- © 2016 Mineralogical Society of America
Tectonically emplaced mantle rocks, such as ophiolites, abyssal peridotites, and orogenic peridotite massifs, provide a principle constraint on the composition of and processes in the Earth’s upper mantle (Bodinier and Godard 2003). In the past, these ‘mantle tectonites’ have sometimes received different names because their history and origin has been unclear. Mantle tectonites are now understood to reflect a range of geologic environments regarding their emplacement and their origin (e.g., Dilek and Furnes 2014). The advantage of these rocks compared to mantle xenoliths is the large-scale exposure of textural and compositional relations between different rock types that can be used to identify processes such as melting, magma or fluid transport, chemical reactions, mixing or deformation at a range of spatial scales. A disadvantage of most mantle tectonites is that they commonly display substantial chemical modification of some elements, resulting from widespread serpentinization at low temperatures. In some cases, this may also affect abundances of several of the highly siderophile elements (HSE: Re, Au, PGE: Os, Ir, Ru, Rh, Pt, Pd), however, this can be tested by comparison with unaltered rocks of similar composition. As is discussed in Luguet and Reisberg (2016, this volume), Harvey et al. (2016, this volume) and Aulbach et al. (2016, this volume), peridotite xenoliths have their own alteration issues regarding sulfides and chalcophile elements.
Numerous studies have obtained Os isotope and/or highly siderophile element abundance data on many different types of mantle tectonites. Some of these studies have focused on large-scale chemical and isotopic variations, others on grain size-scale compositional variations to understand small-scale distribution processes. These studies have, together, significantly advanced the understanding of the processes that fractionate the HSE in the mantle at different spatial scales and have provided insights into the behavior of sulfide in the mantle—the phase that typically hosts the vast majority of the strongly chalcophile elements, including the HSE. Osmium isotopes and Re–Os model ages have provided tools to directly date melting of mantle tectonites and have changed views on the efficacy of mixing and melting processes in the mantle.
Here, we review these advances, which have mostly taken place in the past 15 years, aided by new developments in isotope dilution and ICP-MS based techniques and the application of in situ laser ablation ICP-MS. First, we provide a brief summary of the current views about geodynamic environments of different mantle tectonites. Work on Os isotopes and HSE abundances in mantle tectonites of different geodynamic settings will be reviewed subsequently. In the Discussion, we summarize the views on processes and chemical behavior of Os isotopes and the HSE in mantle tectonites.
BREVIA OF CONCEPTS, TERMINOLOGY, AND ANALYTICAL CAVEATS
In this chapter, we normalize isotopic abundance ratios to 188Os following currently accepted conventions: 187Os/188Os = measured 187Os/188Os, 186Os/188Os = measured 186Os/188Os, 187Os/188Osi = initial 187Os/188Os at a given age. Up to the late 1990s some workers used the 187Os/186Os ratio to compare variations in 187Os in natural materials. After it became clear that some minerals and rocks also display variations of radiogenic 186Os from the decay of 190Pt (Walker et al. 1997; Brandon et al. 1998, 1999), it was suggested to use 188Os as the stable reference isotope (e.g., Shirey and Walker 1998). In the present work, conversion of early 187Os/186Os and 187Re/186Os data to 187Os/188Os and 187Re/188Os, respectively, was performed by multiplication with a value of 186Os/188Os = 0.1203. This value for 186Os/188Os was commonly obtained by early less precise and accurate isotopic ratio determinations. Later high-precision measurements of Os isotopic compositions of mantle-derived rocks using N-TIMS and faraday cup detection have yielded lower 186Os/188Os for the Earth’s mantle and the bulk silicate Earth (0.119838 ± 0.000003, 2 s.d., Brandon et al. 2006). The measurement of Os isotopic ratios via OsO3− ions requires that the raw data is corrected for interferences produced by the minor isotopes of O. Mass-dependent fractionation of Os isotopes in nature and during analytical procedures is commonly corrected assuming 192Os/188Os = 3.0827 (Luck and Allègre 1983, Shirey and Walker 1998). High-precision Os isotopic data require more elaborate measurement and correction protocols (e.g., Brandon et al. 2005a, 2006, Luguet et al. 2008a; Chatterjee and Lassiter 2015).
The deviation of 187Os/188Os of a sample from an arbitrary ‘average’ chondritic composition (present 187Os/188Os = 0.12700, 187Re/188Os = 0.40186) at age T is given as γOst and was calculated using the equation and parameters given in Shirey and Walker (1998). The decay constant of 187Re used for calculations is 1.666 × 10−11 year−1 (Smoliar et al. 1996; Selby et al. 2007). The average chondritic 187Os/188Os and 187Re/188Os have no specific meaning other than as a reference for comparing different materials. Initial 186Os/188Os of samples were calculated using λ190Pt = 1.48×10−12 year−1 (Brandon et al. 2006). Rhenium-depletion model ages TRD (Ch) and Re–Os model ages TMA (Ch) have been defined previously relative to a chondritic evolution model (Ch) using the parameters mentioned above (Walker et al. 1989; Shirey and Walker 1998). Alternatively, these model ages may be calculated relative to the Re–Os evolution of the primitive-mantle model composition, e.g., TMA (PM). The primitive mantle has a slightly higher 187Os/188Os (0.1296) and 187Re/188Os (0.4346) than the ‘average’ chondrite reference values (Meisel et al. 2001).
Normalization of concentration data
In the literature, normalizations of HSE abundances in mantle rocks are sometimes performed relative to mean abundances in CI chondrites using data from compilations (e.g., Anders and Grevesse 1989; Horan et al. 2003; Lodders 2003). One disadvantage of this approach is that the HSE composition of the Earth’s mantle (and of the bulk Earth) likely does not match CI chondrites (Walker et al. 2002a,b; Horan et al. 2003; Becker et al. 2006; Fischer-Gödde et al. 2010, 2011). However, it does have the advantage of using a measureable reference frame for normalization. An alternative approach to assess igneous fractionation of the HSE in mantle and crustal rocks is to normalize to a primitive-mantle model composition (PM, sometimes also referred to as primitive upper mantle, PUM and bulk silicate Earth, BSE), and to arrange elements according to their incompatibility as it is commonly performed for lithophile elements (e.g., Hofmann 1988). The HSE concentrations in PM used for normalization and the sequence of HSE in normalized concentration diagrams are those given in Becker et al. (2006) and Fischer-Gödde et al. (2011). This theoretically has the advantage of providing comparison with the primitive mantle composition. Alternatively, a composition of ‘average depleted spinel lherzolites’ (based on mantle xenoliths and tectonites) has been defined in the literature for comparative purposes (Pearson et al. 2004). Here we use, in different situations to reflect different aims, both normalization to primitive mantle and to ‘average’ chondrite values calculated with equal weighting from ordinary, enstatite and carbonaceous chondrites, from data compiled in Walker (2009) and from Fischer-Gödde et al. (2010). We also use both logarithmic and linear scales to best display the variations present in each particular figure. The sequence of HSE in normalized concentration diagrams of terrestrial rocks commonly follows the sequence of increasing enrichment in basalts and komatiites (i.e., Os < Ir < Ru < Rh < Pt < Pd < Au < Re ≈ S), which is similar, but not always identical, to the depletion in many peridotites. Elemental patterns in some peridotites, that differ from this general depletion sequence, reflect re-enrichment in Re, Au, Pd, and multi-stage histories.
Precision and accuracy of concentration data and analytical issues
Previous studies have indicated that some of the early analytical techniques used to determine HSE abundances or Re–Os systematics did not always produce complete recovery of Ir, Os, and Ru, even at test portion masses of > 10 g (e.g., Shirey and Walker 1998; Meisel and Moser 2004; Becker et al. 2006; Lorand et al. 2008; Meisel and Horan 2016, this volume). In the discussion of processes that fractionate the HSE, we will primarily focus on more recent data that have been obtained either by Carius tube digestion at enhanced temperatures (T > 230 °C), by high-pressure asher (typically > 300 °C), or by improved NiS fire-assay techniques (Gros et al. 2002). If isotopic ratios were analyzed by ICP-MS or, in the case of Os, N-TIMS, these methods yield combined analytical uncertainties (1 s.d.) of concentrations that may range between better than a few % for Re and 15 % for Au for well-homogenized whole rock powders of lherzolites and test portion masses of about 2 g or more (Meisel and Moser 2004; Pearson et al. 2004; Becker et al. 2006; Lorand et al. 2008; Fischer-Gödde et al. 2011). Heterogeneity of abundances of carrier phases of the HSE in powders of some peridotites is a well-known problem (‘nugget effect’). In addition to the nugget effect, complete and reproducible digestion of refractory platinum-group element minerals (PGM) in some harzburgites or dunites, may represent a challenge. Incomplete digestion of refractory alloy phases may bias ratios of Os, Ir, and Ru. For further details, see Meisel and Horan (2016, this volume).
HIGHLY SIDEROPHILE ELEMENTS IN MANTLE TECTONITES FROM DIFFERENT GEODYNAMIC SETTINGS
Summary of mantle tectonites and their geodynamic settings
Mantle tectonites include peridotite sections of ophiolites, abyssal peridotites and orogenic peridotites that often, but not exclusively, occur in orogenic belts (also known as peridotite massifs, alpine or alpinotype peridotites). These different mantle tectonites can be distinguished by their geodynamic setting, and associated emplacement history and pressure–temperature (P, T) evolution, but also by their chemical composition. Most of the rocks concerned record a relatively simple cooling history from lithospheric mantle conditions (T of 1000–1300 °C and P of the garnet, spinel, or plagioclase lherzolite stability field) to some lower T and P equivalent to crustal conditions. Owing to their origin from in situ lithospheric or asthenospheric mantle conditions, these rocks are sometimes also referred to as ‘high-temperature peridotites’. In contrast, ‘low-temperature’ orogenic peridotites are former high-temperature peridotites that have been subducted as part of a package of crustal rocks in collision zones (e.g., the Alpe Arami peridotite, Nimis and Trommsdorff 2001; peridotites of the Western Gneiss region, Norway, Brueckner et al. 2010; Zermatt-Saas ophiolite, Barnicoat and Fry 1986). Low-temperature peridotites were partially re-equilibrated at high P/T conditions, but in some cases, this partial re-equilibration is hardly noticeable and chemical and textural features inherited from the high-temperature history of the peridotites predominate (e.g., at the Lanzo peridotite massif; Pelletier and Müntener 2006).
Improved understanding of the geodynamic evolution of passive continental margins (Dilek et al. 2000), the transition to ocean spreading and the role of ocean spreading rate in the lithological composition of the oceanic crust (Dick et al. 2006) have led to improved interpretations of the origin and geodynamic environments of mantle tectonites and ophiolites (Dilek et al. 2000; Dilek and Furnes 2014). It is now understood that high-temperature peridotite tectonites derived from continental lithospheric mantle may be exhumed during slow extension of continental lithosphere and the formation of sedimentary basins or small ocean basins. Well-known examples are the island of Zabargad in the Red Sea (Brueckner et al. 1988), the Pyrenean peridotite bodies in southern France (Vielzeuf and Kornprobst 1984; Bodinier et al. 1988), the peridotite bodies of NW Italy (Ivrea-Verbano Zone, Lanzo; Ernst 1978; Sinigoi et al. 1983; Shervais and Mukasa 1991; Mazzucchelli et al. 2009) and some of the mantle tectonites in the Alps and in Italy that sometimes have been referred to as ‘ophiolites’ (for instance the External Ligurian peridotites; Rampone et al. 1995). Some mantle tectonites were exhumed in oceanic environments as indicated by their alteration and association with ophicalcitic breccias, basalts, gabbros and cherts. Such rocks, for instance the Internal Ligurian peridotites of the Tethys ocean, do not show the classical Penrose-type ophiolite sequence and are most similar to exhumed mantle in modern ultraslow spreading environments, e.g., like parts of the SW Indian ridge or the Gakkel ridge (Dick et al. 2000, 2006; Michael et al. 2003). The classical Penrose-type ophiolite stratigraphy, which is believed to be representative of moderate to fast spreading ocean ridges, is represented by the Samail ophiolite in Oman and the Troodos ophiolite (Cyprus). However, it should be noted that these ophiolite complexes were at least partly affected by convergent plate margin processes (Dilek et al. 2000; Dilek and Furnes 2014).
Indeed, many ophiolites probably formed close to subduction zones and were later incorporated into the crust by collision of terranes or continental fragments. Evidence for the proximity of subduction zones is mostly derived from the composition of associated igneous rocks such as calcalkaline basalts or boninites. To what degree subduction processes affected the mantle tectonites is not always clear. For instance, mantle rocks in the northwestern segments of the ophiolites in Oman may have been influenced by supra-subduction zone melting processes or by migration of magmas that formed in subarc mantle, as is indicated by the abundance of podiform chromitite deposits in these rocks and the calcalkaline and boninitic affinities of the crustal rocks (Boudier et al. 2000; Ishikawa et al. 2002). In contrast, the southern massifs of the Samail ophiolite in Oman show little evidence for such rocks and the crust is predominantly MORB-like in composition (Koga et al. 2001; Pallister and Knight 1981). Some peridotite massifs contain abundant pyroxenite layers which sometimes carry chemical and isotopic evidence for the significant presence of recycled crust components (e.g., Beni Bousera, Ronda, Bohemian massif; Pearson et al. 1991a,b, 1993; Becker 1996a,b). Such compositions only occur in mantle tectonites from areas that may have undergone lithospheric delamination and previous episodes of subduction. Some ‘ophiolites’, such as the Ligurian ophiolites (N. Italy) and similar complexes in the Alps, were not affected by convergent processes and are more properly assigned to purely extensional environments (e.g., Rampone et al. 1995, 1996; Piccardo and Guarnieri 2010).
In the following sections, we will describe the HSE and Os isotopic characteristics of different types of mantle tectonites in the context of their formation environments, as far as these have been constrained. These sections contain basic information on the formation environment and evolution of the ultramafic bodies together with the Re–Pt–Os isotopic and HSE composition of their various mantle lithologies. We will proceed from abyssal peridotites and other mantle tectonites exhumed in extensional geodynamic environments to peridotite massifs and ophiolites affected by magmatic processes at convergent plate margins. Interpretations of these compositions will then follow in the Discussion. The geological settings covered, locations, available HSE data and key references are summarized in Table 1.
HSE IN ABYSSAL PERIDOTITES FROM SPREADING OCEANIC LITHOSPHERE
Rocks from slow spreading ridges share many characteristics with mantle tectonites exhumed in passive continental margin or transitional oceanic environments (see later sections). That is, a spectrum of peridotite compositions is often present, including lherzolites, harzburgites and replacive dunites. However, in some cases (e.g., 15° 20′ N fracture zone, Atlantic Ocean; Harvey et al. 2006), a greater degree of serpentinization is present, sometimes with little primary mineralogy remaining, possibly due to the nature of emplacement and exposure of abyssal peridotites, either with little overlying crust (slow–ultra slow spreading) or bounded by transform faults. This can be important for the budgets of the HSE, as is discussed in the first section of the Discussion.
Abyssal peridotites contain major and trace element evidence for significant melt depletion, and isotopic evidence for that melt extraction being ancient, with long-term depletion of incompatible elements. Early studies found Os isotope evidence for this depletion, with 187Os/188Os ratios between 0.1208 and 0.1304 in abyssal peridotite whole rocks from several global localities (Martin 1991; Roy-Barman and Allègre 1994; Snow and Reisberg 1995). These ratios range from close to the estimate for the primitive upper mantle (0.1296; Meisel et al. 2001) to values which equate to Re depletion at ~ 1.2 Ga (TRD, see the following section on Continental tectonites), assuming all Re was stripped from the residue during melting (Shirey and Walker 1998). In reality, Re remains present in abyssal peridotites, although typically at much lower abundances than in the PM. This means that the actual age of depletion is older than calculated for a TRD age, because evolution of 187Os/188Os didn’t cease entirely after depletion. However, all abyssal peridotite sample suites display evidence for recent open system behavior, most probably in the form of Re addition (e.g., Harvey et al. 2006), but also sometimes Os loss and enrichment in 187Os (Snow and Reisberg 1995). This is apparent in the sub-horizontal trends within suites which show similar 187Os/188Os over a range of 187Os/188Os ratios (Fig. 1), and is consistent with the evidence for extensive serpentinization during fluid-rock interaction (e.g., Harvey et al. 2006). This seawater interaction can also be coupled with elevated and lowered 87Sr/86Sr and 143Nd/144Nd ratios, respectively (Snow and Reisberg 1995). However, despite the extremely radiogenic isotopic composition of seawater (187Os/188Os ~ 1.05; Levasseur et al. 1998), the modelled effects of seawater interaction on Os isotopes are small except at very high fluid/rock ratios (see Discussion section: Influence of low-temperature alteration processes on the HSE in bulk rocks and minerals), due to the very low Os concentration in seawater (~11 fg/g; Levasseur et al. 1998) compared to mantle samples (~ 1–5 ng/g). Moreover, a comparison of the rims and cores of abyssal peridotites from Gakkel Ridge in the Arctic Ocean found no systematic difference in Os contents and only a very small increase in 187Os/188Os from core to rim (Liu et al., 2008). A possible alternative source of radiogenic Os is by reaction with percolating melts from enriched lithologies. There is, however, a much larger effect of seawater interaction on Re/Os ratios (cf. ~ 7.3 pg/g Re in seawater; Anbar et al. 1992), with examples of sample rims reset while sample cores display a co-variation between Re/Os and Al2O3 contents, which must be a primary melt depletion feature (Liu et al. 2008).
In general, the processes of alteration mean that the measured Re-Os elemental and isotopic values may not accurately represent the long-term history of abyssal peridotites, casting doubt on the accuracy of TRD ages. Nonetheless, all abyssal peridotite suites consist primarily of samples with Os isotope ratios ranging from close to the PM estimate to sub-chondritic values (Fig. 1), reflecting long-term evolution in a low Re/Os environment following ancient melt depletion. Seawater interaction can only increase 187Os/188Os, so both alteration and minor ingrowth of 187Os since depletion would only serve to reduce the apparent age.
Snow and Reisberg (1995) proposed an ‘uncontaminated’ range for abyssal peridotites of 0.1221–0.1270, with a mean of 0.1246. Both Snow and Reisberg (1995) and Roy-Barman and Allègre (1994) identified that this range was less radiogenic than the range of early MORB analyses. Further analyses of samples from a forearc region and from slow or ultra-slow spreading ridges have significantly extended the known range of Os isotope compositions; whole-rock 187Os/188Os of 0.119, 0.117, and 0.114 were found, respectively, from the Izu-Bonin forearc (Parkinson et al. 1998), ODP Hole 1274a (15° 20′ N transform, mid-Atlantic; Harvey et al. 2006), and Gakkel Ridge (Arctic; Liu et al. 2008). The unradiogenic samples of the forearc setting were first thought to indicate that subduction zones might be ‘graveyards’ for ancient depleted mantle (Parkinson et al. 1998). While mantle in subduction zones may be extremely depleted, the findings from the 15° 20′ N transform and Gakkel Ridge indicate that such portions of ancient depleted mantle are likely present throughout the upper mantle.
Sulfide compositions display greater Os isotopic variation than whole-rocks (some plotting at 187Re/188Os ratios up to ~ 12) and can be divided into two broad groups: rounded intragranular grains and more skeletal interstitial sulfides (Alard et al. 2005; Harvey et al. 2006). The latter typically have higher Re/Os and more radiogenic Os isotope signatures (see Discussion), and the rounded, included sulfides possess the least radiogenic 187Os/188Os, lower than the host whole-rocks, reflecting depletion and isolation since an ancient melting episode.
Assuming that isochron information is typically compromised due to recent open-system behavior (see Fig. 1), then minimum Re depletion ages must be utilized; these are shown by the horizontal dashed lines in Figure 1. The least radiogenic whole-rocks from Gakkel and sulfides from the 15° 20′ N transform equate to TRD ages in excess of 2 Ga. The six rounded sulfides from one sample from Hole 1274a actually display a near-isochronous relationship. The age of this errorchron is ~ 2.05 Ga, consistent with the TRD ages for these sulfides. Sulfides from South-West Indian ridge peridotites (Warren and Shirey 2012) typically have more radiogenic compositions, closer to the PM value, and their sub-horizontal array suggests relatively recent resetting of their Re/Os ratios (Fig. 1). However, when combined with data from Alard et al. (2005) and with Pb isotope data (Warren and Shirey 2012), the broader array appears to give an age approaching 2 Ga.
