS inclusions coeval with diamond formation (Aulbach et al. 2009a). However, like the Kaapvaal craton, the subsequent evolution of the Slave craton has been dominated by a series of collision and subduction-related events recorded in the multiple generations of diamond-hosted eclogite- and pyroxenite-hosted BMS (Aulbach et al. 2009a,c).

The relationship between the age of the SCLM and the overlying crust

In regions where the overlying crustal record is preserved, ages derived for crust generation from the crust itself can be compared to peaks in model ages preserved in SCLM-derived BMS grains derived from peridotite xenoliths (e.g., Pearson et al. 2007; González-Jiménez et al. 2013, 2014). Where the crustal record is imperfect, for example when ancient crust has been eroded or has become obscured through burial or tectonism, the geochronological evidence preserved in relative probability plots of BMS model ages still provides valuable information regarding the history of crust formation and SCLM stabilization. Peaks in model ages on relative probability diagrams of Re–Os TRD or TMA (e.g., Figs. 18, 20) are attributed significance as corresponding to the stabilization of SCLM, which may have been coupled to major crust formation events. Both bulk-rock and BMS Re–Os ages frequently show a correspondence between the crust and the underlying mantle, sampled as xenoliths, from which it is inferred to be derived (Pearson et al. 1995a,b,2002; Handler et al. 1997; Carlson et al. 1999; Chesley et al. 1999; Menzies et al. 1999; Hanghøj et al. 2001; Irvine et al. 2001,2003; Lee et al. 2001; Griffin et al. 2002; Schmidt and Snow 2002; Carlson and Moore 2004; Marchesi et al. 2010; González-Jiménez et al. 2013, 2014). However, there are some examples where underlying mantle is apparently older than the overlying crust (Parkinson and Pearce 1998; Peslier et al. 2000; Handler et al. 2003; Smit et al. 2010).

In the Udachnaya kimberlite, Siberia, Griffin et al. (2002) used sulfide model ages to construct a relative probability plot to date the formation of the Siberian craton, though the BMS enclosed in olivine macrocrysts are complex in nature. Of the 53 BMS grains that had 187Re/188Os of < 0.07 (the cut-off chosen by Griffin et al. 2002, as ensuring a model age that was not affected by Re-addition or mixing with Re-rich secondary sulfides) 45 of the BMS grains produced TMA ages between 2.5 and 3.6 Ga, and 35 of the grains gave TMA of > 2.8 Ga. This lead Griffin et al. (2002) to conclude that the Siberian SCLM formed during the period 3–3.5 Ga, with a peak in lithosphere stabilization at around 2.9 Ga, and little evidence for significant additions to the lithosphere afterwards. However, the recent Re–Os isotope study of Yakutian diamond-hosted BMS by Wiggers de Vries et al. (2013) demonstrated that the assemblage of the Siberian craton is more protracted than the initial 2.9–3.5 Ga growth phase suggested by Griffin et al. (2002). The ages obtained by Wiggers de Vries et al. (2013) for the different diamond populations demonstrate two major periods of eclogitic and lherzolitic Yakutian diamond formation at ca. 2.1–1.8 Ga (Mir, 23rd Party Congress and Udachnaya kimberlites) and at ca.1.0–0.9 Ga (Mir and 23rd Party Congress kimberlites). These correspond to the collision between different terranes of the Siberian Craton during the formation of a Palaeoproterozoic supercontinent at ca. 2.0–1.8 Ga and accretion leading to formation of the supercontinent Rodinia ca. 1.1 Ga. The very radiogenic initial Os isotope ratios of the eclogitic and lherzolitic BMS grains (187Os/188Os = 0.14–2.22) imply the incorporation of radiogenic Os from subducted oceanic lithosphere, a conclusion that is also relevant to the Kaapvaal and Slave cratons.

