- © 2016 Mineralogical Society of America
Terrestrial magmatism is dominated by basaltic compositions. This definition encompasses mid-ocean ridge basalts (MORB), which account for more than eighty percent of Earth’s volcanic products and which are formed at divergent oceanic plate margins, as well as intraplate volcanic rocks such as ocean island basalts (OIB), continental flood basalts (CFB) and continental rift-related basalts, and highly magnesian ultramafic volcanic rocks that dominantly occur in Archean terranes, termed komatiites. All of these broadly basaltic rocks are considered to form by partial melting of the upper mantle, followed by extraction from their source regions and emplacement at the Earth’s surface. For these reasons, basalts can be used to examine the nature and extent of partial melting in the mantle, the compositions of mantle sources, and the interactions between the crust and mantle. Because much of Earth’s mantle is inaccessible, basalts offer some of the best ‘proxies’ for examining mantle composition, mantle convection and crust–mantle interactions. By contrast, at arcs, volcanism is dominated by andesitic rock compositions. While some arcs do have basaltic and picritic magmatism, these magma types are rare in convergent plate margin settings and reflect the complex fractional crystallization and often associated concomitant assimilation processes occurring in arcs. Despite the limited occurrence of high MgO magmas in arc volcanic rocks, magmas from this tectonic setting are also important for elucidating the behavior of the HSE from creation of basaltic compositions at mid-ocean ridges to the subduction of this crust beneath arcs at convergent plate margins.
The highly siderophile elements (HSE; comprising Re and Au, along with the six platinum-group elements [PGE] Os, Ir, Ru, Rh, Pt, and Pd) combined with the 187Re–188Os and 190Pt–186Os systems that are embedded within these elements, have found significant utility in the study of basaltic rocks (e.g., Shirey and Walker 1998; Carlson 2005; Day 2013). The greatest strengths of the HSE lies in the fact that they strongly partition into metal or sulfide phases, and so record evidence for processes that are not revealed from other isotope systems commonly used in high-temperature geochemical studies (e.g., He–O-Sr–Nd–Hf–Pb). Partial melting over much of Earth’s geological history has resulted in significant fractionation of the HSE between the mantle and the crust (oceanic and continental). The HSE show contrasting behavior during melting, with the platinum-PGE (PPGE; Pt, Pd), Re, and Au usually behaving as moderately compatible to moderately incompatible elements during melting and crystallization, and the iridium-PGE (IPGE; Os, Ir, and Ru) acting as highly compatible elements (Barnes et al. 1985). The differential response of the HSE to partial melting is demonstrated by differences in both the absolute and relative abundances of the HSE in mantle-derived melts and in residual mantle rocks themselves. High degree melts, such as komatiites (e.g., Puchtel et al. 2009) show a smaller enrichment of PPGE over IPGE than relatively lower degree melts, such as MORB (e.g., Rehkämper et al. 1999; Bezos et al. 2005) (Fig. 1a). Mantle peridotites often show a complementary depletion of PPGE relative to the IPGE that reflects the degree of melt depletion (Fig. 1b), consistent with preferential removal of Re > Au > Pd > Pt > Rh > Ir ≥ Ru ≥ Os (Pearson et al. 2004; Becker et al. 2006; Fischer-Gödde et al. 2011). In the broadest sense, these observations suggest that the HSE in mantle and mantle-derived melts are controlled by both: (i) the degree of melting and; (ii) the mineralogy of mantle rocks. The IPGE are preferentially retained in mantle rocks at low degrees of melting, consequently, low-degree melts such as MORB have relatively low IPGE abundances.