As well as constraints on the 187Os evolution of the convecting mantle, the combined 186Os–187Os systematics of abyssal peridotites from the Kane transform area of the Atlantic Ocean have been studied (Brandon et al. 2000). The average 186Os/188Os of these samples is 0.1198353 ± 0.0000007, identical to the mean from alloys and podiform chromitites (Walker et al. 1997; Brandon et al. 2006), indicating the general absence of significant fractionation of Pt and Os in the abyssal and ophiolite environments. The Kane samples display a co-variation of 187Os/188Os with Pt/Os ratio which would likely not have been preserved if full Os-isotopic equilibration during melting had taken place (Brandon et al. 2000). No covariation of 187Os/188Os and Re/Os, due to seawater interaction, exists. The variability of 187Os/188Os could either be ascribed to differing ages of past depletion or to variable degrees of depletion, perhaps with garnet present in which Re is thought to be compatible (Righter and Hauri 1998). Brandon et al. (2000) proposed that Re is only depleted by about 40% in these rocks, therefore requiring very ancient melt depletion to produce the most unradiogenic samples. This ancient melting is not evident in 143Nd/144Nd, indicating decoupling of the two isotope systems, perhaps due to the Nd budget being predominantly hosted by clinopyroxene which is continually involved in partial melting, whereas the Os budget is likely dominated by included sulfides which are isolated from moderate degrees of partial melting and thus retain an ancient signature (Brandon et al. 2000). The later work of Harvey et al. (2006), outlined above, supports the influence of shielded sulfides, which control much of the whole-rock Os signature.
The apparent disconnect between abyssal peridotites and their overlying crust found in the early Os isotope abyssal studies (also see Discussion) is clearly seen in refractory Macquarie Island peridotites (Southern Ocean) and their surprisingly enriched overlying crust (Dijkstra et al. 2010). Here, a slow spreading and low productivity ridge would not be expected to account for the 20–25% near fractional melting suggested by the very high Cr numbers for spinel (0.40–0.49) in the peridotites. Although some authors have suggested a minor or absent role for abyssal peridotites in the generation of oceanic crust (e.g., Liu et al. 2008; Dijkstra et al. 2010), a compilation of abyssal peridotite data by Lassiter et al. (2014), including new analyses of Lena Trough peridotites (187Os/188Os: 0.118–0.130, average 0.1244), is remarkably similar to the distribution of 187Os/188Os in xenoliths entrained in ocean island basalts (Fig. 2 and references in caption). The authors argue that this range of 187Os/188Os for both suites represents the composition of the convecting mantle, and is inconsistent with a refractory ‘slag’ hypothesis for abyssal peridotites (cf. Rampone and Hofmann 2012). One issue with this interpretation, however, is that OIB xenoliths likely do not represent a deep source mantle for OIB melts, and instead might sample the lithospheric mantle which is plausibly genetically related to abyssal peridotites. Nonetheless, on an 187Os/188Os–Al2O3 diagram (sometimes called an ‘aluminachron’ (Fig. 2), where Al2O3 is used as a proxy for melt- and Re-depletion), abyssal peridotites and OIB xenoliths produce best-fit lines with similar ‘initial’ values, but differing slopes (the intersection of the correlation with the 187Os/188Os axis at Al2O3 = 0 yields the initial 187Os/188Osi at the time of the partial melting event). The similarity of the most depleted ‘initial’ values suggests that the age of Re depletion is similar for the two suites. So rather than the different slopes reflecting different depletion ages, the steeper trend of the abyssal suite instead suggests additional recent depletion of Al during partial melting to form new oceanic crust (Lassiter et al. 2014). This argues for a role for abyssal peridotites in the formation of mid-ocean ridge basalts. It remains possible, however, that the trends instead represent mixing between melts and residues and that the differing slopes reflect different conditions (e.g., depth, fS2 etc.) of such mixing.
Analyses of the range of HSE in abyssal peridotites showed that they are not present in strictly chondritic proportions (Snow and Schmidt 1998), and thus may not be consistent with the theory that HSE in the silicate Earth were derived from a late veneer of primitive chondritic material, after core formation had ceased (Chou 1978). Snow and Schmidt (1998) proposed that mantle HSE patterns instead reflected remixing of the outer core into the mantle. However, subsequent analyses using improved digestion techniques (Becker et al. 2006) cast doubt on the robustness of the Os, Ir, and Ru data in that study (obtained by NiS fire assay), reducing the magnitude of the observed non-chondritic signature. Moreover, later work highlighted the importance of metasomatism and melt–rock reaction processes in producing non-chondritic HSE patterns in mantle rocks. Rehkämper et al. (1999) found that abyssal peridotites broadly contained HSE in chondritic proportions and that HSE ratios were inconsistent with an outer core input. Where non-chondritic ratios were identified in the Horoman peridotite, a petrogenetic model showed that these ratios were consistent with sulfide addition associated with melt percolation. Alard et al. (2000) then identified PPGE-rich (Pt-group) and IPGE-poorer (Ir-group; Barnes et al. 1985) interstitial sulfides that were introduced during melt infiltration. These sulfides have the potential to produce non-chondritic HSE patterns in whole-rocks and also have suprachondritic Re/Os and 187Os/188Os.
A study of Kane fracture zone peridotites (Atlantic Ocean) identified a range of HSE systematics in different lithologies (Luguet et al. 2003). Harzburgites have low Pd/Ir ratios and are sulfide-poor. Refertilized harzburgites often have higher concentrations of Pd, while lherzolites have approximately chondritic proportions of HSE and between 100 and 300 μg/g S, which encompasses the estimate for the PM (250 ± 50 μg/g; Lorand 1990; McDonough and Sun 1995; Palme and O’Neill 2014). Peridotites from the 15°20′ N fracture zone are typically more depleted (Marchesi et al. 2013) than those from Kane (Fig. 3), and show complete consumption of sulfide in some cases, presumably with HSE (particularly the IPGE) then hosted by PGM. Both this study, and an earlier one looking at two sites with differing alteration from Gakkel Ridge (Liu et al. 2009), found there to be no significant mobilization of the HSE during serpentinization, but S contents were reduced. The same is also true of weathering, except for Re and Pd in some cases. For the 15° 20′ N fracture zone, there was also no observed mobilization of HSE by S-undersaturated melt, which is somewhat surprising given that S-undersaturated melt would be expected to dissolve sulfide. Presuming that sulfide was dissolved into the melt, the implication from the 15° 20′ N samples is that all HSE, with the exception of Re, are retained until sulfide is almost exhausted. However, this finding may be dependent on the phase relations in any given sulfide system, as fractionation of sulfide melt from solid sulfide would be expected to fractionate PPGE from IPGE (e.g., Mungall et al. 2005). In contrast to studies advocating melt percolation as a means to fractionate HSE (e.g., Alard et al. 2000), Liu et al. (2009) contend that supra-chondritic Ru/Ir and Pd/Ir in Gakkel peridotites cannot be reconciled with melt enrichment and therefore instead support an inherent primitive origin for such ratios.
HSE in mantle tectonites from continental extensional domains and continent–ocean transitions
Mantle tectonites exhumed in passive continental margin or transitional oceanic environments share many characteristics with similar rocks from ultraslow spreading ridges (see the sections on abyssal peridotites). These environments often display the complete spectrum of peridotite compositions, including lherzolites, harzburgites, and replacive dunites and because of their compositional variety, mantle rocks from these environments have been the focus of detailed petrological and geochemical studies of lithophile element behavior, HSE abundances and Re–Os isotopic studies. Many of these tectonites have been exhumed in the course of the development of small oceanic and sedimentary basins in the Alpine–Mediterranean realm (Piccardo and Guarnieri 2010).
In the Pyrenees, numerous small, serpentinized peridotite bodies (typically km2 size or less) occur as lenses in high-grade gneiss–granulite–sediment rock associations (e.g., at Lherz, Bestiac, Turon de Tecouere). The mantle and lower crustal rocks were presumably exhumed during extension and subsequent compressional movements between Iberia and the European plate in the Mesozoic to Cenozoic (Vielzeuf and Kornprobst 1984). The mantle rocks are predominantly variably serpentinized spinel lherzolites and harzburgites, with occasional spinel and garnet facies pyroxenitic banding (Bodinier et al. 1987, 1988). Melt infiltration affected incompatible trace elements, such as the light REE, in the mantle rocks to a variable extent (Vasseur et al. 1991). The small ultramafic body near the village of Lherz (Lers), the type locality of lherzolite, has been studied in detail and has yielded textural and geochemical evidence that the lherzolites in that body formed by reactive infiltration of incompatible element-depleted melt into older harzburgites (Le Roux et al. 2007). The peridotites at Lherz are a key example that shows how reactive transport of basic silicate melt may re-enrich depleted mantle rocks in incompatible major elements via precipitation of pyroxenes, a process called refertilization. The pyroxenites may represent leftover cumulates and reaction products from these processes. However, mechanical mixing of pyroxenite and harzburgite has also been proposed as a mechanism capable of producing the refertilization at Lherz which is commonly attributed to melt reaction (Riches and Rogers 2011).
Early Re–Os work on peridotites from Pyrenean ultramafic bodies by Reisberg and Lorand (1995) yielded positive correlations between measured 187Os/188Os and 187Re/188Os (the Re–Os isochron diagram, Fig. 4a), as well as Al2O3 contents (Fig. 4b), respectively. Al2O3 contents have been used as a preferred melt extraction index (see also Fig. 2) and proxy for the Re/Os ratio, because Re abundances are typically believed to have been partially affected by serpentinization, whereas Al is largely considered immobile during alteration processes (Reisberg and Lorand 1995; Shirey and Walker 1998). The positive correlation of 187Os/188Os with Al2O3 was interpreted to reflect past melt extraction, assuming the mantle rocks were cogenetic and their different Al2O3, Re contents, and Re/Os ratios reflected different degrees of partial melting. The intersection of the ‘initial’ 187Os/188Os and a chondritic evolution curve then gives a model age of 2.3 Ga (Reisberg and Lorand 1995). Figures 4b and 4c show this model age concept using the range of measured 187Os/188Os in bulk rocks of chondrites (Walker et al. 2002a; Fischer-Gödde et al. 2010). The same approach was applied by Reisberg and Lorand (1995) to peridotites from the Ronda peridotite massif (see below). The ancient Re–Os model ages of these peridotite massifs, their coincidence with Sm–Nd model ages of overlying crustal rocks and their geodynamic position have been used to argue that these bodies represent fragments of exhumed subcontinental lithospheric mantle that have undergone Proterozoic melt extraction (Reisberg and Lorand 1995; Burnham et al. 1998). It is plausible to infer that the melt extraction processes may have occurred in an ocean ridge environment and consequently, the model ages would record the ancient formation of lithospheric mantle from asthenosphere.
Subsequently published HSE concentration data for the same and additional samples show some features that are not only characteristic of peridotites from the Pyrenees, but also of mantle tectonites from many other locales. Here we will outline the differences between lherzolites and harzburgites, because these different lithologies have been well studied for their bulk rock compositions, as well as their sulfide and other accessory phase mineralogy and mineral compositions. The lherzolites (Fig. 5a) display limited abundance variations for Os, Ir, Ru, and Rh, and variable abundances of Pt, Pd, Au, Re, and the chalcogen elements S, Se, and Te (Pattou et al. 1996; Lorand et al. 1999, 2008, 2010, 2013; Becker et al. 2006; Luguet et al. 2007; Fischer-Gödde et al. 2011; König et al. 2012, 2014; Wang and Becker 2013).
The highest abundances of Pt, Pd, and Re in the Pyrenean lherzolites occur in samples that yield 187Os/188Os and major element compositions similar to estimates of the composition of the primitive mantle (Meisel et al. 2001; Becker et al. 2006). Ratios of the HSE in these samples suggest broadly chondritic proportions of the HSE, with the exception of Ru and Pd, which are suprachondritic compared to other HSE. In contrast, harzburgites (Fig. 5a) from the Lherz body are commonly strongly depleted in Rh, Pt, Pd, Re, and chalcogens, whereas abundances of the Ir group PGE (IPGE; Os, Ir, and Ru) were retained at similar to slightly higher levels than in lherzolites.
Study of the accessory phase mineralogy of peridotites from the Pyrenees has indicated the presence of variable proportions of different sulfide types (pentlandite, pyrrhotite, chalcopyrite, pyrite), alloy phases (Os-rich, Pt-rich, Au-rich) and other types of platinum-group metal phases such as Pt-bearing tellurides (Fig. 6) (Luguet et al. 2007; Lorand et al. 2008, 2010; Lorand and Luguet 2016). The majority of these phases are likely low-temperature exsolution products that formed during cooling of once homogeneous high-temperature phases such as sulfide liquids and monosulfide solid solution. The exsolution origin of such phases is reflected in strong chemical fractionations of some HSE (notably Pt, but sometimes also Pd and Au) and related elements (e.g., Bi, Te, Se, and S) that are only observed on the grain scale, but not in corresponding bulk rocks. However, some alloy phases, for instance Pt–Ir- and IPGE-rich alloys may have been inherited from previous episodes of high degrees of melting (Lorand et al. 2010). The significance of these observations are discussed further below and by Lorand and Luguet (2016, this volume), and Reisberg and Luguet (2016, this volume).
Balmuccia, Baldissero and Lanzo peridotite bodies
In northern Italy, several peridotite bodies occur that also represent fragments of continental lithospheric mantle in an extensional continental margin setting. The peridotite bodies at Balmuccia, Baldissero, and Lanzo were all derived from the southern European passive continental margin that had developed following the Variscan orogeny. Towards the end of the Variscan orogeny during the lower Carboniferous and upper Permian, the lower crust and presumably also existing continental lithospheric mantle were flooded with MORB like magma from the asthenosphere (Quick et al. 2009; Snoke et al. 1999; Voshage et al. 1990). The peridotite bodies of Balmuccia and Baldissero, mostly spinel lherzolites with subordinate harzburgites, discordant dunites and pyroxenites, show different distributions of their TRD (Fig. 10 in Wang et al. 2013). In Balmuccia the model ages of the lherzolites show a single distribution peak of Paleozoic model ages, with a harzburgite yielding the only Proterozoic model age (Note: samples with 187Os/188Os < 0.1254 yield Precambrian TRD[PM] model ages, see Fig. 7). At Baldissero, a bimodal distribution of TRD occurs with a Paleozoic and a Proterozoic peak (Fig. 10 in Wang et al. 2013). Lithophile element, Re–Os, Sm–Nd isotopic, and HSE abundance data and textural relations can be interpreted such that depleted Proterozoic mantle (the harzburgites) were variably refertilized by MORB-like magma during the Paleozoic (Mazzucchelli et al. 2009, Mukasa and Shervais 1999, Obermiller 1994, Rivalenti et al. 1995, Wang et al. 2013). The greater compositional homogeneity of peridotites from Balmuccia compared to those from Baldissero (Fig. 5b, c) suggests that the former body was fluxed and re-equilibrated with melt more efficiently than the latter. IPGE concentrations in harzburgites in both bodies are lower than in lherzolites, which is opposite to observations from some other suites of peridotites (Pearson et al. 2004; Becker et al. 2006). Re–Os data suggests that some of the Cr-diopside-rich websterites at Balmuccia may have formed during these or earlier events of reactive melt infiltration. However, most Al-augite-rich clinopyroxenites yielded Jurassic model ages (Wang and Becker 2015c). Spinel and plagioclase bearing lherzolites from the Lanzo peridotite massif are similar to lherzolites from Baldissero in their HSE patterns (not shown in Fig. 5) and in their distribution of 187Os/188Os data (Fig. 7a, Becker et al. 2006; Fischer-Gödde et al. 2011).
External and Internal Ligurian peridotites
The External Ligurian peridotites are now recognized to represent mantle rocks of the subcontinental lithospheric mantle of the south European realm (but more distal than Lanzo, Balmuccia and Baldissero), presumably exhumed during the early- to mid-Mesozoic (Rampone et al. 1995; Piccardo and Guarnieri 2010). In contrast, the Internal Ligurian peridotites have been interpreted to derive from depleted mantle of ultraslow spreading ocean floor of the Jurassic Tethys Ocean (Rampone et al. 1996, 1998; Piccardo and Guarnieri 2010). In both cases, plagioclase-spinel lherzolites are the predominant rock type (with subordinate pyroxenites).
Detailed petrological and geochemical work in these and other studies has shown that the Ligurian peridotites have been variably affected by melt infiltration and refertilization (Rampone et al. 2004). In spite of the somewhat different tectonic setting, the Re–Os and HSE composition of External and Internal Ligurian peridotites is similar to other lherzolites (Figs. 5d and 7a; Snow et al. 2000; Luguet et al. 2004; Fischer-Gödde et al. 2011). Mantle lherzolites and pyroxenites with evidence for melt infiltration and chemical characteristics similar to lherzolite massifs from N Italy have been described from the suture zone in the Alps (e.g., Totalp, Swiss Alps; van Acken et al. 2008; 2010a,b). The Totalp lherzolite body is notable for its Re-rich composition and slightly suprachondritic Re/Os of the lherzolites (Figs. 5e, 7c), which is different from most other peridotite tectonites. The Re-rich composition of the lherzolites and associated pyroxenites can be related to infiltration of melt with MORB-like isotopic compositions, presumably during the Mesozoic or late Paleozoic.
The peridotite body of Zabargad Island in the Red Sea represents a young example of subcontinental lithospheric mantle, exhumed during post-Miocene extension of the Red Sea (Bonatti et al. 1986; Piccardo et al. 1993).
Spinel-bearing lherzolites, amphibole harzburgites and dunites display evidence for metasomatism by fluids or hydrous melts which led to the formation of amphibole harzburgites (Piccardo et al. 1993). The HSE patterns and S abundances of the lherzolites are similar to comparable rocks elsewhere. However, Cu is notably depleted in these lherzolites (around 10 μg/g). Amphibole-bearing dunite and harzburgite have higher than expected abundances of Pd, Au, Re, and S (Fig. 5f; Schmidt et al. 2000). An orthopyroxenite and a plagioclase wehrlite display high PGE and Au abundances, but low Re, S, and Cu abundances (Schmidt et al. 2000).
The Horoman peridotite body in Japan comprises outcrops of layered dunite, harzburgite and lherzolite that have been interpreted to be the result of variable degrees of melt-peridotite reaction that occurred during percolative melt transport in the mantle. Dunites, harzburgites and spinel- and plagioclase-bearing lherzolites at Horoman are believed to have undergone variable degrees of pyroxene dissolution into percolating olivine-saturated magma (Takahashi 1992; Takazawa et al. 1992, 1996, 1999). Despite the occurrence of highly unradiogenic Pb in the Horoman peridotite (Malaviarachchi et al. 2008), abundances of the HSE and 187Os/188Os data in lherzolites and harzburgites (Rehkämper et al. 1997; Saal et al. 2001) are similar to data from peridotites elsewhere. The correlation of Re abundances with MgO in the peridotites may be the result of refertilization processes (Saal et al. 2001).
HSE in ophiolites that formed at fast spreading ridges with little or no influence from subduction processes
Oman ophiolite, Wadi Tayin Section
The crustal and mantle section of Wadi Tayin in the SE part of the Samail ophiolite (Oman) represents one of the best exposed examples of fast-spreading oceanic crust and upper mantle on Earth (Pallister and Hopson 1981; Hanghøj et al. 2010). Geochemical studies of the crustal rocks in the section indicate that the crust mostly comprises normal mid-ocean ridge-type basalts and gabbros (Pallister and Knight 1981; Koga et al. 2001). Part of the ophiolite likely formed at an ocean spreading center about 90–95 Ma ago, but must have been incorporated into an active subduction–collision zone that led to changes in magma compositions in the NW part of the ophiolite (Tilton et al. 1981; Searle and Cox 1999).