However, more recently, the age of the Siberian craton as a whole has been called into question, as the vast majority of Re–Os dates for the region have been derived from a single kimberlite province, i.e., Udachnaya. Doucet et al. (2015) argue that although the Siberian SCLM contains components of material that retain an Archean signature, a wide range of Re-depletion ages have been derived (3.4–1 Ga and commonly ≤ 2 Ga for peridotite xenoliths), consistent with craton-forming activity not limited to the Archean. Using data derived from a second Siberian kimberlite, Obnazhennaya, Ionov et al. (2015) have recently proposed that the lithospheric mantle beneath the Siberian craton was formed in at least two events, one in the late Archean and the other in the Paleoproterozoic.

An example of the utility of relative probability plots in coupling melt depletion and the crustal growth history of non-cratonic SCLM was recently published by González-Jiménez et al. (2013). Here, using a similar approach to that employed by Griffin et al. (2002) for olivine-hosted BMS, they determined TRD of BMS grains from peridotite xenoliths recovered from the Calatrava volcanic field, Spain. Their study revealed that episodes of mantle magmatism and/or metasomatism in the Iberia microplate were linked to supercontinent assembly and/or breakup at ca. 1.8, 1.1, 0.9, 0.6, and 0.3 Ga (Fig. 20). In addition, they found that the mantle and crust constituting the Iberian microplate have coexisted since at least Paleozoic–Proterozoic time. Rhenium-depletion ages of 68 carefully screened BMS grains, either possessing 187Re/188Os < 0.07 (n = 12) or 187Re/188Os > 0.07 but with uncorrelated Re/Os, and hence no systematic ingrowth of 187Os (n = 56), produced peaks on a relative probability plot coinciding with major tectonic events recorded in the crust (Fig. 20). The formation of the Iberian lithosphere at ca. 1.6–2.0 Ga coincides with crust-mantle differentiation that occurred during the assemblage of the Columbia supercontinent (ca. 1.8–1.5 Ga). Peaks in TRD at 1.3–0.9 Ga coincided with the amalgamation of Rodinia at ca. 1.25 Ga and its breakup at ca. 0.75 Ga.

Peaks in TRD at 0.6 Ga and 0.3 Ga were interpreted by González-Jiménez et al. (2013) as representing the beginning of continental collisions to form Gondwana and the initiation of the breakup of Pangea respectively (Orejana et al. 2009). These interpretations should be treated with some caution given the caveats associated with interpreting Phanerozoic TRD in peridotites. The interpretation of the older sulfide TRD are broadly consistent with the earlier study of Marchesi et al. (2010), which examined individual BMS grains recovered from the Ronda orogenic massif, Spain. This earlier study also considered the possibility that residual BMS grains with Os model ages of ca. 1.2–1.4 Ga may represent the reworking of older SCLM in the Mesoproterozoic. Critically, if the massif became part of the lithosphere at about 1.2–1.4 Ga, the preferred interpretation of whole-rock Os isotopic data (Reisberg et al. 1991; Reisberg and Lorand 1995), then the residual BMS grains with Os model ages of ~ 1.6–1.8 Ga from Marchesi et al. (2010) could be interpreted as inherited grains that survived within the convecting mantle for hundreds of millions of years. The studies of Marchesi et al. (2010) and González-Jiménez et al. (2013) agree that the Proterozoic Os model ages exhibited by BMS grains in the Iberian SCLM tend to correspond to different stages of generation of crust, which was later recycled in the Gondwana supercontinent.

Evidence for ancient lithospheric mantle pre-dating the oldest exposed cratonic crust has also been found in the Kimberley craton, north-western Australia in BMS-bearing diamonds recovered from the ca. 20 Ma Ellendale lamproite pipes (Smit et al. 2010). Lherzolitic diamond–hosted BMS grains yield an age of 1426 ± 130 Ma, with an initial 187Os/188Os ratio of 0.1042 ± 0.0034. The upper limit on the 187Os/188Os initial ratio suggests a Re depletion age of 2.96 Ga, indicating the presence of SCLM beneath Ellendale since at least the Mesoarchaean. This is supported by independent evidence for the presence of deep SCLM below the King Leopold Orogen and the Kimberley craton by seismic tomography (van der Hilst et al. 1998; Kennett 2003; Fishwick et al. 2005) and cool cratonic geotherms for the West Kimberley province (Griffin and Ryan 1995). Archean SCLM also extends to the south-east of the Kimberley craton, with mantle xenolith TRD ages as old as the Neoarchean (2.2–2.9 Ga; Luguet et al. 2009) or possibly even older, as suggested by an imprecise Re–Os isochron age of 3.4 Ga (Graham et al. 1999). However, the age of the mantle below the Kimberley craton is significantly older than the oldest exposed crust within the North Australian Craton (ca. 2 Ga; Page et al. 1995; Tyler et al. 1999; Worden et al. 2008). Therefore, despite early Proterozoic convergent margins and new crust formation (Tyler et al. 1999; Griffin et al. 2000), remnants of the underlying continental mantle were preserved (Smit et al. 2010). This indicates that crust and mantle in the Kimberley region were decoupled (Luguet et al. 2009), with preservation of the pre-existing lithosphere during accretion (Smit et al. 2010).

The generation of BMS grains during ancient melt-depletion events that no longer represent any known event recorded in the overlying crust is the most likely explanation for some populations of BMS. These could either be related to domains where ancient crust is not exposed, either because it has been removed or because newer mantle, containing an old heterogeneous component has become part of the lithosphere. This is particularly pertinent to the subject of osmium and lead isotope heterogeneity in the oceanic mantle.

The inherent heterogeneity within the oceanic mantle

Isotopic studies of oceanic samples have long demonstrated the occurrence of isotopic variations in the mantle that must have been created by geologic processes much older than the 180 My age of the oldest ocean floor (e.g., Gast et al. 1964; Sun and Hanson 1975; O’Nions et al. 1977). In addition, compositional heterogeneities in the oceanic mantle at length scales of thousands of kilometres (e.g., Hart 1984; Meyzen et al. 2007) to sub-km scales have often been reported (e.g., Shirey et al. 1987; Dosso et al. 1999; Meibom et al. 2002; Standish et al. 2002; Warren et al. 2009). These observations contrast with studies of the SCLM, where Re–Os ages for the mantle often show a correspondence to the age of the overlying crust (e.g., Griffin et al. 2002; Pearson et al. 2002), though examples exist of the SCLM having ages older than the overlying crust (e.g., Peslier et al. 2000; and discussion above).

Although the Pb isotope composition of the oceanic mantle has been of interest to geochemists for nearly half a century (e.g., Allègre 1969), the role that BMS plays as a reservoir for mantle Pb has been less well explored, despite the early realization that Pb isotope heterogeneity may be present at a small scale (e.g., Saal et al. 1998, and more recently Maclennan 2008). Recent interest in the Pb elemental and isotopic composition of mantle BMS stems from a potential solution to the first Pb paradox (Allègre 1969), which is the observation that global basalts have ubiquitously radiogenic Pb isotope compositions (Hofmann 1997) that fall at much higher 206Pb/204Pb than the geochron. If estimates of the bulk silicate Earth Pb isotope composition are correct (i.e., broadly chondritic for U, Th, and Pb; e.g., Palme and Jones 2003) and that the U/Pb ratio was set at ~ 4.50 Ga when core formation ended (Kleine and Rudge 2011), then there must be a reservoir of unradiogenic Pb. Observations of Pb isotopes in both bulk-rock peridotite and individual BMS grains suggest that the mantle may represent one of the reservoirs for unradiogenic Pb based upon isotopic measurements of peridotites and their minerals (e.g., Malaviarachchi et al. 2008; Burton et al. 2012; Warren and Shirey 2012), and predictions based upon trace element depletion (Godard et al. 2005; Kelemen et al. 2007; Hanghøj et al. 2010).