A study of PGE and Re abundances and 187Os/188Os in the lower crustal gabbros indicated low Re concentrations and systematically higher PGE concentrations compared to MORB (Peucker-Ehrenbrink et al. 2012). The Os isotopic compositions of some gabbros may have been affected by circulation of seawater. The HSE abundances and 187Os/188Os of parts of the exposed mantle section were studied across an 11 km transect from the exposed Moho into high- and then low-temperature peridotites underneath (Hanghøj et al. 2010). Platinum-group element concentration data on harzburgites of similar composition have also been published by Lorand et al. (2009). The high-temperature peridotites likely represent textures and compositions of the mantle inherited from the ocean ridge stage, whereas the low temperature peridotites underneath may represent mantle modified by deformation, re-equilibration and fluid transport during obduction of the ophiolite. The mantle rocks at Wadi Tayin comprise serpentinized harzburgites and replacive dunites that are strongly enriched in fluid-mobile incompatible lithophile elements (e.g., Rb, Pb), which may reflect late alteration or, alternatively, retention of small quantities of melt during peridotite-melt interaction (Hanghøj et al. 2010). The strong fractionation of the REE in most of these samples is significantly greater than in abyssal peridotites and suggests that these rocks can be regarded as highly depleted melting residues in which the LREE were strongly depleted by fractional melting (Hanghøj et al. 2010). The dunites are usually interpreted as forming by magmatic dissolution–precipitation processes that dissolve pyroxenes and increase the modal amount of olivine (Kelemen et al. 1995). Harzburgites typically show consistent HSE abundances with IPGE greater than most abyssal peridotites, slight depletion in Pt, and enrichment in Pd. Dunites, however, show far greater variability, including their Os/Ir ratio, and range from moderately depleted abundances of Re, Pd, and Pt to variable enrichments of Re, Pd, and Pt, sometimes a factor of 2–3 times higher than values commonly observed in lherzolites (Fig. 8). The enrichments of Re, Pd, and Pt in the harzburgites and dunites may have resulted from shallow precipitation of magmatic sulfide from S-saturated magmas, although S concentrations in the mantle rocks are low (typically a few tens of μg/g, Hanghøj et al. 2010) compared to Pd, Re, and Cu abundances (Fig. 8). The initial 187Os/188Osi (at 90 Ma) of the harzburgites and dunites are remarkable in that they display a large range from as low as 0.113 to suprachondritic values of 0.15 in dunites (Fig. 9). As in other mantle tectonites, most samples are in the chondritic to subchondritic range, however, some samples with suprachondritic 187Os/188Os either require interaction with magma with radiogenic 187Os/188Os, or have lost a substantial amount of their original inventory of Re.
Taitao ophiolite (Chile)
The Taitao ophiolite on the Taitao Peninsula in S. Chile is believed to represent part of the oceanic lithosphere formed about 6 Ma ago on the Chile Ridge, which is presently subducting under South America (Guivel et al. 1999). The ophiolite was obducted during or soon after its magmatic formation and was affected by hydrothermal alteration and a metamorphic overprint related to subduction, obduction and contact metamorphism imposed by young granitoid intrusions. The Taitao ophiolite displays a somewhat dismembered Penrose style sequence of serpentinized harzburgites, gabbros, basic dikes, pillow basalts and sediments (Schulte et al. 2009 and references therein). The chemistry of the basic rocks hints that at least some of these magmas may have been affected by subduction zone processes, similar to basalts from the active Chile Ridge (Klein and Karsten 1995). The serpentinized harzburgites display some variability in their HSE patterns ranging from samples that display variable depletions of Re and Pd, depletion of Pd but not Re, and samples showing positive or negative anomalies of Pt relative to Ru and Pd (Fig. 10; Schulte et al. 2009). Basic rocks tend to have very low abundances of IPGE, with variable positive Pt anomalies and strong enrichment of Re (Schulte et al. 2009). Measured 187Os/188Os range from 0.117 to 0.128, with many samples scattering around a 1.6 Ga reference line in an isochron diagram (Fig. 11). Because of the relatively large range in 187Os/188Os and the strongly depleted major element composition of the harzburgites, the slope in the 187Os/188Os–Al2O3 diagram (Fig. 12) is different from other suites of peridotites (Figs. 2 and 7). Schulte et al. (2009), however, interpreted the HSE data of the mantle rocks to reflect a two-stage partial melting history at 1.6 Ga and 6 Ma ago. Textural evidence indicates that some harzburgites may have been affected by melt impregnation processes, which may have led to some of their chemical and isotopic variability. The initial 187Os/188Osi (6 Ma) of the basic rocks ranges from chondritic to suprachondritic (γOsi = −1 to +342). The suprachondritic composition may either reflect the presence of a rhenium-enriched component in the mantle source or the influence of seawater/altered crust during the emplacement of the magmas.
High-temperature orogenic peridotites from convergent plate margin settings
Many ophiolites have originally been emplaced near subduction zones and commonly even their mantle sections were affected by magmas that formed in supra-subduction zone environments (see below). Among high-temperature orogenic peridotites, evidence for the influence on mantle rocks by magmas that may have formed in convergent plate margin settings is not very common and, in fact, is somewhat ambiguous. Here, we discuss examples of mantle tectonites that were emplaced during or in the aftermath of subduction and collision processes. In the case of the Ronda and Beni Bousera ultramafic massifs these bodies represent mantle exhumed during the collapse of the Betic orogenic belt in the western Mediterranean during the Cenozoic (van der Wal and Vissers 1993; Blichert-Toft et al. 1999). In the southern Bohemian massif, similar processes occurred during collapse of the core zone of the Variscan belt during the Carboniferous (Medaris et al. 2005). The principal evidence is mostly derived from geodynamic reconstructions in combination with the lithophile element and isotope geochemistry of peridotites and pyroxenites. Notably, garnet-bearing pyroxenite layers in these peridotite massifs show strong evidence that they formed from magmas with crustal geochemical and isotopic signatures (e.g., Eu anomalies, enrichments of LREE, graphite with δ13C suggestive of organic protoliths, sediment-like Sr–Nd–Pb isotopic compositions; (Pearson et al. 1991a,b, 1993; Becker 1996a).
Ronda (Southern Spain)
The Ronda peridotite has been a classic study area of mantle processes (Frey et al. 1985; Reisberg and Zindler 1986; Reisberg et al. 1989). It shows a transition from garnet lherzolite to spinel lherzolite and plagioclase-bearing peridotites (Obata 1980). Initially the peridotites were regarded as residues of partial melting (Frey et al. 1985); however, later the significance of melt infiltration into continental lithospheric mantle was recognized and the latter process also may have caused partial re-equilibration of the peridotites at shallow pressure-temperature conditions (Bodinier et al. 2008). Re–Os model ages of depleted peridotites yield an average age of melt extraction in these rocks of 1.3 ± 0.4 Ga (Reisberg et al. 1991, Reisberg and Lorand 1995; Becker et al. 2006). The HSE patterns (not shown) and 187Os/188Os (Fig. 7d, h) of the peridotites are similar to data on fertile to depleted peridotite tectonites exhumed in extensional tectonic settings (Lorand et al. 2000; Becker et al. 2006; Fischer-Gödde et al. 2011). Pyroxenite layers from Ronda have suprachondritic Re/Os and 187Os/188Os, and Pd and Pt are enriched relative to IPGE (Reisberg et al. 1991; Marchesi et al. 2014). The depletion of Re in some pyroxenites relative to Pd (see Discussion section on HSE fractionation during the formation of mantle pyroxenites) may reflect multi-stage melting processes (Marchesi et al. 2014).
Beni Bousera (Morocco)
The Beni Bousera peridotite massif crops out on the Moroccan side of the Alboran Sea and shares a similar history with the Ronda body. Re–Os and HSE concentration data on peridotites are comparable with data from Ronda (Fig. 7d, h, Kumar et al. 1996; Pearson and Nowell 2004; Pearson et al. 2004; Luguet et al. 2008b; Fischer-Gödde et al. 2011). Studies of the Re–Os systematics in pyroxenite layers from Beni Bousera yielded highly variable Re/Os and 187Os/188Os, the latter reflecting radiogenic ingrowth, but also partly incorporation of unradiogenic Os from reaction with the host peridotites (Kumar et al. 1996, Pearson and Nowell 2004). The Re–Os model ages cluster near 1.3 Ga, similar to results from some peridotites, and similar to Lu–Hf ages of some, but not all pyroxenites. The spectrum of Re–Os model ages and Lu–Hf isochron ages is consistent with other evidence that suggests a complex multi-phase history of both the Ronda and the Beni Bousera bodies (Loubet and Allègre 1982; Marchesi et al. 2014). Luguet et al. (2008b) and Marchesi et al. (2014) found variations of Pt/Os and Re/Os in some bulk rocks and sulfides from pyroxenites at Beni Bousera and Ronda, respectively. These rocks were interpreted to represent likely equivalents of the sources of mantle plume-derived picrite and komatiite lavas with elevated 186Os signatures (Brandon and Walker 2005 and Discussion section).
Southern Bohemian Massif (Lower Austria, Czech Republic)
In the Bohemian Massif, kilometer-sized bodies comprised of serpentinized high-temperature garnet lherzolites, spinel harzburgites, and dunites occur enclosed in high-pressure granulites and amphibolite facies gneisses (e.g., Carswell and Jamtveit 1990; Becker 1996b, 1997; Medaris Jr et al. 2005). As in the peridotite massifs of the Betic cordillera, the garnet pyroxenite layers in the peridotites show chemical and isotopic compositions that suggest that they precipitated from basic magmas that formed in mantle contaminated by recycled sedimentary material (Becker 1996a). Detailed Re–Os work on layered peridotite–pyroxenite rocks indicates that peridotite-derived Os and Cr are mobilized during melt–rock reaction that led to the formation of layered pyroxenite-dunite rocks (Becker et al. 2001, 2004). The pyroxenites in these rocks show suprachondritic initial 187Os/188Os which may be inherited from subducted materials as indicated by initial Sr–Nd isotopic compositions. The variation of 187Os/188Osi in modally layered lithologies indicates Os isotopic disequilibrium on the cm-scale resulting from magmatic infiltration processes. Another, yet different, example of metasomatic overprint that affected HSE abundances in peridotites in the Bohemian Massif are Mg-rich peridotites with relatively high IPGE contents (e.g., up to 10 ng/g Os), but not quite as high Pt, Pd, and Re abundances (Ackerman et al. 2013). These rocks occur with pyroxenites and Fe-rich cumulate rocks with high Pt, Pd, and Re abundances (Fig. 13).
In peridotites from Ronda, Beni Bousera, and Lower Austria, measured 187Os/188Os are subchondritic or chondritic, similar to peridotites from extensional tectonic settings. Pyroxenites show high, but variable Re/Os and suprachondritic γOsi. However, unlike some data on lithophile elements, these features are not necessarily indicative of the influence of subducted crust or subduction zone fluids. High Re/Os (and γOsi) seem to be a hallmark of mantle pyroxenites and may be acquired by magmatic fractionation in the crust or during melting and transport of magmas in the mantle (e.g., Pearson and Nowell 2004; van Acken et al. 2010b; Marchesi et al. 2014; Wang and Becker 2015c). This topic will be discussed in later sections.
Highly siderophile elements in peridotites and melt-reacted lithologies of ophiolites influenced by convergent plate margin magmatism
In comparison to ophiolites with little subduction influence, convergent plate margin ophiolites typically comprise more depleted harzburgitic mantle sections and thicker ultramafic sequences in the lower crust. This is due to the greater degree of partial melting that usually occurs in the fluid-fluxed supra-subduction zone setting. However, the presence of hydrous melts and fluids also promotes the formation of melt-reacted lithologies such as dunites, pyroxenites and, in particular, chromitites, in the mantle sections of ophiolites from convergent plate margins. Such melt–rock reaction, and the lithologies it produces, is diverse and depends principally on the melt/rock ratio and the degree of saturation of silica and sulfide in the melt. The variable impact of melt–rock reaction on HSE-host phases is of course, critical to the behavior of the HSE, and melt–rock reaction is thus a major process by which HSE are fractionated and heterogeneity is generated. This fractionation of the HSE, including that which occurs during chromitite formation, likely plays an important role in defining the HSE characteristics of magmas at Earth’s surface, particularly those of convergent margin ophiolites and in volcanic arc systems (e.g., Dale et al. 2012b).
At the same time, there is the potential for sulfide to be exhausted during moderate to high degrees of mantle melting, particularly if sulfur solubility is increased (Jugo 2009) due to an elevated oxygen fugacity of the sub- or back-arc mantle, relative to typical depleted MORB mantle (e.g., Carmichael 1991; Kelley and Cottrell 2009). Given the extremely chalcophile nature of the HSE (e.g., Mungall and Brenan 2014; with the exception of Re; Brenan 2008), sulfide exhaustion would cause HSE behavior to depart significantly from the typical mid-ocean ridge setting where sulfide is thought to remain in the residue.
Commonly, convergent margin ophiolites contain substantial units of podiform chromitite, enveloped in dunite, which require high degrees of melt depletion and are probably formed through a process of melt–rock reaction, particularly when a hydrous melt is present and the melt/rock ratio is high, or as cumulates from melts formed through high degrees of melting (Ballhaus 1998; Zhou et al. 1998). Chromitites are known to contain variable but high concentrations of HSE (Prichard and Lord 1996), particularly the IPGE, indicating their presence in high concentrations in the chromitite-forming melts. Further concentration of HSE occurs primarily because chromitites contain associated platinum-group mineral grains (PGM) which form due to a local oxygen fugacity-induced reduction in solubility of the HSE (Finnigan et al. 2008). This reduction in oxygen fugacity occurs locally around chromite crystals because of their preference for trivalent transition metal cations, particularly Cr3+ and Fe3+ ions. The IPGE have lower solubilities in silicate melts than PPGE, on the order of tens versus hundreds of ng/g (e.g., O’Neill et al. 1995; Borisov and Walker 2000; Brenan et al. 2005; Ertel et al. 2006), and hence Os, Ir, and Ru are particularly enriched in PGM from chromitites. Although chromitites and platinum-group minerals (PGM) are covered more comprehensively in O’Driscoll and González-Jiménez (2016, this volume), we include a brief Os isotope summary in the Discussion because ophiolitic chromitites are a major source of PGM, and they have a direct bearing on determining both the Os isotopic composition of the convecting mantle and the degree of mantle heterogeneity. Here we focus mainly on HSE behavior in the range of mantle lithologies present in ophiolites, rather than the specifics of PGM mineralogy and its role in HSE behavior (cf. O’Driscoll and González-Jiménez 2016, this volume).
Troodos Ophiolite (Cyprus)
Two complementary studies of melt percolation in the Troodos ophiolite found fractionated HSE abundances and variable 187Os/188Os in a range of residual and melt reaction products (Büchl et al. 2002, 2004). A sequence of spinel lherzolites, minor dunites and clinopyroxene-bearing harzburgites was found to have a large range of initial 187Os/188Os (at 90 Ma) from subchondritic (0.1168) to mildly suprachondritic (0.1361); a second unit, consisting of harzburgites, dunites and chromitites, has an even larger and more radiogenic range 0.1234–0.1546. The subchondritic values can readily be explained by ancient melt depletion of Re (> 800 Ma), as for abyssal peridotites and most other mantle rocks. The suprachondritic Os compositions, as with those from the Oman ophiolite described earlier (Hanghøj et al. 2010) and many other ophiolites (see Figs. 14 and 15), require the addition of a radiogenic melt component (unless samples have experienced significant recent Re loss), likely during the formation of the Troodos around 90 Ma ago. The ultimate source of this radiogenic Os is not known, and could relate to seawater contamination prior to concentration in chromitites (because a radiogenic signature is also evident in the most Os-rich chromitite samples) or to crustal contamination during emplacement, but the former at least is difficult to reconcile with the very low Os concentrations in seawater (Levasseur et al. 1998). Another possible mechanism, that would be applicable to both mid-ocean ridge and supra-subduction ophiolites, is the production of radiogenic melts due to preferential sampling of radiogenic interstitial sulfides (Alard et al. 2005; Harvey et al. 2011) or due to the presence of enriched domains in the mantle (cf. pyroxenites; Reisberg et al. 1991; Pearson and Nowell 2003). However, melting of enriched domains is not consistent with the refractory boninitic melt that typically produces HSE- and Cr-rich chromitites. Given the apparent global distinction in Os isotopes between ophiolites of convergent and mid-ocean ridge origin (Fig. 15), the most plausible explanation for a significant part of the radiogenic signature is a flux from the subducting slab, with Os mobilized in oxidized chlorine-rich fluids (Brandon et al. 1996; Becker et al. 2004). In this scenario, despite the extreme fractionation of Re from Os in oceanic crust, the low Os contents and relatively young age of subducted mafic crust would suggest that a sedimentary input may be required to provide sufficient radiogenic Os to impart that signature on the Os-rich mantle.
The process(es) of dunite formation also induces significant HSE fractionation. Harzburgites, which could be simple residues of melting or, as Büchl et al. (2002) conclude, the product of melting during melt-percolation at low melt/rock ratios, have largely uniform IPGE patterns and concentrations that only range by roughly a factor of two (Fig. 16). Palladium and Re abundances do, however, vary over approximately an order of magnitude in harzburgites (Büchl et al. 2002). In contrast, a dunite rim and core, the product of high melt/rock ratios, together with a websterite and a boninite all display high and remarkably uniform concentrations of Pt (6.5–12.2 ng/g), moderately variable Pd and Re, and two or more orders of magnitude variation in Os content. Qualitatively, it seems that dunites and websterite could be produced by some sort of reaction and mixing process between harzburgite and boninitic melt, retaining high Pt but removing/diluting Os; requiring Os to be mobilized. This is supported by modelling of HSE ratios (dominated by mixing of harzburgitic and magmatic sulfides) and REE in clinopyroxene during open system melting (Büchl et al. 2002).
Shetland Ophiolite Complex (UK)
Harzburgites from Unst, Shetland, have Os isotope compositions ranging from γOs of 2 to −6 (using an O-chondrite reference frame; 187Re/188Os = 0.422, 187Os/188Os = 0.1283). Most Os isotope ratios are consistent with an ambient convecting mantle signature (see Os isotopic heterogeneity in the mantle in Discussion) but there is evidence of both melt depletion at ~ 1.2 Ga and also radiogenic Os addition for some samples (O’Driscoll et al. 2012).
Dunites have a wider range of 187Os/188Os than harzburgites (γOs492Ma of −22 to 12), reflecting the effects of melt–rock reaction involved in their formation (O’Driscoll et al. 2012). Chromitites have the narrowest range of 187Os/188Os, from γOs +0 to +3.5. This relative homogeneity is perhaps surprising given the higher melt/rock ratios involved in producing chromitite, but this is set against the extremely high Os concentrations, and low Re abundances, that allow for accurate estimation of the initial Os isotope composition. In part, the range for dunites (and harzburgites) may reflect difficulties in age correcting over 492 Ma (as this is dependent on measured Re and Os concentrations—with the potential for recent disturbance). Overall, however, a radiogenic Os flux is required to explain the supra-chondritic γOs values. As discussed for the Troodos Ophiolite, there are various possible sources of the radiogenic Os, but a flux from the downgoing slab may be the most plausible mechanism.
Shetland Ophiolite samples display huge variations in HSE concentrations, with some chromitites containing up to ~ 100 μg/g Pt (Prichard and Lord 1996; O’Driscoll et al. 2012) while some dunites contain less than 100 pg/g Pt. The most HSE-rich chromitites (from Cliff) have Ir and Ru contents that are roughly two orders of magnitude higher than the range of chromitites analyzed from the Qalander, Luobusa, and Zambales ophiolites (Fig. 16). Moreover, these chromitites have unusual HSE patterns with PPGE/IPGE ratios greater than unity and Pd concentrations up to 156 μg/g (O’Driscoll et al. 2012), compared with typical IPGE-rich chromitites which have Pd and Pt contents approximately four orders of magnitude lower (Zhou et al. 1996, 2000, 2014; Ismail et al. 2014). The range of HSE abundances between chromitites from different localities is, in itself, huge. The two other localities analyzed have more typical HSE patterns, albeit in one case also enriched by one to two orders of magnitude. The degree of P-PGE enrichment has been linked to the thickness and sulfide content of the ultramafic dunite sequence and ultimately to the degree of melting, and, in the case of the extremely PPGE-enriched Cliff chromitites, also linked to hydrothermal redistribution from surrounding ultramafics (Prichard and Lord 1996).
There are also large variations in the HSE concentrations and patterns of dunites, which show an overall depletion in Pt, relative to IPGE, and are enriched in Pd in many cases. Rhenium concentrations are low in almost all harzburgite, dunites, and chromitites, although enrichment in Re does also occur in some dunites.
Zambales Ophiolite (Philippines)
The Zambales Ophiolite contains two distinct blocks, which differ in the composition of their chromitites. The Acoje Block contains chromitites with high-Cr spinel, while the Coto Block is characterized by more Al-rich spinel (Zhou et al. 2000). A comparative study of these two blocks found variations and similarities in the HSE budget of the two chromitite types. As in other studies (e.g., Ahmed et al. 2006; Ismail et al. 2014) high-Cr chromitites are found to be richer in HSE than those with high-Al spinel. In this case, however, the IPGE contents vary significantly (e.g., Ru = 8–38 ng/g for Coto, and 62–70 ng/g for Acoje), while Pt and Pd contents and ratios are similar in the two types (Fig. 16) (Zhou et al. 2000). Dunites are also found to vary, particularly in Pt content, with the Acoje Block having more Pt-rich dunites. These spinel compositions and HSE contents are linked to the parental magmas of the chromitites. The Cr-rich Acoje chromitites were likely generated by interaction with a refractory boninitic melt, while the Coto chromitites probably had a more tholeiitic source. Boninitic melts are typically sulfide undersaturated, and thus may form with, and retain, high HSE abundances, compared to tholeiitic melts which are commonly saturated in sulfide thus inducing its precipitation and a reduction in HSE content of the remaining melt (Zhou et al. 2014).
Qalander Ophiolite (Iraq)
The Qalander Ophiolite is a poorly preserved mélange-type complex, containing serpentinized dunites and harzburgites which surround two types of podiform chromitite; high-Al and high-Cr. The harzburgites and dunites analyzed have comparable HSE patterns overall (Ismail et al. 2014), broadly similar to PM estimates (Becker et al. 2006), except offset to higher concentrations (Fig. 16) particularly for Os (4–9 ng/g Ir, 10–17 ng/g Os). As with other chromitite occurrences, Cr-rich and Al-rich types have differing relative proportions of HSE, although they almost all possess high IPGE/PPGE ratios (see Zhou et al. 2014; cf. Shetland, above). Cr-rich chromitites are the most strongly enriched in IPGE, and have the highest IPGE/PPGE ratios. Al-rich chromitites have significantly higher PPGE concentrations, above those of peridotite, while the Cr-rich type has PPGE at the low end of the peridotite range.
Egyptian ophiolites and podiform chromitites, Oman N massifs
The Os isotope composition of PGM from chromitites of the Proterozoic Eastern Desert ophiolite, Egypt and in the Phanerozoic Oman ophiolite were analyzed by Ahmed et al. (2006). It was found that PGM from different regions of each ophiolite have distinct 187Os/188Os ratios, from sub- to broadly chondritic in some regions, to significantly suprachondritic in others (0.1293 for the Proterozoic Eastern Desert ophiolite and up to 0.1459 for the Oman ophiolite). At the same time, there are also distinct compositions of the chromitites themselves, with (i) concordant lensoid forms with intermediate-Cr spinel, which are relatively PGE-poor, and (ii) discordant, dyke-like chromitites, with high Cr spinel, which are PGE-rich. The authors conclude that the variety of chromitites, and the Os–HSE signatures that they contain, reflects the variety of formation processes. The radiogenic chromitites of the Eastern Desert are thought to be affected by crustal contamination, whereas the radiogenic, Cr- and HSE-rich chromitites from Oman reflect high degree melting and an input from a subducting slab, most likely in a supra-subduction zone setting (Ahmed et al. 2006), although here we note that some workers prefer a MOR origin and obducted emplacement for the Oman ophiolite (see earlier section).
Feather River ophiolite (California)
A suite of serpentinized peridotites from the Feather River ophiolite has been compared with serpentinized abyssal peridotites and used as a means of establishing the chemical impacts of serpentinization at a range of water/rock ratios and depths in the mantle (Agranier et al. 2007). The serpentinites have elevated concentrations of seawater-derived fluid mobile elements, such as boron, although typically lower than abyssal peridotites. Feather River serpentinites do not have corresponding seawater-affected supra-chondritic 187Os/188Os ratios (measured range: 0.1175–0.1279). Nonetheless, there is a probable covariation between Os abundance and Os isotope composition in Feather River rocks, which may reflect incorporation of seawater-derived radiogenic 187Os/188Os. Agranier et al. (2007) contend that the serpentinites formed at lower water/rock ratios (greater depth) than is typical for abyssal rocks, and are therefore more representative of bulk serpentinized lithosphere.
In summary, melt percolation in the supra-subduction zone environment generates substantial lithological heterogeneity, which is accompanied by significant Os isotope and HSE variability, both between lithological groups (harzburgites, dunites, chromitites, pyroxenites) and within groups. There is compelling evidence for addition of melt-derived radiogenic 187Os to parts of the mantle sections of ophiolites (see above and Figs. 14 and 15), most probably due to a degree of Os fluxing from the downgoing slab, although other possibilities exist. However, the precise mechanism for such a transfer is not yet clear. The process of melt–rock reaction during melt percolation results in a decoupling of Al2O3 and 187Os/188Os (Fig. 14), which for other suites is considered a fairly robust method for determining the approximate ages of depletion for suites of peridotites, where measured Re contents are often unreliable (Meisel et al. 2001; Lassiter et al. 2014).
Highly siderophile elements in the mantle sections of ophiolites of uncertain origin
Luobusa ophiolite (Tibet)
Chondrite-normalized HSE concentrations for harzburgites, dunites and chromitites from the Luobusa ophiolite are presented in Fig. 16. The concentrations of Ir, Pt and Pd are broadly comparable between two different studies (Becker et al. 2006; Zhou et al. 1996), but the low Os/Ir ratios of the Ni-S fire assay data of Zhou et al. (1996) are not supported by the high temperature (345 °C) isotope dilution data of Becker et al. (2006), suggesting either different petrogenetic histories for the two sample sets or an unidentified analytical issue for Os in the Zhou et al. data. To err on the side of caution, we will assume the latter here and disregard the very low Os/Ir ratios in the harzburgites and chromitites.
The harzburgites appear to represent residua after MORB extraction (Zhou et al. 1996). The HSE abundances are similar to the PM mantle estimate (Becker et al. 2006), and do not indicate significant melt depletion, except perhaps for Pt (although data for Re—the most incompatible HSE—is only available for two samples). The Cr-numbers of Cr-spinel in melt-reacted dunitic rocks are higher than those in the harzburgites, suggesting interaction of a boninitic melt with the residual peridotite, which also removed pyroxene (Zhou et al. 1996). As a result, melts became more boninitic and saturated in Cr-spinel, which precipitated to form chromitite pods within the dunite zones. The inferred boninitic melts suggest a subduction-related origin for this ophiolite. Chromitites have distinct, strongly fractionated HSE patterns with high IPGE/PPGE ratios (e.g., normalized Ir/Pt ratios ~ 100). The concentrations of IPGE in the chromitites are an order of magnitude or more greater than those of the harzburgites, while Pt abundances are approximately five times lower in the chromitites than the harzburgites, and are comparable to the dunites (Fig. 16). These concentrations and patterns are similar to other Cr-rich chromitites from the Qalander and Zambales ophiolites (Zhou et al. 2000; Ismail et al. 2014). Dunites have similar PPGE contents to the chromitites, but without the enrichment in IPGE, due, presumably, to a lack of PGE saturation, and consequent PGM formation, during dunite formation.
Jormua ophiolite (Finland)
Serpentinites, the oxides they contain, and podiform chromitites have all been analyzed for Re–Os abundances and Os isotopes (Tsuru et al. 2000). As with most abyssal peridotites that have undergone serpentinization, Os concentrations, although somewhat variable (1.5–11.7 ng/g) are broadly similar to those of the convecting upper mantle. Rhenium abundances are more variable; most samples are depleted in comparison with PM (Becker et al. 2006) but some experienced (probably recent) Re enrichment. Whole-rock samples have experienced open-system behavior, with respect to Re–Os isotopes, but chromite to Cr-rich magnetite separates have extremely low Re/Os and largely homogenous initial 187Os/188Os values, with a mean initial γOs of approximately −5, suggesting closed-system behavior. Other parts of the ophiolite contain chromitites with initial γOs between +1 and +3. Tsuru et al. (2000) concluded that the positive values may indicate the presence of MORB-type and subcontinental lithospheric mantle sources. Addition of radiogenic Os by melt percolation may be another mechanism to explain the Os isotope data.
Outokumpu ophiolite (Finland)
The Cr-rich nature of residual chromites and boninite-like volcanic rocks suggest a supra-subduction origin for this ophiolite, but an origin in a continental rift zone has also been proposed (Walker et al. 1996). The key conclusion of an Os isotope study (Walker et al. 1996), mainly of chromite, was that this mantle section displayed broadly chondritic 187Os/188Os ratios, and hence Re and Os abundances, which were used to support the ‘late veneer’ model (Chou 1978). In detail, however, there were variations from a ‘residual’ sub-chondritic laurite (Ru(Os,Ir)S2) to fluid addition with a composition of around 0.4 γOs. In this case, however, the radiogenic signature is thought to be derived either from seawater contamination or from a crustal input during emplacement, akin to that proposed for the Eastern Desert Ophiolite, Egypt (see previous section).
Tethyan ophiolites (Turkey)
Harzburgites and dunites from Tethyan ophiolites at Koycegiz, Marmaris, Tekirova, Adrasan and Lake Salda in Turkey have been analyzed by Aldanmaz et al. (2012). Both mid-ocean ridge and supra-subduction zone geochemical signatures have been identified in different parts of the ophiolites, and these have differing HSE systematics. The mid-ocean ridge harzburgites have broadly chondritic Os/Ir and supra-chondritic Pd/Ir and Rh/Ir, similar to PM estimates (Becker et al. 2006), although some PPGE/IPGE enrichment is ascribed to sulfide addition. They also have a sub-chondritic range of measured 187Os/188Os of 0.1223–0.1254, and have correspondingly depleted Re/Os ratios (Aldanmaz et al. 2012). In contrast, the peridotites of supra-subduction zone affinity have more variable HSE patterns and a wider range of 187Os/188Os from 0.1209 to 0.1318, which is −5.3 to 3.3 in γOs90Ma units, relative to O-chondrite evolution. The greater heterogeneity of supra-subduction zone peridotites, compared to those of mid-ocean ridge affinity, reflects a more complex evolution.
Eastern Alps ophiolites (Austria)
Peridotitic units of Eastern Alps ophiolites (the Reckner, Hochgrossen, Kraubath, Steinbach and Bernstein peridotites; including two chromitites) have been found to have remarkably uniform 187Os/188Os ratios (~ 0.1266–0.1281), clustering around the chondritic evolution curve (Meisel et al. 1997), with the exception of one locality (Dorfertal) which has an Os isotope composition consistent with a minimum age of Re depletion of ~ 1.6 Ga. The authors considered the uniformity of Os composition to be somewhat surprising given the uncertain age and affinity of the samples. One important finding of that study was the robustness of Os isotopes, given a high degree of serpentinization, compared with other geochemical data, and even petrographic and field methods.
Mayari-Cristal ophiolite (Cuba)
The key finding of a study of PGM in the Mayari-Cristal ophiolite was the scale of Os isotope heterogeneity present within single hand specimens, thin sections and down to a scale of several millimeters that separated two PGM with contrasting Os isotope ratios (187Os/188Os: 0.1185 and 0.1232; Marchesi et al. 2011), which equate to Re depletion ages of 1370 and 720 Ma, respectively (O-chondrite reference). Given that the budget of Os for these PGM is thought to be sourced from several m3 of mantle, this has intriguing implications for mixing (or the lack thereof) of distinct percolating melts in the mantle (Marchesi et al. 2011).
Influence of low-temperature alteration processes on the HSE in bulk rocks and minerals
Here we briefly discuss the influence of low-temperature (non-magmatic) processes on the bulk rock, sulfide, and PGM composition of mantle tectonites. Ultrabasic rocks affected by oxidative weathering are usually not used for bulk rock chemical analyses to study high-temperature processes. Sulfides are at least partially oxidized by these processes, thus, it is expected that the abundances of chalcophile elements will be disturbed in non-systematic ways. Because areas of ultramafic rocks affected by oxidative weathering are easily identified by their brown color, stemming from ferric iron bearing secondary weathering products, such altered areas can be normally identified and removed.
The influence of serpentinization on HSE abundances and 187Os/188Os
Serpentinization represents another common low temperature alteration process of ultrabasic rocks. Serpentinization reactions occur during the reaction of igneous and metamorphic ultrabasic rocks with seawater or freshwater under a range of geologic conditions and temperatures (e.g., Evans et al. 2013 and references therein). For instance, these processes occur today in oceanic mantle exposed on the seafloor and at greater depth where heated seawater moves within deep-reaching fractures. Similar processes occurred in ultramafic parts of ophiolites during their exhumation on or beneath past seafloors, during tectonic obduction or by reaction with fluids and meteoric water of variable origin during continental collision (Hirth and Guillot 2013). During serpentinization of peridotites, water reacts with olivine, pyroxenes, spinel (to a lesser extent) and sulfides that formed at high temperatures. Depending on temperature and progress of reaction, the new minerals formed include serpentine minerals (chrysotile, lizardite, at higher temperatures antigorite), magnetite and other secondary minerals such as brucite (see for example Bach et al. 2004).
The influence of serpentinization on the abundances of HSE in mantle tectonites has not been studied in much detail. Early Re–Os studies of serpentinized peridotites (e.g., Snow and Reisberg 1995) have emphasized that serpentinization of peridotites in the oceanic lithosphere occurs under reducing conditions. Because of the low fO2 environment caused by the local production of hydrogen and methane (Evans et al. 2013), secondary sulfides (heazlewoodite, millerite, godlevskite), Fe–Ni alloy phases (awaruite) and native metals (Au, Cu) may form (Klein and Bach 2009) and thus, the HSE are able to retain a low valence. The extent to which the HSE are retained in these secondary phases compared to the original abundances in the unaltered bulk rocks and how much of the HSE may be lost into the fluids is poorly constrained. The similarities of abundances of Os, Ir, Ru, Rh, Pt, and Pd in fresh and variably serpentinized peridotites with similar lithophile element composition have been used to argue that serpentinization at reducing conditions results in only minor changes in the abundances of these elements in serpentinized ultramafic bulk rocks that are difficult to resolve from analytical or intrinsic variations in such rocks (e.g., Becker et al. 2006; van Acken et al. 2008; Fischer-Gödde et al. 2011; Marchesi et al. 2013; Foustoukos et al. 2015). This contention is supported by abundances of these elements in serpentinized komatiites, which often preserve correlations between PGE and Mg or Ni, which were unequivocally produced by igneous fractionation processes (e.g., Brügmann et al. 1987; Puchtel et al. 2004, 2005).
The influence of serpentinization on Re and Au abundances is more difficult to predict, as no systematic studies exist and the applicability of experimental studies of Re behavior in specific hydrothermal fluids is difficult to evaluate (Xiong and Wood 1999; Pokrovski et al. 2014). Compared to Pd, Re is often depleted in serpentinized harzburgites, as expected for strongly depleted residues of partial melting; however, it may also be more enriched than Pd in normalized concentration diagrams (e.g., Figs. 3, 5, 9, 10, 16). It is difficult to judge if these abundances reflect secondary addition of Re from seawater (which has very low Re abundances) that has dissolved sulfides elsewhere, or, if re-enrichment of Re occurred before alteration (e.g., by precipitation of liquid sulfide from silicate melts, as may be plausible from observations of unaltered peridotites). Similar uncertainties arise in serpentinized lherzolites. Correlations of Re with indicators of melt extraction or refertilization such as Al, Ca or Mg/(Mg + Fe2+) in peridotites have been interpreted as evidence for limited mobilization of Re by low-temperature alteration processes (e.g., Becker et al. 2006). In mantle pyroxenites that were affected by variable degrees of serpentinization, Re seems to be unaffected by alteration because it is typically systematically more enriched than Pd and Pt. Such a behavior is expected from crystal fractionation products of basic melts (van Acken et al. 2010b). The behavior of gold during serpentinization of mantle peridotites has not been studied systematically either. Although Au, in some cases, follows Pd and Re in its geochemical behavior in unaltered peridotites (Fischer-Gödde et al. 2011), it shows scattered distributions in element variation diagrams that are not well understood. Because of the known mobility of Au in hydrothermal systems in basic and ultrabasic rocks (Pokrovski et al. 2014) and the enrichment of Au in some serpentinite-hosted sulfide deposits (e.g., the Lost City hydrothermal field, Mid Atlantic Ridge), it is to be expected that Au may be rather mobile during serpentinization.
The question of whether or not the Os budget of serpentinized peridotites can be measurably affected by radiogenic 187Os from seawater has been discussed in several publications (e.g., Martin 1991; Roy-Barman and Allègre 1994; Snow and Reisberg 1995; Brandon et al. 2000; Standish et al. 2002; Alard et al. 2005; Harvey et al. 2006). Cenozoic seawater has highly variable and mostly very radiogenic 187Os/188Os ranging over 0.5–1 (Peucker-Ehrenbrink and Ravizza 2000), however, the concentration of Os in seawater is extremely low (about 3.8 fg/g Os, (Sharma et al. 1997). These low abundances are in stark contrast to the ng/g levels of Os in peridotites. Figure 17 illustrates the effects of simple peridotite–seawater mixing, assuming 187Os/188Os of 0.122 and 0.127 and 3.9 ng/g Os in unaltered peridotite and modern seawater with 187Os/188Os of 1 and 3.8 fg/g Os. Very high water–rock ratios of 103–104 are required in order to disturb the 187Os/188Os of peridotite bulk rocks at the % level or higher. Lower values of 187Os/188Os in seawater, such as 0.5, would not alter this conclusion. For comparison, water–rock ratios of significantly less than 100 have been calculated for rock units of the Oman ophiolite (McCulloch et al. 1981). Some workers have suspected that Mn hydroxide films in cracks and on surfaces may pose a problem because these phases tend to scavenge Os from seawater (Martin 1991; Roy-Barman and Allègre 1994). Although leaching studies of serpentinized peridotites have not yielded clear indications of contamination, it is preferable to remove such surfaces or avoid such rocks altogether. Many abyssal peridotites are strongly serpentinized, yet they are characterized by chondritic to subchondritic 187Os/188Os, similar to unaltered or weakly serpentinized post-Archean peridotite xenoliths or other tectonites. Thus there appears to be no need to invoke late addition of radiogenic Os by serpentinization. Positive linear correlations of 187Os/188Os with Al2O3 contents in serpentinized peridotites provide the strongest argument against a significant influence of serpentinization on 187Os/188Os in such rocks (Reisberg and Lorand 1995). These correlations are a primary magmatic feature of mantle rocks (e.g., Handler et al. 1997; Peslier et al. 2000; Meisel et al. 2001; Gao et al. 2002).
Suprachondritic 187Os/188Os occasionally occur in bulk rocks of strongly serpentinized abyssal peridotites (Standish et al. 2002) and from serpentinized harzburgites and dunites of ophiolite sections and peridotite massifs (e.g., Becker et al. 2001; Büchl et al. 2002; Hanghøj et al. 2010). Standish et al. (2002) reported small-scale variations of 187Os/188Os in serpentinized harzburgites and dunites. In the latter study, isotopic differences in chromite (187Os/188Os = 0.124–0.148) compared to bulk rocks (187Os/188Os = 0.118–0.158) were interpreted to result from serpentinization and the addition of seawater-derived radiogenic Os in the altered portion of the rocks. Considering the Os concentration differences between seawater and peridotites, it is not clear how sufficient 187Os can be added from seawater to raise the 187Os/188Os to values higher than 0.15. The Os isotopic data in Standish et al. (2002) cannot be reconciled with low-temperature alteration in a simple way, because Cr rich spinels sometimes have more radiogenic Os than their bulk rocks, and samples with the highest 187Os/188Os are characterized by unusually low Os concentrations (below 1 ng/g). Other workers have interpreted chondritic to slightly suprachondritic initial 187Os/188Os in serpentinized dunites and harzburgites to result from the interaction between magmas with suprachondritic 187Os/188Os and mantle rocks, which, because of magmatic dissolution of sulfide liquid or chromite, may also cause a decrease of Os abundances in peridotites (Becker et al. 2001; Büchl et al. 2002; Hanghøj et al. 2010). Alard et al. (2005) and Harvey et al. (2006) have interpreted different generations of sulfides in serpentinized peridotites from the Atlantic Ocean to reflect magmatic impregnation from percolating magma with suprachondritic 187Os/188Os, similar to observations from continental peridotites (Burton et al. 1999; Alard et al. 2002; Harvey et al. 2011; Luguet and Reisberg 2016, this volume). To conclude, the effects of serpentinization on the 187Os/188Os of serpentinized peridotite are likely minor and difficult to resolve from Os isotopic heterogeneities in mantle rocks inherited from high-temperature igneous processes.