The first hint that the unradiogenic Pb reservoir was at least partially hosted in the mantle came from a Pb isotope study of the Horoman peridotite (Malaviarachchi et al. 2008). Here, bulk-rock plagioclase lherzolites were analyzed for their Pb isotope ratios and some plotted to the unradiogenic side of the 4.5 Ga geochron. Among abyssal peridotites, one clinopyroxene analysis out of three samples analyzed has yielded an unradiogenic composition (Warren et al. 2009). To date, further Pb isotope data on pyroxene mineral separates from abyssal peridotites are not available due to the analytical difficulties of working at Pb concentrations of ~ 10 ng·g−1.

More recently, Burton et al. (2012) and Warren and Shirey (2012) measured Pb isotope compositions in individual BMS grains extracted from abyssal peridotites recovered from the MAR, Gakkel Ridge and SWIR (Fig. 16). For the MAR, Burton et al (2012) found that 206Pb/204Pb ratios ranged from 18.4063 to 18.4672 for interstitial BMS, i.e., all radiogenic compared to the geochron. In contrast, enclosed BMS ranged from 16.2276 to 17.7839, i.e., extending to values considerably less radiogenic than the geochron. Similarly, Warren and Shirey (2012) reported BMS 206Pb/204Pb ratios ranging from 17.034 (less radiogenic than the geochron) to 19.640 (considerably more radiogenic than the BMS grains of Burton et al. 2012). As seen in Os isotope ratios of abyssal peridotite BMS grains (e.g., Alard et al. 2005; Harvey et al. 2006), Pb isotope ratios of abyssal peridotite BMS show little systematic covariation with basalt Pb isotopic compositions from the area of the ridge from which they were recovered. For example, BMS grains from Gakkel abyssal peridotites span the full range of Gakkel Ridge basalt compositions (Warren and Shirey 2012; Blusztajn et al. 2014), but do not show the break in Pb isotope ratios at 15° E observed among basalts (Goldstein et al. 2008).

The recent discoveries of ancient Pb isotope signatures preserved in BMS grains recovered from abyssal peridotites (Fig. 16) suggests that at least some of the missing unradiogenic Pb resides in the upper mantle. This conclusion is supported by the observation that the Pb isotope composition of bulk-rock peridotites from a variety of continental settings also spans both sides of the geochron (Warren and Shirey 2012). In addition, Os and Pb isotopic compositions are correlated (Fig. 21), with both isotope systems yielding similar model ages. For the combined dataset from the MAR, Gakkel and SWIR, the Pb isotopic composition of mantle BMS grains aligns along the ~2 Ga data array for oceanic basalts and peridotite mineral separates, while the Re–Os datasets plot along the 2 Ga Re–Os reference isochron (Harvey et al. 2006; Burton et al. 2012; Warren and Shirey 2012). This “age”, combined with the observed Pb–Os isotopic correlation (Fig. 21) and a very small (≪ 1 km) length scale of peridotite compositional variability can be explained either by crustal extraction at 2 Ga or an average mixing age for the mantle. A single depletion event that can explain the isotope datasets from three widely separated ridges (Gakkel, SWIR and MAR) seems implausible and is inconsistent with episodes of peak crustal formation at 2.7, 1.0, 0.6, and 0.3 Ga (Condie and Aster 2010). In contrast, the mixing of ancient components back into the mantle by subduction has long been suggested as an explanation for the heterogeneity of the oceanic basalt array (e.g., Chase 1981; Hofmann and White 1982; Zindler and Hart 1986; Kellogg et al. 2007).