Low-temperature decomposition of primary sulfides in peridotites
Work on sulfide compositions in peridotites and results of experimental data at typical mantle P–T conditions also noted that sulfides in peridotites, in particular sulfides on grain boundaries, display exsolution assemblages from a homogeneous sulfide phase, typically monosulfide solid solution (mss, (e.g., Lorand and Luguet 2016, this volume). The result of these decomposition processes, which depends on the cooling history, is a heterogenous assemblage of intergrown sulfides (commonly pentlandite, pyrrhotite and chalcopyrite), and other minerals, notably platinum metal bearing alloys and Te-, Bi-, Se-rich phases (Alard et al. 2000; Lorand et al. 2010, 2013; Luguet et al. 2003, 2004, 2007). Because of these subsolidus processes, it is not uncommon that some elements (e.g., Pt, Te, Au) become strongly redistributed from sulfides into other trace phases in which they are a major element (e.g., Pt alloys, tellurides, selenides). As a consequence these elements may display negative anomalies in normalized concentration diagrams of exsolved sulfide phases (Alard et al. 2000; Lorand et al. 2010) that are not present on the bulk-rock scale. A detailed discussion of phase assemblages and their composition will be given elsewhere in this volume (Harvey et al. 2016, this volume; Lorand and Luguet 2016, this volume).
The influence of melt infiltration and partial melting on HSE abundances in mantle tectonites
Since the early 1980s, studies of lithophile element geochemistry and Sr–Nd–Pb isotope compositions have shown that mantle tectonites have undergone variable degrees of partial melting during past melting events. Typically this is indicated by their depletion in moderately and highly incompatible elements (e.g., Frey et al. 1985; Johnson et al. 1990) and unradiogenic Sr and radiogenic Nd isotopic compositions (e.g., Jacobsen and Wasserburg 1979; Polvé and Allègre 1980; Reisberg and Zindler 1986). The compositional pattern of major elements in mantle tectonites is such that most abyssal peridotites and ophiolites genetically related to convergent plate margins are harzburgites (and subordinate dunites), whereas lherzolites tend to occur more often in ultra-slow spreading environments, subcontinental settings or continent-ocean transitions. These compositional differences mirror different degrees of partial melting in these settings and are broadly consistent with the polybaric melting column model of upwelling upper mantle (Langmuir et al. 1992). The model predicts that below mid-ocean ridges or other regions of shallow mantle upwelling such as back arc basins, the highest degrees of melting and harzburgitic residues are expected at the top of the mantle, whereas lherzolites should occur at greater mantle depth where less melting occurs.
Subsequent work has established that many peridotites show petrologic and geochemical evidence for a multi-stage history of high-temperature processes (summarized by Bodinier and Godard 2003). These multi-stage processes include melt extraction and later melt infiltration and reaction with existing peridotite, which induces chemical changes in mantle rocks that range from kinetically controlled fractionation of incompatible trace elements (e.g., Vasseur et al. 1991) to significant modal mineralogical changes (Le Roux et al. 2007). The latter processes are capable of converting harzburgites into lherzolites (“refertilization”) by stagnation of magma or repeated influx of magma saturated in a multiphase assemblage of pyroxene(s) ± Al phase (plagioclase, spinel, or garnet) + sulfide in deeper parts of the lithospheric mantle. Melt–rock reaction in shallow mantle tends to produce tabular dunite, rather than lherzolites (Kelemen et al. 1995, 1997) or plagioclase-pyroxene bearing impregnations, dikes and pockets in otherwise depleted harzburgite (Edwards and Malpas 1996; Seyler et al. 2004). As a consequence of these processes, the inventory of incompatible elements and their isotopic composition in these metasomatically modified rocks is mostly derived from the magma that produced these changes (for instance, the LREE-depleted compositions of lherzolites from Lherz in the Pyrenees and their isotopic compositions must have been inherited from the infiltrating magma, Le Roux et al. 2007). Melt infiltration and chemical reaction with peridotite has been recognized as an important process in many mantle tectonites from different tectonic settings (e.g., Pyrenees, Ronda, Ligurides, Ivrea Zone, Lanzo, Horoman, abyssal peridotites, ophiolites). It may be ubiquitous in melting columns, mantle diapirs and in the deep lithosphere and should be considered normal for open-system melting environments. In the following, we first discuss some general compositional constraints from peridotites that may be linked to melting processes. We then address the influence of reactive melt infiltration on sulfide–silicate equilibration and discuss partitioning of the HSE.
Behavior of the HSE during partial melting of harzburgites and lherzolites
A general observation is that harzburgites have similar abundances of Os, Ir, and Ru (IPGE, Barnes et al. 1985) to lherzolites, whereas concentrations of other PGE, Re, and Au are typically much lower in harzburgites than in lherzolites (Figs. 3, 5, 9, 10, 16). On the other hand, basalts and komatiites often have higher chondrite-normalized concentrations of Pt-group PGE (PPGE: Rh, Pt, Pd, Barnes et al. 1985), Au and Re than IPGE (Hertogen et al. 1980; Brügmann et al. 1987; Rehkämper et al. 1999; Puchtel et al. 2004; Bezos et al. 2005). These studies have pointed out that the main host phase of the HSE in lherzolites at high temperatures should be sulfide. Thus, the stronger depletion of Rh, Pt, Pd, Au, Re, and sulfur in harzburgites compared to lherzolites likely reflects the consumption of sulfide in peridotite during high degrees of melting (e.g., Barnes et al. 1985; Morgan 1986; Lorand 1988; Keays 1995). The details of sulfide dissolution and HSE partitioning into basic magma have remained unclear, particularly for melting processes at P–T conditions that should yield lherzolite residues. Many workers have advocated sulfide–silicate partitioning (e.g., Brenan et al. 2016, this volume, and references therein). For chalcophile element partitioning, the assumption has been that during local partial melting in the mantle, a homogeneous sulfide liquid or solid should coexist in equilibrium with silicate melt, olivine, pyroxenes and an Al-rich phase (Keays 1995; Morgan 1986; Rehkämper et al. 1999). The amount of sulfide liquid dissolved into the silicate melt is controlled by ambient pressure, temperature and FeO content of the melt (Mavrogenes and O’Neill 1999; O’Neill and Mavrogenes 2002; Jugo et al. 2005). Another partitioning process, mss–liquid sulfide partitioning, that was also proposed to control HSE abundances (Bockrath et al. 2004) will be discussed below. During melting of upwelling asthenosphere or deep lithosphere, at temperatures > 1250 °C, it is expected that mantle rocks and coexisting magmas were chemically and isotopically equilibrated, as is commonly assumed for lithophile elements (Hofmann and Hart 1978).
At high degrees of melting, partitioning of Os, Ir, Ru, Rh, and Pt may be controlled by the solubility of alloys of these elements in silicate melt (Pearson et al. 2004; Fonseca et al. 2011, 2012; Mungall and Brenan 2014; Brenan et al. 2016, this volume). The significance of this for the composition of harzburgites will be discussed later. Here, we specifically focus on processes during low and moderate degrees of melting in the deeper regions of the melting column where sulfide should be stable in the residue and sulfide–silicate partitioning has been proposed as the main control on the distribution of the HSE and other chalcophile elements (Barnes et al. 1985; Morgan 1986). However, it has been unclear if sulfide exists as a solid phase (mss), liquid sulfide, or both. Recent improvements in the accuracy and precision of liquid sulfide–silicate partition coefficients (Dsulf/sil) indicate values in the range of 105–106 and 104 for the PGE and Au, respectively (Li and Audétat 2013; Mungall and Brenan 2014; Brenan et al. 2016, this volume), whereas Re is much less chalcophile (Dsulf/sil ≈ 300–800, Fonseca et al. 2007; Brenan 2008). Assuming a simple fractional melting process (batch melting yields similar results as long as the elements are not highly incompatible), element concentrations in the residues can be calculated according to the mass balance equation Cr = Co (1-F)((1/Db)−1), with Cr = concentration of an element in the residue, Co = total concentration of an element in the bulk system (residue + melt), Db = bulk partition coefficient of an element between residue and melt, F = melt fraction. As long as sulfide is present in the mantle residue and it is equilibrated with silicates and silicate melt, the high Dsulf/sil require nearly constant concentrations of all PGE in peridotites (Fig. 18a), because bulk partition coefficients of the PGE in lherzolites are ≫ 1: At 0.02 wt% S in fertile lherzolite and 35 wt% S in monosulfide solid solution, DPGEb > 0.00057 × 105 + 0.9994 × 0.1 = 57, assuming DPdsil.min./sil.melt < 0.1 with other PGE likely having higher Dsil.min./sil.melt (Mungall and Brenan 2014). Gold should also be retained in lherzolites that have lost a significant fraction of melt (DAub ≥ 0.00057 ×5 ×103 = 3, assuming DAusil.min./sil.melt < 0.01 (Mungall and Brenan 2014), whereas Re should be moderately depleted at relevant fO2 in normal upper mantle (FMQ – 1), as its Db is always below 1 in cases where no garnet occurs in the residue (DReb ≤ 0.00057 ×800 + 0.9997 ×0.1 = 0.6, assuming DResil.min./sil.melt < 0.1 (no garnet), Mallmann and O’Neill 2007; Brenan 2008).
The situation in mantle rocks, however, has been found to be more complicated; one indication being the difficulty in reproducing peridotite HSE patterns by sulfide–silicate equilibrium partitioning (Fig. 18a). In the following, we discuss evidence suggesting that many mantle peridotites are in chemical disequilibrium regarding chalcophile element partitioning at the scale of hand specimen to grain boundaries. An alternative partitioning scenario, such as mss–sulfide liquid–silicate liquid equilibrium, is also discussed below.
Melt infiltration induces chemical disequilibrium of chalcophile elements in mantle peridotites
Studies of chalcophile element abundances in sulfides of different textural position, in mantle xenoliths and in peridotite tectonites, have shown that significant compositional differences may exist between sulfides that occur as inclusions in olivine (and sometimes pyroxenes and spinel) and sulfides present at grain boundaries. The former are rich in Ir-group PGE and depleted in Pd, Au, and Re, while the latter may or may not be depleted in IPGE and have higher Pd, Re, and Cu (Alard et al. 2000, 2002; Luguet et al. 2001, 2003, 2004). Although these different assemblages are sometimes complicated by internal separation into multi-phase assemblages (pentlandite, pyrrhotite, and other phases) that occurred late during slow cooling, it is clear from their different compositions that included and grain boundary sulfides were not chemically equilibrated during their formation. The sulfide assemblages on grain boundaries are sometimes associated with pyroxene–spinel assemblages that have been interpreted to have formed during melt infiltration and refertilization. From this observation, it follows that reactive melt infiltration likely led to sulfur saturation in these magmas and precipitation of the sulfides located on grain boundaries (e.g., Alard et al. 2000). The reaction of silicate melts and sulfide segregation processes are not only indicated by the different sulfide assemblages in the peridotites, but also by the HSE abundances in mineralogically zoned boundaries between pyroxenites and host peridotites and disequilibrium sulfide assemblages in mantle pyroxenites (see sections on mantle pyroxenites below).
Some authors have proposed that sulfide melts may be mobile in mantle rocks, and thus may change the Re–Os and PGE systematics of mantle rocks (Gaetani and Grove 1999). The existing data on peridotites, however, do not support pervasive or widespread sulfide melt mobility, as linear correlations between 187Os/188Os, Re, and S abundances and lithophile elements such as Al, Ca, or Mg in peridotites would not be maintained over long periods of time in the mantle (Figs. 4, 7; e.g., Reisberg and Lorand 1995; Meisel et al. 2001; Becker et al. 2006; Wang and Becker 2013), although minor mobility is not precluded due to scatter in the datasets. The role of fluids as metasomatic agents in the redistribution of HSE and other chalcophile elements has been invoked in some cases (e.g., Lorand and Alard 2010). One possibility is that such fluids are the end products left after crystallization of mantle-derived melts or, if they are of external origin, may have been derived from crustal sources at lower temperatures during the exhumation history of mantle tectonites. Regardless of the origin of the fluids, what is not yet clear is the effect of these small-scale observations on the mass balance of bulk rocks. In summary, silicate melts are the main metasomatic agents that, by way of coupled precipitation of sulfide melt, pyroxenes and an Al phase, clearly produce significant modifications of HSE abundances and 187Os/188Os at magmatic temperatures in the mantle.
Detailed surveys of the accessory mineral inventory of peridotites (e.g., Fig. 6) have revealed the occurrence of Pt–Ir alloys, Ru–Os-bearing sulfides and Os–Ir–Ru alloy phases (Luguet et al. 2007; Lorand et al. 2010; O’Driscoll and González-Jiménez 2016, this volume). These phases are expected to become stabilized by decreasing fS2 shortly before or during the exhaustion of liquid sulfide in harzburgite residues at moderate to high degrees of melting (e.g., Fonseca et al. 2012; Mungall and Brenan 2014; Brenan et al. 2016, this volume). Thus, their occurrence in harzburgites (e.g., at Lherz; Luguet et al. 2007) is not unexpected.
However, such phases have also been detected in lherzolites from Lherz that formed by refertilization, albeit they occur in smaller proportions than in harzburgites (Lorand et al. 2010). If the alloy phases were indeed inherited from more depleted parent rocks, their presence in some lherzolites may also reflect chemical disequilibrium between these phases and the more abundant sulfide minerals that were precipitated as sulfide liquid from silicate melt. The impact of such inherited and presumably ‘residual’ alloy phases on bulk rock budgets of lherzolites that formed by refertilization appears rather limited. For instance, the bulk rock Os/Ir ratios of lherzolite tectonites is rather homogeneous and overlaps chondritic values (Fig. 19a, Pearson et al. 2004; Becker et al. 2006; Fischer-Gödde et al. 2011; Liu et al. 2009; Wang et al. 2013). Because of the different solubilities of Os and Ir metal in silicate melt (e.g., Mungall and Brenan 2014), chondritic Os/Ir are not a priori maintained in residual peridotites at higher degrees of melting (as witnessed by the larger scatter of this ratio in harzburgites). Pt/Ir and Pt/Os in lherzolites range from chondritic to mildly subchondritic. Only rarely do lherzolites display enrichments of Pt that are decoupled from Pd, Au, and Re (e.g., Fig. 5b, c) and might be ascribed to the excess presence of Pt minerals. In this context, it is noteworthy that ratios of Ir, Os, and Ru in mantle tectonites tend to be more scattered in harzburgites than in lherzolites (Fig. 19). The difference in homogeneity of the different rock types may either reflect digestion problems in the laboratory, i.e., the difficulty of complete dissolution of refractory platinum group metal alloys in harzburgites (Meisel and Horan 2016, this volume, and references therein), or it may be due to dissolution of refractory alloy phases in coexisting sulfur-undersaturated melt at high temperatures.
Osmium isotopic disequilibrium within mantle peridotites
Evidence for small-scale chemical disequilibrium regarding chalcophile elements is provided by Re–Os data that suggest that grain- to hand specimen-scale Os isotopic disequilibrium is common in the mantle. Burton et al. (1999) found that different mineral separate fractions from mantle xenoliths showed differing 187Os/188Os that were not related by isochronous behavior. Leaching experiments of powders of refertilized mantle xenoliths and tectonites show that 187Os/188Os frozen in during the Archean or Proterozoic survived Phanerozoic refertilization, most likely because of the preservation of ancient chromite or olivine that contained inclusions of HSE carrier phases (Chesley et al. 1999; Becker et al. 2006; Wang et al. 2013). Alard et al. (2002, 2005) showed that the sulfide populations with different PGE compositions also display systematic differences in Re/Os and 187Os/188Os. In peridotite xenoliths and abyssal peridotites, sulfides on grain boundaries tend to have chondritic to suprachondritic Re/Os and 187Os/188Os, whereas sulfides in inclusions also display subchondritic values (Harvey et al. 2006, 2011; Warren and Shirey 2012). The heterogeneous 187Os/188Os in different bulk rocks of essentially all suites of peridotites, xenoliths or tectonites from different geodynamic environments (e.g., Figs. 1, 2, 4, 7, 9, 11; and Luguet and Reisberg 2016, this volume) also represents a manifestation of disequilibrium on the scale of hand specimen and outcrops. In principle, such variation may have been caused by differences in the age of partial melting and melt infiltration. However, evidence for grain-scale initial Os isotopic heterogeneity at times of melt infiltration (in cases where the timing can be constrained) suggests that mixing of residues and melts with different 187Os/188Os during reactive melt infiltration did not result in full Os isotopic equilibrium. A good example are the ultramafic tectonites in the Pyrenees and in the Italian and Swiss Alps (Baldissero, Balmuccia, Lanzo, Totalp), where episodic melt infiltration into Proterozoic continental lithospheric mantle during Paleozoic and Mesozoic extension only partially re-equilibrated 187Os/188Os values. All these data and observations suggest that disequilibrium must have been maintained even at high temperatures in the upper mantle and in the presence of silicate melt. The widespread heterogeneity of initial 187Os/188Os at the grain boundary- to centimeter-scale in mantle rocks also suggests that sulfide liquids are efficiently trapped even during recrystallization processes.
Alongside evidence from textures and lithophile elements (e.g., Rivalenti et al. 1995; Müntener et al. 2005; Le Roux et al. 2007; Mazzucchelli et al. 2009), the extent of re-equilibration is manifested in the scatter of HSE abundances displayed by different suites of peridotites, in the abundance of harzburgite rocks in outcrops and in the distribution of Re–Os model ages in these bodies. At Lherz, Lanzo and Baldissero Re depletion ages of peridotites display bimodal distributions of Proterozoic and Phanerozoic ages, with harzburgites or depleted lherzolites typically showing older model ages (i.e., lower measured 187Os/188Os) than lherzolites (Reisberg and Lorand 1995; Burnham et al. 1998; Becker et al. 2006; Fischer-Gödde et al. 2011; Wang et al. 2013). In contrast, at Balmuccia and Totalp, depleted lherzolites and harzburgites are rare and display Proterozoic Re depletion ages. Model ages of fertile lherzolites at these locales range from Phanerozoic to future ages (van Acken et al. 2008, 2010a; Wang et al. 2013). Of note is that the scatter of the concentrations of Os, Ir, and Ru in fertile peridotites at these localities is more limited than in other lherzolite bearing tectonites (compare Fig. 5b with 5a and 5c).
Osmium isotopic heterogeneity is also prevalent in abyssal peridotites, which are commonly presumed to represent melting residues of MORB-type magmas. Harvey et al. (2006) have shown that sulfides in harzburgites from the 15° 20′ N fracture zone (Atlantic Ocean) preserve small-scale isochronous relationships that date back to the Paleo-Proterozoic. Such preservation of early- to mid-Proterozoic 187Os/188Os values in bulk rocks and sulfides has also been reported in other abyssal peridotites (Parkinson et al. 1998; Alard et al. 2005; Liu et al. 2008; Warren and Shirey 2012). Further evidence of small-scale disequilibrium is apparent in studies of platinum-group minerals from ophiolites. Platinum-group minerals from the Mayari-Cristal Ophiolite, Cuba, have been found to have diverse 187Os/188Os ratios even on the scale of a single thin section (Marchesi et al. 2011). The most extreme example found was the presence of two PGM only a few millimeters apart, with 187Os/188Os ratios of 0.1185 and 0.1232 (Marchesi et al. 2011), which give TRD ages of 1370 and 720 Ma, respectively (ordinary chondrite reference evolution line; Walker et al. 2002a). The mechanism of formation for such PGM is not well known, but given that the budget of Os for these PGM is thought to be sourced from at least several m3 of mantle (total Os equivalent to ~ 1 m3 mantle), this would imply little if any mixing of percolating melts, or a lack of equilibration between mineral grains and subsequent percolating melts.
The influence of disequilibrium between mantle and magmas on HSE distributions
The predicted behavior of the HSE can be compared with HSE patterns of peridotites. Relatively ‘constant’ concentrations have been noted for the Ir group PGE in many studies of lherzolite tectonites. However, Rh and Pt display a tendency towards higher concentrations in lherzolites (e.g., Fischer-Gödde et al. 2011). In some (but not all) suites of peridotites, Pd correlates with fertility indicators such as Al2O3 abundances (e.g., Becker et al. 2006). Some workers (e.g., Lorand et al. 1999; Pearson et al. 2004) have noted that the variable depletion of Pd in lherzolites is difficult to reconcile with partial melting and very high sulfide–silicate partition coefficients (> 104 to 105). The smooth depletion of Pd, Au, and Re relative to other HSE in lherzolites from Balmuccia and elsewhere (e.g., Fig. 18) is inconsistent with equilibrium partitioning and the liquid sulfide-liquid silicate partitioning data. It is also difficult to explain by other equilibrium partitioning processes involving sulfides, e.g., mss–liquid sulfide (see below). Furthermore, concentrations of Os, Ir, and Ru in peridotite tectonites of similar lithophile element composition display considerable scatter (e.g., Fig. 5), as do Os isotopic compositions. For lherzolites, at least, the different concentrations cannot entirely be an artifact of heterogeneous distribution of sulfide grains within sample powders or the rock (Meisel and Moser 2004; Meisel and Horan 2016, this volume). Instead, these concentration variations may reflect the compositional variability of sulfide grains in the rock; as indicated by variable Ir and Ru concentrations in peridotitic sulfides (e.g., Alard et al. 2000). As there is indisputable evidence for widespread, or even ubiquitous, chemical and isotopic disequilibrium of the HSE in peridotites, it is plausible that the distribution of chalcophile elements between peridotite and magma is partly controlled by the composition of sulfide liquids from infiltrating primitive magmas and partly by mixing processes between such liquids and sulfide liquids already present in the rocks (e.g., Lorand et al. 1999; Alard et al. 2000; Pearson et al. 2004; Lorand et al. 2010).
In the melting model shown in Figure 18b apparent sulfide–silicate partition coefficients were used to match the patterns of peridotites from the Balmuccia peridotite massif. Apparent partition coefficients take into account the extent to which the HSE composition of peridotites displays the effects of mixing, and thus the influence of the original infiltrating melt compositions, rather than just sulfide melt–silicate melt equilibrium. It is clear that the fractionations inherited from the melt contribute to the lowering of Db, compared to the equilibrium case. The differences will be particularly notable for Pd and Au. As Pd in depleted lherzolites is commonly slightly depleted, the apparent bulk distribution coefficient for this element should be < 1 and apparent sulfide–silicate distribution coefficients in the model in Figure 18b would be about 1300; far lower than the 105–106 range for sulfide–silicate equilibrium (Mungall and Brenan 2014). For Pt and Rh apparent partition coefficients may also be lower. Gold abundances in depleted lherzolites are lower than in fertile lherzolites and this, coupled with the slight enrichment of Au in primitive basaltic magmas, suggests that Au also has an apparent bulk distribution coefficient < 1. Consequently, apparent sulfide–silicate distribution coefficients for Au are significantly lower (about 200 in the case of Fig. 18b) than equilibrium values (4000–10000; Mungall and Brenan 2014). Rhenium and other moderately chalcophile elements with equilibrium sulfide–silicate partition coefficients < 1500 are not sensitive enough to identify chemical disequilibrium, as the influence of the silicate mineral–silicate melt partition coefficients is substantial. Combined sulfide–silicate and silicate mineral–silicate melt partition coefficients of these elements yield bulk partition coefficients < 1, whether or not equilibrium is assumed.
Figure 20 displays the variation of Re concentrations versus Pd concentrations in various suites of mantle tectonites (note that in more strongly serpentinized peridotites, such as from the Oman ophiolite, Re may also be affected by late-stage alteration). Both elements tend to correlate in harzburgites and in depleted lherzolites, however, in more fertile rocks, Re displays larger variations (0.07–0.4 ng/g) at relatively constant Pd (5–9 ng/g). The most likely explanation for this observation is that sulfide and other HSE carrier populations in harzburgites and depleted lherzolites reflect mixing and full disequilibrium, whereas pre-existing phases in fertile lherzolites may have partially reacted and equilibrated with a larger fraction of silicate melt and sulfide liquid. The data also suggest that HSE carriers in fertile peridotites of some suites (e.g., Balmuccia and Baldissero) must be more depleted in Re than other suites, which may be a property of the melts that precipitated sulfides during reactive infiltration. The curved trend defined by some data in Figure 20 may be related to the quantity of melt that reacted and precipitated sulfide liquid in the rock. The systematic behavior of Pd, Au, Re, and of other chalcophile elements such as S, Se, Te, Cu, and Ag in most peridotites and in MORB (Wang and Becker 2015b) indicates that the relative depletion and enrichments of these elements in peridotites and in MORB may be described by apparent bulk partition coefficients. Melt compositions calculated by this approach may yield similar concentrations of Pd, Au, and Re as in primitive MORB, although the latter almost certainly require a more complicated fractionation history (e.g., Langmuir et al. 1992; Rehkämper et al. 1999; Bezos et al. 2005; Mungall and Brenan 2014; Wang and Becker 2015c).
An alternative model of HSE partitioning during mantle melting was presented by Bockrath et al. (2004) and Ballhaus et al. (2006). These authors proposed that residual mss may coexist with liquid sulfide over a significant pressure-temperature range in the mantle. Partitioning between these phases may control the HSE abundances in residues and silicate melts. However, because of uncertainties in the position of the sulfide liquidus in different experimental studies, the stability of mss in the asthenosphere or deeper lithosphere is debated (see Fonseca et al. 2012; Mungall and Brenan 2014). The relevance of mss–liquid sulfide partitioning in the upper mantle can be evaluated on the basis of existing partitioning data for chalcophile elements and the composition of mantle rocks, basalts and their sulfides. Melting models of bulk rock compositions of lherzolites that employ mss–liquid sulfide partition coefficients (Fig. 18c) display a poor match for Pt, Pd and Au. However, it must be acknowledged that bulk partition coefficients are strongly influenced by the silicate mineral–silicate melt partition coefficients. Only for olivine-silicate melt partitioning does sufficient data exist for Pt, Pd, and Au (see Eqns. (11)–(13) in Mungall and Brenan 2014, which yield low Dolivine/silicate melt for these elements at fO2 of 10−9–10−10 bar). Pyroxene-silicate melt partition coefficients for these elements are poorly constrained, and thus Db may be higher. As for sulfide liquid-silicate partition models, Re fits well because its Db is strongly controlled by the large mass fraction of silicates and the well-determined mineral–silicate melt partition coefficients.
In principle, mss–liquid sulfide partitioning may account for the different patterns of Ir-group and Pt-group PGE in sulfide inclusions and sulfides on grain boundaries in peridotites (e.g., Ballhaus et al. 2006). However, the behavior of Re concentrations in sulfide inclusions versus grain boundary sulfides argues against this process. Equilibrium mss-–liquid sulfide partitioning would predict higher Re and Os concentrations in residual sulfides compared to coexisting sulfide liquids, because both elements are compatible in mss (DOsmss/sul liq = 3–7, DRemss/sul liq = 3, Brenan 2002; Ballhaus et al. 2006). Although sulfide inclusions in silicates of peridotites may have higher Ir and Os than sulfides on grain boundaries (e.g., Alard et al. 2000, 2002), Re is depleted in the former and enriched in the latter, commonly accompanied by correlated Re/Os (Alard et al. 2005). Recently, it has been proposed that some harzburgites contain sulfides with high Se/Te ratios similar to what is expected from mss–liquid sulfide partitioning (König et al. 2014, 2015). However, because of the low concentrations of these elements, the mass balance of such phases in strongly depleted peridotites is difficult to constrain, and they may also reflect precipitation of sulfide from somewhat more fractionated magma with high Se/Te and Re/Os (Wang and Becker 2015a). Work on Cu and Ag abundances in peridotites has shown that the relative behavior of these elements in bulk rock lherzolites is consistent with the systematics predicted by sulfide liquid-silicate partitioning but not with mss–liquid sulfide partitioning (Wang and Becker 2015b).
The differing 187Os/188Os of the two sulfide populations suggests that sulfides precipitated on grain boundaries during melt infiltration did not equilibrate with included sulfides, which is a basic requirement for equilibrium mss–sulfide liquid–silicate melt partitioning models. Thus, as shown before in the discussion of sulfide liquid–silicate melt partitioning, none of the proposed partitioning processes that are potentially relevant during partial melting yields a satisfactory quantitative description of the HSE composition of many mantle peridotites. Sulfide melt–silicate melt partitioning seems to be the best match for the observed HSE pattern in lherzolite bulk rocks. However, at least for Pd, Au, Re, and S, their ratios in lherzolites may be mostly inherited from the melts that infiltrated depleted precursor rocks (e.g., harzburgites; Fig. 20). The origin of the HSE fractionation in the infiltrating melts and their sulfide liquids will be discussed below.
HSE fractionation during the formation of mantle pyroxenites
Mantle pyroxenites are important because they represent products of magmatic fractionation in the mantle and thus yield information on the composition of relatively ‘primitive’ magmas (Bodinier and Godard 2003). Pyroxenites are cumulates that formed by reactive infiltration and fractional crystallization of primitive to more evolved basic magmas. Websterites (‘Cr diopside suite’) and orthopyroxenites sometimes display mineralogically zoned reaction domains with peridotites, which have formed due to melt infiltration into the surrounding peridotite (e.g., Bodinier et al. 1987, 2008; Becker et al. 2004). Quite often, clinopyroxenites (‘Al augite suite’) appear to have formed from more evolved compositions and the absence of reaction zones may indicate their formation at shallower levels (e.g., Sinigoi et al. 1983; Suen and Frey 1987).
Only limited data are available for HSE abundances and Os isotopic compositions in mantle pyroxenites from tectonites, including pyroxenites from Ronda (Reisberg et al. 1991; Reisberg and Lorand 1995; Marchesi et al. 2014), Beni Bousera (Kumar et al. 1996, Pearson and Nowell 2004; Luguet et al. 2008b), Lower Austria (Becker et al. 2001, 2004), Troodos (Büchl et al. 2002), Totalp (van Acken et al. 2008, 2010b), Hori Bory (Ackerman et al. 2013) and Balmuccia (Wang and Becker 2015c). The HSE patterns of pyroxenites in mantle tectonites are broadly similar to data from sulfides in pyroxenite xenoliths. In general, the relative fractionation of the HSE is similar to that in basalts, but with higher concentrations of Os, Ir, Ru, Rh, Pt, and Pd than in MORB. Websterites and orthopyroxenites often display HSE patterns that are less strongly fractionated than clinopyroxenites (Fig. 21).
Concentrations of S and Re in pyroxenites are similar or lower than in MORB, but often higher than in lherzolites. Abundances of other HSE in pyroxenites are similar or lower than in lherzolites (Fig. 21). Some pyroxenites display a depletion of Re relative to Pd, which may have been caused by multi-stage melting (Marchesi et al. 2014). The occurrence of cm-scale Os isotopic heterogeneity between alternating pyroxenite–peridotite layers (Becker et al. 2001, 2004; Büchl et al. 2002; van Acken et al. 2008) is another indication of the difficulty of small-scale Os isotopic equilibration between silicate melt and existing sulfide populations. A study of a zoned clinopyroxenite–websterite–orthopyroxenite rock from Lower Austria that represents a former reaction zone between high-temperature silicate melt and peridotite has shown that Sr and Nd isotopic compositions were equilibrated across a 10-cm distance of the rock at the time of its formation (Becker et al. 2004). In contrast, both γOsi and Os concentrations display strong gradients over the same distance, indicating disequilibrium. HSE compositions of sulfides in single thin sections of Totalp pyroxenites vary from those with Ru/Ir, Pd/Ir, and Re/Ir similar to peridotitic sulfides, to those with high ratios of these elements, typical of melt compositions (van Acken et al. 2010b). The detailed processes that resulted in the close association of these different sulfide populations are not yet clear, but they suggest that disequilibrium among sulfides may be common in mantle pyroxenites as well as peridotites.
A comparison of Re/Os and Pd/Ir in pyroxenites with data on ocean ridge basalts and gabbros from the lower oceanic crust indicates considerable overlap (Fig. 22). This observation suggests that significant fractionation of HSE ratios in magmas already occurs by precipitation of sulfide liquid during magmatic transport and reaction in the mantle (Wang and Becker 2015c). In contrast to Re/Os, which shows large variations in magmatic products over several orders of magnitude, the variation of Pd/Ir in the latter is much more limited and Pd and Ir show similar bulk partitioning behavior. Because of the segregation of sulfide liquid from magmas during magmatic transport in the mantle, the HSE compositions of basaltic magmas may preserve little direct information on HSE concentrations of deeper parts of the melting region. Figure 22a also shows that the data fields defined by most magmatic products, particularly the basalts, are offset from the bulk compositions of peridotites, but overlap with ratios in grain boundary sulfides from peridotites. A similar observation was made for variations of Se/Te (Wang and Becker 2015c). This observation may provide the best indication so far that most magmas that contribute to the oceanic crust did not fully equilibrate with the bulk rock of mantle peridotite residues.
HSE fractionation during the formation of harzburgites and replacive dunites
Data on HSE and other chalcophile elements in harzburgites show that many of these rocks have high abundances of IPGE and lower abundances of Rh, Pt, and Pd (e.g., Pearson et al. 2004; Becker et al. 2006; Luguet et al. 2007). These IPGE–PPGE fractionations are generally consistent with fractionation of melting residues at moderate to high (15–30 %) degrees of partial melting (Mungall and Brenan 2014; Brenan et al. 2016, this volume, and references therein). The incongruent breakdown of liquid or solid sulfide occurs at advanced degrees of melting at low fS2 and may play an important role in the stabilization of Os–Ir–Ru and Pt–Ir alloy phases that have been found in such rocks (Lorand et al. 1999, 2010; Luguet et al. 2007; Fonseca et al. 2012; Mungall and Brenan 2014; Brenan et al. 2016, this volume). With progressive melting in the absence of a Fe–Ni-rich sulfide phase, all Re, Au, and Pd should be dissolved in coexisting melts, provided that residues and melts were equilibrated. The abundances of Os, Ir, Ru, Rh, and Pt, and their fractionation in harzburgite residues (e.g., Fig. 20) should be controlled by the solubility of these elements in sulfur-bearing silicate melts and the stability of Os–Ir, Ru–Os, and Pt–Ir phases (Mungall and Brenan 2014).
However, harzburgites may show variations in HSE abundances that are not entirely consistent with a simple melting history as envisioned before. Normalized abundances of Re and S in harzburgites are sometimes higher than normalized abundances of Pd (Figs. 5, 10). These patterns have been interpreted either in terms of precipitation of secondary sulfides from infiltrating melts with high Re/Os and fractionated HSE patterns (Chesley et al. 1999, Pearson et al. 2004, Becker et al. 2006; Wang and Becker 2015a). Alternatively, enrichments of Re and S compared to Pd and Pt (and of Se relative to Te) in some harzburgites have been interpreted to reflect the presence of mss of residual origin (König et al. 2014). The former explanation is consistent with magmatic re-enrichment processes of incompatible elements (e.g., light rare earth elements) in some of these rocks. Some harzburgites display lower abundances of IPGE than expected for depleted mantle peridotite, e.g., < 3 ng/g Ir, instead of 4–5 ng/g expected for residues of moderate to high degrees of melting (Figs. 5, 10). In order to understand this behavior, it is useful to recall that even at high temperatures most peridotites likely contain unequilibrated sulfide melt (maybe also mss), with a range of HSE concentrations. Complete dissolution of some of these sulfide droplets (but not others) into sulfur-undersaturated melt, without concurrent precipitation of IPGE alloy phases, will result in a net decrease of the abundances of all HSE. This process almost certainly plays an important role in the formation of some replacive dunites and associated harzburgite-–lherzolite–pyroxenite rock assemblages (Becker et al. 2001, 2004; Büchl et al. 2002, 2004; Hanghøj et al. 2010; Wang et al. 2013). For instance, the variable IPGE abundances and strong depletions of Pt, Pd, Re and other chalcophile elements in discordant dunite bodies in lherzolites at Balmuccia indicate that the magmas were undersaturated in sulfur, which caused the dissolution of sulfides from the lherzolitic protoliths of the dunites (Fig. 5, Wang et al. 2013).
The harzburgites from Wadi Tayin (Oman ophiolite) display normal abundances of IPGE and tend to show primitive mantle-like or even slightly suprachondritic abundances of Pt, Pd and Re (Lorand et al. 2009; Hanghøj et al. 2010). Some of the harzburgites show selective enrichments of Pt that also have been noted from abyssal peridotites and other ophiolites (Fig. 10) and peridotite massifs (Fig. 5). The Pt enrichments may indicate the precipitation of Pt-enriched sulfide liquid from silicate melt that may have dissolved Pt from destabilized Pt–Ir alloys at high degrees of melting. Dunites from Wadi Tayin are similarly enriched in HSE, but show more fractionated Re/Os and PPGE/IPGE ratios. Because the dunites are thought to reflect pathways of olivine-saturated magmas, the enrichments of Pt, Pd, and Re in dunites and harzburgites likely reflect sulfide segregation from magmas enriched in these elements (Fig. 23). Although this process appears to have occurred pervasively, the initial 187Os/188Os (at around 90–95 Ma) in the mantle section at Wadi Tayin were not equilibrated (Fig. 23). The high abundances of Pt, Pd, and Re in otherwise incompatible element depleted mantle rocks suggest that sulfide saturation may play an important role in the uppermost mantle underneath fast-spreading ocean ridges. Dunites from the Troodos ophiolite also display ‘melt-like’ HSE compositions (Büchl et al. 2002). A common property of dunites is that their initial 187Os/188Os ratios extend to suprachondritic values (γOsi ranging from −3 to +17, e.g., Fig. 23 and Becker et al. 2001), suggesting that some of the parent magmas had suprachondritic Os isotopic compositions. However, as the case of the dunites from Balmuccia shows, not all dunites are characterized by an enrichment of Pt, Pd and Re and melt like HSE patterns.
PGE enrichments also occur in podiform chromitites, which are magmatic precipitates associated with dunites and harzburgites in ophiolites that formed in the proximity of convergent plate margins. Because chromitites may represent economically relevant sources of PGE, these high-temperature magmatic ore deposits will be discussed in Barnes and Ripley (2016, this volume).
Summary—Mantle melting and mantle–magma interaction—different sides of the same coin
Models of partial melting of mantle tectonites must consider the natural open-system behavior relevant for melting column models, diapiric upwelling of partially molten mantle or conversion of lithospheric mantle to asthenosphere by melt infiltration (as was suggested to have occurred in the magmatic history of some mantle tectonites, e.g., Müntener et al. 2005). Thus, melt infiltration and melting should occur more or less simultaneously, provided that porosity and permeability permit melt infiltration. The composition of the residues will change with time until external processes cause upwelling and melting to stop and the mantle to cool. The HSE concentration and 187Os/188Os data on mantle tectonites with well-constrained ages (e.g., Oman ophiolite) show that the extent of sulfide–silicate equilibrium in these melting processes must be limited. Several different types of sulfide (presumably mostly liquids, but also mss and other solid phases at lower temperatures) may exist at high temperatures in peridotite (see also Lorand and Luguet 2016, this volume). Residual sulfides with subchondritic 187Os/188Os occur as inclusions in silicates and are inherited from ancient melting processes. These sulfides may represent residual sulfide liquids or mss, or both. Sulfide liquids with chondritic to suprachondritic 187Os/188Os and higher Re/Os and Pd/Ir are precipitated from infiltrating silicate melt and mostly reflect the composition of these melts with variable reaction with peridotite. Hybrid sulfide liquids may form locally where magmas and peridotite react and magmas became oversaturated in sulfur. In addition, relic PGM phases such as Pt–Ir alloys inherited from depleted protoliths may survive these magmatic processes. An important aspect of melt infiltration in the lherzolite stability field is the co-precipitation of sulfides with pyroxene ± Al phase assemblages. Only such a process can explain correlations of Re, Re/Os, and sulfur concentrations with fertility indicators such as Al2O3. As it is likely that the same processes were also responsible for the correlations between 187Os/188Os and Al2O3 in many suites of mantle peridotites, the mass balance with inherited Re-depleted sulfides suggests that the infiltrating melts had suprachondritic 187Os/188Os (the origin of such melts will be discussed later). This notion is supported by Os isotopic measurements on grain boundary sulfides in peridotites and by initial Os isotopic compositions of most mantle pyroxenites (Alard et al. 2002, 2005; Harvey et al. 2010, 2011, 2016, this volume; Wang and Becker 2015c).