Analysis of the isotopic composition of individual BMS grains from abyssal peridotites allows mantle heterogeneity to be probed at very small length scales, which is particularly powerful when Pb and Os isotopes can be analyzed in the same grain (Fig. 21; Warren and Shirey 2012). Base metal sulfide grains recovered from the same drill core (Harvey et al. 2006; Burton et al. 2012) or dredge haul (Alard et al. 2005; Warren and Shirey 2012; Blusztajn et al. 2014) have large magnitude Os and Pb isotopic variations, demonstrating that these can be preserved in the oceanic mantle at length scales of significantly less than 1 km. For example, the Os and Pb isotope heterogeneity in BMS from serpentinized peridotite, reported by Harvey et al. (2006) and Burton et al. (2012) came from a single 6 cm section of quarter-core from Hole 1274a on the 15° 20′ N Fracture Zone, MAR. Blusztajn et al. (2014) found considerable Pb isotope variation among BMS grains in some individual samples from the Gakkel Ridge and SWIR, including one sample with BMS grains that cover 25% of the Pb isotope range of oceanic basalts. Pyroxenes from abyssal peridotites have also been used to show considerable Nd (e.g., Cipriani et al. 2004), Sr (e.g., Warren et al. 2009) and Hf (e.g., Stracke et al. 2011) isotope variability, but these isotopes so far have only been measured in mineral separates and not individual grains.

The scale and magnitude of oceanic Pb and Os isotope heterogeneity will remain poorly constrained, especially compared to current knowledge of the SCLM, until a larger dataset for abyssal peridotite BMS becomes available. However, what is clear is that the composition of basalts alone cannot be relied upon to constrain either the extent of mantle depletion or the degree of small scale variations in the mantle from which they were derived. Sulfides indicate the occurrence of Os and Pb heterogeneities in the upper mantle with extreme compositions that are not resolvable in pooled basaltic melts. Either these anomalies reside in BMS within domains that are too refractory to contribute to the melting process or their size is such that more voluminous melts smooth out the most extreme isotopic signatures of the source (e.g., Warren and Shirey 2012).

Base metal sulfides in mantle peridotite and diamonds preserve evidence for a variety of processes, including melt depletion, re-fertilization, metasomatism, and alteration. The ability of sulfides in mantle rocks to preserve different processes has led to an unraveling of these different generations of processes in bulk mantle samples. Combined with new advances in analytical sensitivity and spatial resolution, an improved understanding of the geological history of mantle samples on all scales is being achieved. The emergence of improved amplifier resistors for TIMS and MC ICP-MS instruments has recently seen a leap in possible analytical resolution for silicates (e.g., Koornneef et al. 2015; Sarkar et al. 2015) and meteoritic material (Peters et al. 2015). It seems likely that these methods will be applied to problems where high precision analyses of micro-metric BMS grain and / or alloys are required in the near future.

Base metal sulfides hosted in peridotites and pyroxenites of the oceanic lithosphere carry a memory of melt depletion and melt infiltration that is an evident effect of the plate tectonic cycle of ridge melting and slab subduction. For the first time, isotopic ages can be anchored to their mineral hosts. By looking at primary BMS grains, we can peer through the prevalent alteration that masks much of the work done on the bulk-rock scale. Future work will increase the now sparse sites on the ocean floor that have been studied and will allow us to approach a global understanding of specific episodes of depletion and melting, with the ultimate goal to relate these episodes to specific plate geometries and geodynamic cycles.

Geochronology on BMS grains in diamonds has achieved the long-held goal of producing ages for every diamond locality that carries amenable BMS inclusions. With the caveat that all diamonds in kimberlite are xenocrysts, age arrays that have geological significance emerge from the diamond suites in many well-mined kimberlites. These arrays are being used to identify when portions of the deep continental lithosphere have been added and reworked, giving a new and deep dimension—one not evident from the surface—to the evolution of the continents. Future work will more solidly anchor these deep mantle lithosphere ages to crustal tectonothermal histories, so that we can achieve a thorough picture of crustal differentiation through time and the nature of the earliest stable crust forming processes.

The authors would like to thank Ambre Luguet, Claudio Marchesi, and Akira Ishikawa for thought-provoking reviews of this chapter and James Day for editorial handling. JH was supported by a NERC Advanced Research Fellowship ( NE/J017981/1), a Blaustein Visiting Professorship at Stanford University, and a visiting investigator appointment at the Carnegie Institution, Washington. JMW received support from the US National Science Foundation through grant OCE-1434199; SBS received support from NSF grant EAR-1049992.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.