Different modeling approaches, both complicated and simple may produce appropriate HSE compositions of basalts from model mantle compositions (e.g., Rehkämper et al. 1999; Bezos et al. 2005; Harvey et al. 2011; Mungall and Brenan 2014). As discussed here and elsewhere (e.g., Lorand et al. 1999; Pearson et al. 2004; Lorand and Alard 2010; Fischer-Gödde et al. 2011; König et al. 2014; Wang and Becker 2015a), models that employ equilibrium distribution of the HSE between mantle phases have difficulties in accounting for some of the detailed compositional variations of the compatible HSE in bulk peridotites. Studies of HSE in bulk rocks of mantle peridotites and pyroxenites and their trace phases indicate that in high temperature magmatic processes in the mantle, disequilibrium between different HSE host phases and silicates may be the rule (e.g., Burton et al. 1999; Alard et al. 2000, 2002, 2005). In spite of these complexities, a useful assessment of the bulk distribution behavior of the HSE is possible and their relative behavior is consistent with abundance data in komatiites and basalts. The data on bulk rocks and sulfides of mantle pyroxenites and sulfides from grain boundaries in peridotite tectonites and in xenoliths indicate that infiltrating melts show relative fractionation of the HSE and S similar to the fractionation pattern of basalts, with mantle normalized abundances of S ≈ Re > Au > Pd > Pt ≥ Rh > Ru > Ir ≥ Os. The HSE data on peridotites and pyroxenites suggest that the composition of infiltrating melts also affects the composition of peridotites (e.g., Figs. 5, 7 20). Notably, enrichments and depletions of Re in peridotites may be caused by precipitation of sulfides with suprachondritic Re/Os. If the abundances of Re, Au, Pd, Pt, and other chalcophile elements in mantle peridotites are predominantly controlled by sulfide segregation from primitive basic magma, the question arises, which partition process produced the relative fractionation among these elements in these magmas to begin with? The answer may lie in the increasing importance of alloy solubility in silicate melt during moderate to high degrees of melting in the shallow mantle, near or beyond the exhaustion of sulfide in the residues. At these conditions, the concentrations of the HSE in silicate melts may be controlled by residual PGE alloys, the different solubility of Pt, Rh, Ru, Ir, and Os and possibly silicate mineral-oxide-melt partitioning (Mungall and Brenan 2014; Brenan et al. 2016, this volume). Thus, basic melt infiltrating the asthenosphere and lithosphere at greater depth likely carries the HSE and 187Os/188Os signature of oceanic crust produced in previous Wilson cycles. This conclusion is consistent with suprachondritic initial 187Os/188Os of mantle pyroxenites and some peridotites that were affected by melt infiltration and coexisting harzburgites with subchondritic 187Os/188Os, which may represent ancient remnants of shallow oceanic mantle.
Os isotopic heterogeneity in the mantle
The compatibility of Os during partial mantle melting, and the existence of two radioactive decay systems producing isotopes of Os, makes it an ideal element with which to investigate mantle heterogeneity (Hart and Ravizza 1996; Burton et al. 1999). The relative compatibility of Os and Re is primarily controlled by their differing preference for sulfide over melt (See section above: Behavior of HSE during partial melting). This produces strong fractionation of moderately incompatible Re from compatible Os during partial melting of the mantle, giving rise to very high Re/Os ratios in crust-forming melts (see Gannoun et al. 2016, this volume) and correspondingly low, sub-chondritic 187Os/188Os ratios in depleted mantle. In turn, crust recycled back into the mantle is potentially traceable due to its distinct Os isotope signature. Likewise, small degree melts within the mantle may also produce variations in Re/Os and thus, over time, in 187Os/188Os. Due to the chalcophile affinity of Os, Re–Os isotope variations can provide different, yet complementary, information to lithophile isotope systems, and can display behavior that is decoupled from lithophiles (e.g., Class et al. 2009).
The 190Pt–186Os decay system, in contrast to the Re–Os system, does not typically produce resolvable differences in 186Os/188Os ratios in mantle rocks due to the much smaller decay constant compared to 187Re, and due to the lower degree of fractionation between parent and daughter. Only in specific cases of high-degree melting do Pt concentrations significantly exceed those of the mantle, such as in some volcanic arc settings (Dale et al. 2012b) and in komatiites (e.g., Puchtel and Humayun 2001; Fiorentini et al. 2011); but in the latter case Os in the melt approaches mantle concentrations and thus fractionation of Pt and Os remains limited. Recycled crust has only moderately high Pt/Os (Dale et al. 2009a; Peucker-Ehrenbrink et al. 2012) which is not sufficient to produce anomalous compositions given the subsidiary Os concentrations of crust, relative to mantle. Nevertheless, 186Os enrichments have been identified in some intraplate magmas (Brandon et al. 1998, 2003; Puchtel et al. 2005) and in a later section we briefly discuss whether mantle processes are a plausible mechanism by which to produce these enrichments.
In this section, we focus on broad-scale mantle heterogeneity, whereas disequilibrium on a hand specimen scale, or smaller, is covered in the previous section Os isotopic disequilibrium.
187Os/188Os mantle composition and heterogeneity
The bulk Os isotope composition of the silicate Earth was likely set by late accretion of material with a bulk primitive composition, after core formation had ceased (Kimura et al. 1974; Chou 1978). However, neither the 187Os/188Os composition (Meisel et al. 2001) nor the relative HSE abundances of PM estimates (Becker et al. 2006) match those of any known chondrite group. This difference has been reconciled by (i) late accretion of differentiated planetesimal core material and primitive chondritic material (Fischer-Gödde and Becker 2012), (ii) by a hybrid model for the enrichment of Earth’s HSE involving late accretion to a fractionated mantle signature (which may be a residue from metal-silicate segregation, cf. Righter et al. 2008; Walker 2009), or (iii) by mantle processes accounting for the combination of non-chondritic ratios involving Ru and Pd and chondritic ratios of other HSE in fertile lherzolites (e.g., Lorand et al. 2010). See Day et al. (2016 this volume) for further discussion.
The processes of continental crust production and incomplete rehomogenization of recycled oceanic crust have likely both served to reduce the 187Os/188Os of the peridotitic mantle below that of the primitive mantle. Thus, heterogeneous distribution of 187Os in the mantle is due to the timing and degree of melt depletion and the presence of enriched domains, which may either be recycled crustal materials or domains within the mantle fertilizes by low-degree melts.
A compilation of 187Os/188Os data for global peridotites (excluding pyroxenites), grouped according to the tectonic settings used in this chapter and in this volume, is shown in Figure 24, and a summary of the averages and ranges for each setting/sample type is shown in Table 2. Cratonic and circum-cratonic xenoliths, which won’t be discussed further here, are both typically strongly unradiogenic, reflecting their severe and early melt depletion and subsequent isolation from the convecting mantle (see Aulbach et al. 2016, this volume, and references therein). All major tectonite and xenolith groups (continental/continent-ocean transitional tectonites, high-T convergent tectonites, ophiolites, abyssal peridotites, oceanic mantle xenoliths, sub-continental lithosphere xenoliths and sub-arc xenoliths) have a considerable ‘peak’ in probability of 187Os/188Os between 0.125 and 0.128, indicating a degree of effective large-scale homogenization in the convecting mantle and younger lithosphere, albeit incomplete. Moreover, most groups have remarkably similar total ranges of 187Os/188Os (when excluding up to 3% of the most extreme data), between 0.026 and 0.029 units, with the exception of high-T convergent margin tectonites (n = 48) which have a range of 0.023, and sub-continental lithospheric mantle xenoliths, with a larger range of 0.037 (although in this latter case the primary data may be compromised by secondary processes such as weathering and reaction with host melts. Greater than 85% of samples from each tectonic setting fall within a narrower range of 187Os/188Os of around 0.015 units (the range of each group varies from 0.013 for all ophiolites, to 0.019 for continental/continent–ocean transition tectonites).
In detail, however, each grouping displays a variable distribution of Os isotope composition, and the positions of the modal and mean 187Os/188Os compositions differ between many of the groupings. One caveat here is that the data plotted on Figure 24 are present-day measured 187Os/188Os ratios, to reflect the current degree of overall mantle heterogeneity, and thus do not account for any isolation of portions of lithosphere sampled in this dataset. If these portions were exposed to gradual convective stirring then some of the ‘older’ depletion ages may have been remixed with more radiogenic ambient mantle. Not all components of the compilation, therefore, necessarily reflect the composition of the ‘convecting’ mantle.
All tectonite groups have ranges that extend to sub-chondritic and supra-chondritic 187Os/188Os ratios, although some extend broadly equally in each sense, while others have a pronounced skew towards less or more radiogenic values. For instance, the ophiolite record has a modal 187Os/188Os of ~ 0.1255, with a broadly equal number of data extending in each sense down and up to values of 0.115 and 0.143, respectively (Fig. 24). At least half of the data fall between 0.1225 and 0.128. In contrast, the dataset for continental/continent-ocean transitional tectonites shows a modal 187Os/188Os of ~ 0.126, close to that of ophiolite ultramafics, but with a range extending down to 0.112 and up to 0.133, with a lower mean value than for ophiolites (Fig. 24). The abyssal peridotite samples of the convecting mantle show a remarkably similar probability profile to the continental/transitional tectonites, with a modal 187Os/188Os of ~ 0.126, and a range from 0.1125 to 0.140; possibly with similar subsidiary peaks at 0.1225 and perhaps even at 0.115 (although this most unradiogenic peak appears important for continental/transitional tectonites, but likely is not significant for abyssal peridotites, given the sample size).
The ‘tails’ to low and high 187Os/188Os reflect, respectively, ancient melt-depleted domains and enriched domains which have not fully re-homogenized with the rest of the convecting mantle through convecting stirring, melt percolation and infiltration. The distribution of the data is further mentioned below in the context of platinum-group mineral studies. Qualitatively, at least, re-enrichment of ophiolitic mantle is supported by the observation that convergent margin ophiolites appear to have more radiogenic 187Os than mid-ocean ridge ophiolites (Fig. 15), and by the absence of a skew to old depleted values in the overall ophiolite 187Os/188Os distribution (Fig. 24; cf. abyssal peridotite and ophiolite curves). The relatively radiogenic distribution of sub-arc xenoliths is also consistent with the process of re-enrichment in the subduction zone environment.
The chromitite and PGM record of Os isotope mantle composition and heterogeneity
Here, we focus only on the Os isotope evidence from PGM, rather than the systematics of PGM formation and composition (see O’Driscoll and Gonzáles-Jiménez 2016, this volume, for a comprehensive review). The utility of chromitites, and the PGM that they typically contain, is that they are Os-rich, Re-poor and tend to be largely robust to subsequent alteration processes caused by metamorphism and/or fluid-rock interaction. The very low Re/Os ratios mean that their 187Os/188Os isotope composition is almost ‘frozen in’ at the point of formation, or at worst require very small corrections for radiogenic ingrowth, even over periods of 3 Ga or greater (Malitch and Merkle 2004). For these reasons, they have been used to estimate the Os composition of the convecting mantle, to assess mantle heterogeneity and to identify potential major mantle melting events through Earth’s history. One caveat to this use, however, is that chromitite formation occurs in zones of high melt flow, and these melts may have imparted a radiogenic 187Os/188Os signature on the chromitite, thus rendering it no longer entirely representative of the ‘average’ upper mantle (e.g., O’Driscoll et al. 2012; see also the section on convergent ophiolites above).
A global suite of ophiolitic chromites was used to provide an estimate of the average 187Os/188Os composition of the convecting mantle (Walker et al. 2002b). Linear regression of the isotope data relative to the age of the chromite provided an evolution curve with a present-day 187Os/188Os composition of 0.1281. Although the uncertainties overlap, this best estimate equates to approximately 5% less ingrowth of 187Os over the life of the Earth when compared to the PM (0.1296; Meisel et al. 2001). This is presumably due to continental crust extraction and the presence of recycled oceanic crust in the mantle, which has not (yet) been efficiently rehomogenized. A study of over 700 detrital PGM from the Josephine Ophiolite, California, found a Gaussian distribution of 187Os/188Os ratios from 0.119 to 0.130 (Meibom et al. 2002). This was interpreted to represent long-term heterogeneity (melt-enriched and -depleted endmembers) which has been partially erased and homogenized by metasomatism and melt–rock reaction processes. Further work on a range of global ophiolites, however, indicated a more complex distribution of Os isotope ratios in Earth’s mantle. Over 1000 detrital PGM from ophiolites in California, Urals, Tibet and Tasmania revealed a variety of 187Os/188Os distributions, from close to Gaussian to skewed towards old, unradiogenic values in the case of Urals, and a bimodal distribution for both Tibet and Tasmania (Pearson et al. 2007). It was proposed that the apparent ‘peaks’ in probability for certain 187Os/188Os ratios are consistent across different ophiolites and across other geological settings such as cratonic xenoliths, and that these peaks reflect global signatures produced by major global mantle melting episodes throughout Earth’s history which match the implied crustal growth record from zircon ages. The composition of the major peak in 187Os/188Os for PGM is 0.1276 (Pearson et al. 2007; adjusted to present-day in Dale et al. 2009b), although the mean composition is likely significantly lower because of the skewed distribution to less radiogenic values. Perhaps notably, when considering representative analyses of convecting mantle composition, this upper limit of 187Os/188Os composition from PGM analysis is less radiogenic than the average of analyzed chromites (0.1281; Walker et al. 2002b), even though many of the PGM are also sourced from supra-subduction zone ophiolites and therefore may be subject to the same process of radiogenic Os addition. Also of note is the fact that ultramafics from most of the tectonic settings have ‘peak’ values that are slightly less radiogenic than the ‘peak’ value from PGM (see Fig. 24; 187Os/188Os ~ 0.1265, compared to 0.1276).
In summary, although global compilations have inherent bias towards exposed and well-studied areas, all the larger datasets (n > 100) for mantle settings that have not been isolated for long periods (cf. cratons), have very similar modal 187Os/188Os compositions of between 0.125 and 0.127, and mean compositions between 0.1243 and 0.1271. Such values equate to around 8 to 18% less ingrowth of 187Os over the life of the Earth than for PM evolution (cf. Meisel et al. 2001), presumably largely due to crustal extraction and long-term isolation – although the exact degree of mantle Re depletion is dependent on the timing of this extraction. These values are somewhat higher than the 5% estimated from chromitites (see above, cf. Walker et al. 2002b), but some of this discrepancy is due to the omission of pyroxenites and other enriched lithologies from this data compilation. The small variance in the isotopic ranges for each setting appears noteworthy in terms of gauging mantle mixing efficiency, but is beyond the scope of this review.
186Os/188Os mantle composition and heterogeneity
Platinum-group minerals and chromitites have been used as recorders of the 186Os/188Os evolution of the mantle. Many PGM are IPGE-rich and have low Pt/Os and hence faithfully record the 186Os/188Os of the mantle at the time when those PGM formed. Brandon et al. (2006) used Os-rich PGM data, together with chondrite analyses, to constrain the terrestrial evolution of 186Os/188Os from an initial of ~ 0.1198269 ± 0.0000014 (2 σ) at 4.567 Ga to a present-day value of 0.1198382 ± 0.0000028.
The potential for large-scale heterogeneity generated by the 190Pt-186Os system is far smaller than that of the 187Re-187Os system, and in most cases is beyond what is distinguishable given current analytical capabilities. Nevertheless, anomalously radiogenic 186Os/188Os ratios have been found in some high-degree melts in intraplate settings in Hawaii, Gorgona Island and Kostomuksha, Russia (Brandon et al. 1998, 2003; Puchtel et al. 2005), coupled with only limited 187Os enrichment. Possible mechanisms to generate such signatures are discussed below.
The range of Pt/Os ratios found in the supra-subduction zone environment indicates that there must be huge 186Os variations on a lithological and mineral scale, if those materials were isolated. Alaskan-Uralian complexes (see Johan 2002 for details ) also display a large range of Pt/Os ratios, but these are beyond the scope of this chapter. Chromitites from ophiolites typically possess very low Pt/Os ratios (~ 0.1, compared with 1.95 for the PM), but can sometimes have Pt/Os of > 10 (see ophiolite sections). Platinum-group minerals from within chromitites and other PGE-saturated ores can have even more extreme Pt/Os; laurites (Ru (Os, Ir)S2), may have ratios of < 0.01 (González-Jiménez et al. 2009) while Pt–Fe alloys can have Pt/Os of > 100,000 (Walker et al. 1997). Extremely high Pt/Os ratios, such as those of the Meratus Ophiolite, Borneo (up to 2000), evolve to much higher 186Os/188Os compositions than those of the bulk mantle, and because PGM are largely robust to subsequent processes, they may show isochronous behavior and can be used to date ophiolitic complexes (Coggon et al. 2011). These PGM, after ingrowth over as little as 200 Ma, have 186Os/188Os ratios that range from a slightly sub-PM value of 0.119801–0.120315. As a guide to the magnitude of this difference, it is at least 30 times greater than the difference between the bulk mantle and the highest 186Os/188Os mantle melt yet discovered (0.000015; Brandon et al. 1999). These data will be discussed further in the subsequent section on the production of HSE-Os signatures in mantle melts.
A recent study of Eoarchaean chromitites from south-west Greenland found 186Os/188Os data proposed to reflect mantle melt depletion events in Earth’s earliest history, during the Hadean at approximately 4.1 Ga and possibly as old as 4.36 Ga (Coggon et al. 2013). In so doing, Coggon et al. (2013) also inferred that the late veneer must have occurred prior to this time, consistent with the message of an ‘early’ late veneer from studies of basaltic meteorites from different parent bodies (Dale et al. 2012a).
The role of recycled oceanic lithosphere in producing HSE and Os isotope signatures in magmas
At least part of the compositional variability observed in mantle melts at Earth’s surface is derived from heterogeneity in the mantle. The biggest single process by which such heterogeneous chemistry is generated must be that of recycling of oceanic lithosphere through subduction (e.g., Hofmann and White 1982). In addition, there are other processes, such as melt percolation within the mantle and lithosphere (e.g., Halliday et al. 1995) that potentially play an important role in producing the variety of magma compositions that we observe at Earth’s surface. Many instances of melt percolation may ultimately be sourced from enriched recycled material, but this is not a requirement in producing variations in fertility in the mantle. Here, we focus on the composition of recycled ultramafic and mafic lithosphere within the mantle, and its impact within the source regions of oceanic magmas.
Prior to subduction, the oceanic lithosphere gains variable amounts of water and trace elements during seawater interaction or hydrothermal alteration, resulting in the formation of serpentine minerals, at the expense of olivine. This alteration may, in more extreme cases, be accompanied by elevated 187Os/188Os and the loss of Os relative to the other IPGE (see abyssal peridotite section), but typically, abyssal peridotites retain mantle-like HSE proportions and 187Os/188Os ratios. Regardless of the precise HSE signature, serpentinization permits water transport deep into subduction zones and beyond into the deep mantle. Together with the hydrous mafic crust, this provides fluxes of fluids from the downgoing slab into the mantle wedge at a range of depths, as well as retention of water beyond the supra-subduction setting. The potential for the slab to transport water beyond the zone of sub-arc melting is likely to be important for promoting small-degree hydrous melting in the mantle, which may have an impact on HSE through refertilization processes.
The impact of subduction zone processes on HSE in convergent margin magmas and recycled oceanic lithosphere
Fluxes into the mantle wedge produce two effects which have a bearing on HSE behavior and Os isotope composition. First, as discussed above, radiogenic Os may, in certain cases (Brandon et al. 1996; Becker et al. 2004), be transferred from the slab into the mantle wedge and then transferred by melts into arc crust and supra-subduction oceanic crust, sampled by ophiolites. Second, fluid addition will promote hydrous melting, allowing otherwise refractory mantle domains to partially melt and permitting melting of the mantle at temperatures below those of the normal geothermal regime.
The evidence for a radiogenic Os flux to arc magma sources is equivocal, due to the difficulty in knowing the precursor 187Os/188Os of the mantle source and other potential sources of radiogenic Os such as arc crust. Nevertheless, the ophiolite record provides a firmer basis for this contention. An additional HSE flux is the loss of Re from metabasic rocks during dehydration (~ 50–60%; Becker 2000; Dale et al. 2007), and likely enrichment of Re in the mantle wedge (Sun et al. 2003a,b). This flux could contribute, over time, to radiogenic 187Os in the mantle wedge and also has implications for the composition of recycled crust which are outlined below. Other HSE may also be mobilized (McInnes et al. 1999; Kepezhinskas et al. 2002; Dale et al. 2009a), but whether the magnitude of flux is sufficient to produce a measurable effect in supra-subduction zone magmas is doubtful, given the relatively high concentrations of these elements in the mantle.
Melting of refractory domains increases the likelihood of sulfide exhaustion, which, under most circumstances, would reduce the compatibility of all HSE, resulting in less fractionated HSE patterns such as those seen in picrites and komatiites (e.g., Puchtel and Humayun 2000). In the Tonga Arc, however, the relative proportions of the HSE are amongst the most fractionated for mantle melts (Dale et al. 2012b), with extreme Pt/Os approaching 15. This fractionation may be caused by increased HSE-rich phase stability during lower temperature hydrous melting (e.g., laurite stable up to 1275 °C; Brenan and Andrews 2001) and/or the promotion of chromitite formation by interaction between hydrous melts and refractory mantle (Dale et al. 2012b). Chromitite formation during melt–rock reaction in the mantle is expected to fractionate HSE significantly, sequestering IPGE in PGM and producing a melt with high (Re + Au + PPGE)/IPGE (see Ophiolite sections).
The role of recycled lithosphere in producing HSE-Os signatures in convecting mantle melts
Many previous attempts have been made to model the effects of recycling oceanic lithosphere, particularly the mafic crustal portion (e.g., Roy-Barman et al. 1996; Brandon et al. 1999, 2007; Becker 2000; Dale et al. 2009b; Day et al. 2009). While we recognize the importance of quantitatively assessing whether a particular process is possible or likely, given the numerous previous attempts and the dependency on the parameters chosen, here we direct the reader to those previous studies and we instead choose to focus on the record of pyroxenites in the mantle, as direct recorders of enriched, hybridized lithologies. Of course, it is important to bear in mind that the sampled pyroxenite database is still relatively small (62 samples with HSE and/or Os isotope data collated in Fig. 25) and thus it is difficult to relate this to the mantle as a whole. That said, the processes identified are broadly applicable.
Both eclogitic and pyroxenitic enriched lithologies are present in the mantle. Eclogites represent unequivocal crustal materials, sampled as xenoliths in intraplate volcanic settings, which retain much of their crustal geochemical signature, albeit modified by subduction processing. The term ‘pyroxenite’ covers a complex array of lithologies and petrogenetic pathways that are beyond the scope of this chapter (see Lambart et al. 2013). In simple terms, pyroxenites are variably hybridized lithologies produced during reaction of peridotite with silica-saturated melt derived from an enriched lithology such as eclogite (or possibly also derived from small-degree melting of peridotite). Reaction with a silica-undersaturated, olivine-saturated melt would instead produce dunite, so depending on the exact mode of formation of particular dunites (some dunites might be cumulates), they may also carry an enriched Os signature, as seen in the ‘convergent margin ophiolite’ section. Unlike eclogites, pyroxenites form a significant part of mantle tectonites, constituting between 1 and 9% of the Beni Bousera mantle tectonite massif (Pearson and Nowell 2004). These pyroxenites at Beni Bousera have been identified as having a recycled crust origin, on the basis of lithophile and stable isotopes. They typically have radiogenic 187Os/188Os ratios, even in samples that are Os-rich (> 2 ng/g).
Pyroxenites and peridotites from the Totalp ultramafic massif, Swiss Alps, preserve a record of refertilization of peridotites by both melt percolation from the pyroxenites and from mechanical stretching and thinning of websterite layers (van Acken et al. 2008). The pyroxenites are strongly enriched in 187Os (187Os/188Os: 0.122–0.866; main range: 0.13–0.16) and in Re, whereas peridotites have a broadly chondritic average γOs value. It is noted, therefore, that refertilization does not completely homogenize Os isotopes, at least not on a small scale, but isotopic differences are reduced due to reaction of pyroxenite melt with peridotite.
A compilation of ultramafic mantle samples, in terms of Pt/Os and Re/Os ratios, is presented in Figure 25. Pyroxenites form a distinct group at elevated Re/Os and Pt/Os ratio, relative to peridotites. The degree of this enrichment of Re is, in itself, consistent with a partially pyroxenite source for some mantle melts with radiogenic Os over a period of ingrowth of 1 Ga or more. Actual measured 187Os/188Os for global pyroxenites, excluding the 10 highest and lowest values from a total of 94 samples, varies from 0.124 to 0.928. Obviously the ability for these pyroxenites to produce sufficiently radiogenic melts as part of a hybrid pyroxenite–peridotite mantle, depends on their Os contents. The Os concentrations also vary substantially, from 0.005–4.6 ng/g, and this generally co-varies negatively with 187Os/188Os ratios. Thus, at some level, the effect of the pyroxenite in the mantle is self-limiting due to reduced Os content. As well as the strongly radiogenic signatures of the pyroxenites themselves, there is also evidence for radiogenic Os addition to peridotitic rocks (Becker et al. 2001; Büchl et al. 2002; van Acken et al. 2008; Marchesi et al. 2014), and this, combined with the radiogenic pyroxenites, will more easily produce radiogenic mantle melts.
One aspect of oceanic crust recycling that has commonly been overlooked is the geochemical distinction between the gabbroic and basaltic parts of the crust. This is now generally fully recognized for HSE – with gabbroic crust being, on average, significantly more Os- and Pt-rich and slightly poorer in Re than MORB—and this has been incorporated into models for crustal recycling (Peucker-Ehrenbrink and Jahn 2001; Dale et al. 2007; Peucker-Ehrenbrink et al. 2012).
An alternative, but related, means by which recycled lithosphere may have an impact on the HSE composition of mantle melts is through the process of sulfide metasomatism. Sulfides with radiogenic 187Os have been sampled in interstitial locations within peridotites (Alard et al. 2005; Harvey et al. 2006, 2010, 2011; Warren and Shirey 2012). The ultimate source of those sulfides is unknown, but derivation from recycled crustal material, of at least some such sulfides, is plausible. Radiogenic, interstitial sulfides can then be readily incorporated into partial melts, whereas unradiogenic residual sulfides remain shielded from melt by the silicates that enclose them. The process of sulfide addition is a similar process to other forms of refertilization, but in this case the lithophile and chalcophile element signatures may be decoupled. However, the overall broad coupling of 187Os/188Os with Al2O3 contents may suggest that this process is typically not large-scale and pervasive (cf. Fig. 2).
186Os–187Os coupled enrichments
Over time, Pt/Os ratios greater than that of the primitive mantle (PM) will develop elevated 186Os/188Os ratios. A Pt/Os ratio of approximately greater than 8 is required, over a 1.5 Ga period, to produce the most 186Os-enriched mantle melt identified to date. Of the current mantle database for peridotites, dunites, and some chromitites, approximately 11% have Pt/Os ratios greater than 4, while only 4% have ratios greater than 8 (Fig. 25). Enriched pyroxenite lithologies, however, commonly have sufficiently high Pt/Os ratios; ~ 55% of the 62 pyroxenites compiled in Figure 25 have Pt/Os >8. However, many rocks with elevated Pt/Os also possess elevated Re/Os which evolves to much higher 187Os/188Os ratios than observed in intraplate magmas with enriched 186Os. Therefore, rocks with Pt/Os, Re/Os and Pt/Re all greater than the PM are of particular interest for the generation of coupled enrichments of 186Os and 187Os, but such rocks are a very minor proportion of the current mantle database (Fig. 25).
This difficulty in generating radiogenic 186Os, without also producing enrichments in 187Os beyond those observed, led Brandon et al. (1998), after Walker et al. (1995), to propose a role for transfer of an outer core Os signature into the plume source of some high-degree melts in intraplate settings. Twenty years later, this remains a possible scenario, despite the alternative mechanisms proposed that are outlined here. The core-mantle interaction model does, however, require an early onset of inner core solidification (by 2.5 Ga, and earlier for 2.8 Ga Kostomuksha komatiites; Puchtel et al. 2005) in order to allow sufficient time for ingrowth to produce enrichments in 186Os and 187Os in the predicted high (Pt–Re)/Os outer core. A more complete discussion of the core-mantle interaction debate can be found in Brandon and Walker (2005) and Lassiter (2006).
Since the emergence of the core-mantle interaction theory, several other possible sources of radiogenic 186Os have been proposed (e.g., Baker and Jensen 2004; Luguet et al. 2008b), though no proposed mechanism is completely convincing. The modification of pyroxenites, refertilization of peridotites and accompanying sulfide removal and/or metasomatism is the most likely alternative to core-mantle interaction (Luguet et al. 2008; Marchesi et al. 2014), but suitable Pt/Os and Re/Os ratios in the current mantle database are the exception, rather than the rule (Fig. 25). One further, more complex, possibility is that signatures may be combined from separate mantle components each with either high Pt/Os or high Re/Os, but not both. As outlined in a previous section, extreme Pt/Os fractionation exists on a variety of scales in Earth’s mantle, particularly during the formation of PGM. What is not yet clear is the fate of such PGM during mantle convection and whether there is sufficient separation and sampling of particular PGM compositions to produce specific signatures in mantle melts.
In summary, processes exist in Earth’s mantle that can account for the 186Os–187Os enrichments observed in intraplate magmas, but currently they appear to be rare.
The relationship between abyssal peridotites and MORB: an osmium isotope perspective
One major debate in the field of HSE chemistry, and a key issue for mantle geology as a whole, is the extent to which abyssal peridotites represent the mantle residues of partial melting at oceanic spreading centres. Osmium isotopes have been a key part of this debate, but the evidence is complex. Early analyses identified a large range of 187Os/188Os compositions in abyssal peridotites, ranging from sub-chondritic to significantly supra-chondritic (see abyssal peridotite section above). The elevated signatures were largely attributed to seawater interaction. After taking into account this process, the remaining abyssal peridotite data appeared to be far less radiogenic than data for mid-ocean ridge basalts, thus casting doubt on a genetic link between abyssal peridotites and MORB. Since that time, two important findings have been made which reduce this discrepancy.
Firstly, was the discovery of interstitial sulfides of magmatic origin possessing radiogenic, supra-chondritic 187Os/188Os ratios (Alard et al. 2005), together with non-chondritic PGE ratios (Alard et al. 2000). A preferential contribution from these interstitial sulfides to a partial melt, compared with the contribution from ancient, unradiogenic sulfides enclosed within silicates, could account for the more radiogenic signatures of MORB and other partial melts of the oceanic mantle, compared with those recorded in bulk-rock abyssal peridotites.
Secondly, but of at least equal importance, was the finding that the Os isotope compositions of MORB (see Gannoun et al. 2016, this volume) were less radiogenic than previous thought. In particular, the range of 187Os/188Os ratios in MORB glasses was found to be considerably less (0.126–0.148) than previous findings (e.g., Schiano et al. 1997), with a lower mean of 0.133 ± 0.009, in part due to an analytical artefact in the original data (Gannoun et al. 2007). This mean value, while reduced, remains in excess of typical values for abyssal peridotites (187Os/188Os: 0.118–0.130). However, it was also found that the constituent phases of basalts had variable 187Os/188Os due to (i) ingrowth over poorly-constrained periods of time since emplacement (Gannoun et al. 2004), and (ii) the timing of crystallization of different phases with respect to the evolution of the melt and its interaction with seawater-modified crust (Gannoun et al. 2007). Most notably, the latter manifests itself in significantly less radiogenic Os isotope compositions in early-formed relatively Os-rich sulfides compared with their (Os-poor) host glasses. In some cases there is a difference of ~ 0.015 in the 187Os/188Os of glasses and corresponding sulfides (e.g., glasses: 0.1383 and 0.1479; sulfides: 0.1249 and 0.1308, respectively), with the sulfides falling in the range 0.1236–0.1310, largely equivalent to the range seen in abyssal peridotites. Moreover, a negative covariation of 187Os/188Os and Os content in MORB sulfides might indicate that MORB sulfides are also affected by interaction with a radiogenic contaminant, casting doubt on the significance of more radiogenic data for Os-poor sulfides.
Although sulfides included within silicates in abyssal peridotites (and other mantle tectonites) are known to possess even lower 187Os/188Os than bulk-rock samples (~ 0.114; Harvey et al. 2006)—and are therefore also lower than estimates of primitive MORB—such shielded sulfides likely contribute little to moderate degree partial melts relevant for MORB genesis. Therefore, in conclusion, not only has the ‘gap’ in composition between abyssal peridotites and MORB been largely bridged by radiogenic interstitial sulfides, but it seems likely that the gap is minor or non-existent when the most primitive parts of the MORB system are analyzed.
Interpretation of Re–Os model ages
Model ages, whereby the isotope ratio of a sample is compared to the evolution of a reference frame such as average chondrite compositions, have been extensively used in geochemistry to give melt depletion ages in systems where recent mobility of elements has obscured any isochronous isotope systematics. The Re–Os system has been of particular use in this regard, due to the contrasting behavior of Re and Os which can result in, for high degree melts, effective Re removal from the source, while Os remains present in high enough abundances (several ng/g) to provide a degree of robustness against alteration and contamination. For Os, the measured 187Os/188Os ratio of a sample (or, for xenoliths, the ratio calculated at the time of the host eruption) is compared to the evolution curve of the mantle (commonly either a chondrite reference or the primitive mantle estimate). For Re depletion ages (TRD) it is assumed that the residue is completely depleted in Re after partial melting and, thus, there is no further ingrowth of 187Os. The advantage of this method is that it provides a relatively robust guide to the long-term evolution of the sample, due to the generally conservative behavior of Os, without the difficulties induced by recent Re addition or loss. In reality, however, only in high degree melting events is complete Re removal attained and in many cases the TRD age merely provides a minimum age. An alternative type of model age uses the measured Re/Os ratio to calculate the time when the 187Os/188Os of the sample intersected that of the reference frame (TMA or TRe–Os). In theory, this can provide a more accurate age, but it suffers the same sensitivity to Re mobility as do attempts to identify Re–Os isochron relationships.
Numerous caveats and potential pitfalls of model age determinations have now been recognized and the reliability and interpretation of Re–Os model ages in peridotites was the subject of a comprehensive review by Rudnick and Walker (2009). Here, we summarize the main issues surrounding such model ages, in the context of the processes and tectonic settings discussed in this chapter.
Perhaps the most obvious issue encountered has already been mentioned above—that of the degree of depletion of Re. Rudnick and Walker (2009) demonstrate that mantle melting at 3.5 Ga to form a basaltic melt would result in vastly different age estimates from TMA and TRD methods: the TMA age for the residue would be 3.5 Ga, because the Re/Os ratio of the residue is used to back-calculate the isotope evolution of the sample, whereas the assumption of complete Re depletion in the case of a TRD age would produce an age of just over 1 Ga. Clearly at this level of depletion TRD ages are not useful and they only become more valuable when Re removal is close to complete (probably a boninitic or komatiitic melt depletion event).
Alternatives to isochron ages and Re depletion ages have been used to gain age information for sample suites where, respectively, Re mobility is suspected or Re removal was not complete. An element of similar compatibility to Re, but less mobile, such as Al2O3, can be used as a proxy for Re on an isochron diagram (Reisberg and Lorand 1995; see earlier). Although there is sometimes much scatter on such plots, they appear to be broadly robust. For large datasets of > 50 samples, but preferably more, probability density function plots provide a means to identify common apparent depletion ages, which lends weight to an argument for those ages having age significance. For instance, a range of 187Os/188Os ratios could be produced by variable degrees of depletion or by the same degree of depletion at different times. The identification of peaks on probability plots might indicate discrete times of melt depletion (perhaps partially obscured by variable depletion, preservation issues and/or inheritance) rather than a more continuous spectrum of compositions which might be expected from a suite of variably depleted samples.
There is significant inherent uncertainty with any TRD age, because they are based on a model evolution curve. There are two aspects to this issue: (i) it is known that Earth’s mantle has broadly chondritic proportions of the HSE, but it is not known which chondrite group – if indeed any in the global collection – supplied Earth’s HSE or whether there was any fractionation of HSE during core formation. Models to account for the apparently supra-chondritic Ru/Ir and Pd/Ir ratios of the PM (Becker et al. 2006; Walker 2009; Fischer-Gödde et al. 2011) may also have implications for the Re–Os isotope evolution of the PM. The choice of type of chondrite or PM estimate to use for the model evolution can result in an age variation of nearly 200 Ma for a 187Os/188Os of ~ 0.124, decreasing with increased age to an uncertainty of ~ 100 Ma at around 2 Ga (187Os/188Os = 0.114). (ii) As with lithophile isotope systems (e.g., Sm–Nd) a choice has to be made whether to use a primitive or depleted mantle reference frame. This can make an even more significant difference to the age given that the estimated 187Os/188Os of the primitive mantle is 0.1296, whereas an ‘average’ depleted mantle composition might be somewhere between 0.1245 and 0.128, depending on whether the average for abyssal peridotites or a combination of chromitites, PGM and high-degree mantle melts is used (Walker et al. 2002b; Pearson et al. 2007; Dale et al. 2009b). This also illustrates the problem of inheritance, which relates to the large degree of Os isotope heterogeneity observed in the convecting mantle and is amongst the most important considerations. This effectively means that for small datasets without additional information there is little way of knowing whether an apparent old age reflects a significant ancient melt depletion event in the context of its tectonic setting, or whether the measured 187Os/188Os is a composite of that event superimposed on an already depleted (or enriched) Os signature. For this reason, larger datasets obviously produce more robust age estimates and plots displaying probability can be used to identify ‘significant’ common ages or ‘peaks’ (Pearson et al. 2007; Rudge 2008).
So far, we have made no mention of potential petrological pitfalls for model ages. These encompass serpentinization, sulfide breakdown, refertilization and melt–rock reaction (Rudnick and Walker 2009). Serpentinization, as discussed in an earlier section, does not typically affect Os isotope systematics except in extreme cases, which can easily be avoided when selecting samples with which to gain age information. Sulfide breakdown is known to occur in mantle xenoliths, due to interaction with the host melt. This commonly results in Os loss which could potentially impact upon the model age if 187Os/188Os is variable between different host phases, and which also leaves the sample more susceptible to contamination and alteration.
Depending on the tectonic setting, some processes may or may not impact on model ages. For instance, melt–rock reaction in the convecting mantle is commonly associated with melting, and is therefore effectively zero age with respect to melting and should not normally affect the model age recorded for that melting event. Such melt–rock reaction also usually produces discordant samples on an 187Os/188Os–Al2O3 diagram, and can thus be identified and avoided for the purposes of dating. Conversely, processes of melt percolation and reaction in the continental lithosphere may occur long after the melt depletion episode of interest and this has the potential to obscure the true age (Rudnick and Walker 2009). These issues mean that samples with the lowest 187Os/188Os give the most reliable ages, but they too may still have experienced radiogenic Os input. The extent to which this process affects ages depends on the amount of addition of sulfide, and the Os isotope composition and concentrations of those sulfides. Such sulfides are typically poorer in Os than enclosed sulfides so significant additions of sulfide may be required to significantly affect the age.
Although the processes of metasomatism and refertilization can have a significant effect on model ages, sometimes leading to recent TRD ages or “future” TMA ages, in some cases these processes can be traced using HSE behavior. For example, it has been recognized, in the cratonic setting, that the oldest TRD ages for a suite of samples are associated with the lowest Pd/Ir ratios, reflecting the most pristine and severe melt depletion signatures (Pearson et al. 2004). Recently, the Se/Te ratio has also been combined with Pd/Ir, in order to further understand the effects of metasomatic sulfide addition on model ages and place limits on the levels of addition that can occur before the model age may no longer be reliable (Luguet et al. 2015).
In summary, there are numerous potential pitfalls and limitations for Re–Os model age determinations but, in the absence of isochron dating, the system remains amongst the most useful for providing the ages of melt depletion of the mantle.
We thank Chris Ballhaus, Al Brandon, James Brenan, Kevin Burton, Rick Carlson, James Day, Mario Fischer-Gödde, Mouhcine Gannoun, Timo Gawronski, Jason Harvey, Akira Ishikawa, Yogita Kadlag, John Lassiter, Ambre Luguet, Jean-Pierre Lorand, Claudio Marchesi, Graham Pearson, Igor Puchtel, Dave Rubie, Steve Shirey, David van Acken, Richard Walker and Zaicong Wang for valuable insight and discussions over the years. Thanks to Jason Harvey, Chuan-Zhou Liu, Wendy Nelson and Jessica Warren for helpful reviews of the manuscript.