Furthermore, because Pt is moderately compatible, Re is moderately incompatible and Os is highly compatible during melt generation, the Re–Os and Pt–Os isotope systems differ significantly from other geologically useful long-lived radiometric systems (e.g., Rb–Sr, Sm–Nd, Lu–Hf, U–Th–Pb), where both the parent and the daughter elements are preferentially concentrated into the melt. In this chapter, we review the distribution of the HSE amongst mantle minerals and their behavior during melting, the HSE abundances and Os isotope compositions preserved at mid-oceanic ridge settings (divergent plate boundaries), intraplate settings, and of magmas formed at arcs (convergent plate boundaries), to examine the behavior of these elements in a range of tectonic settings.
HIGHLY SIDEROPHILE ELEMENT DISTRIBUTION AND BEHAVIOR IN THE UPPER MANTLE
Core formation and the late accretion of impactor material
The HSE have high affinity for both Fe-metal and sulfide over coexisting silicate minerals or silicate melt. Low-pressure metal–silicate partition coefficients determined experimentally are extremely high (between 104 and 1015) (Kimura et al. 1974; Jones and Drake 1986; Peach et al. 1990, 1994; Fleet et al. 1991, 1996; Borisov et al. 1994; O’Neill et al. 1995; Holzheid et al. 2000; Ertel et al. 2001; Fortenfant et al. 2003; Yokoyama et al. 2009; Mann et al. 2012; Brenan et al. 2016, this volume). Consequently, these elements should have been substantially partitioned into Earth’s metallic core, leaving the silicate mantle effectively stripped of the HSE. Yet, HSE concentrations in Earth’s upper mantle are much greater than predicted from low-pressure experimental data (see Day et al. 2016, this volume). Moreover, their relative abundances display a broadly chondritic pattern, rather than reflecting differences in their respective metal-silicate partition coefficients (Fig. 2). However, the siderophile behavior of some of the HSE may be greatly reduced at high P–T conditions, and on this basis it has been suggested that high-pressure equilibration at the base of a deep molten silicate layer or ‘magma ocean’ on the early Earth, may account for their abundances in the upper mantle (Murthy 1991). High-pressure experiments that simulate the conditions of core formation do indeed indicate that the HSE are less siderophile under these conditions (e.g., Mann et al. 2012). However, the range of HSE partition coefficients, even at elevated P–T conditions, cannot account for either the absolute or relative abundances in the terrestrial mantle, suggesting that high-pressure equilibration was not the dominant process controlling their present distribution. Therefore, mantle HSE abundances have long been taken to suggest that between 0.5% and 0.8% by mass of ‘late accreted’ broadly chondritic material was added to Earth after core formation was complete (e.g., Kimura et al. 1974; Chou 1978). Differing absolute abundances, but similar chondrite-relative HSE abundances have also been inferred for the Moon, Mars and other meteorite parent-bodies (Day et al. 2007, 2010a, 2012, 2016 this volume; Brandon et al. 2012; Dale et al. 2012a; Riches et al. 2012; Day and Walker 2015), suggesting that late accretion was a common phenomenon to terrestrial planets, setting the HSE abundances in planetary mantles. In this way, core formation and late addition of meteorite material are thought to have established the HSE abundance in Earth’s silicate mantle, providing a framework for understanding the long-term effects of mantle melting.
Highly siderophile elements in mantle minerals
The behavior of the HSE during partial melting of the mantle is controlled by their distribution amongst sulfides, platinum group metal alloys (PGM) and coexisting silicates and oxides in mantle rocks (see also Harvey et al. 2016, this volume; Lorand and Luguet 2016, this volume; O’Driscoll and González-Jiménez 2016, this volume).
In addition to their strongly siderophile (iron-loving) behavior, the HSE are also known to be highly chacophile (sulfur-loving), with sulfide in mantle rocks exerting a dominant control over the behavior of the HSE (e.g., Mitchell and Keys 1981), despite the extremely low abundance of these minerals (the proportion of sulfide in mantle rocks is thought to be in the range 0.0014–0.008%, Luguet et al. 2003). The exact magnitude of partitioning of the HSE between sulfide and silicate, however, remains poorly constrained, with values ranging from 1000 to > 106 (Fig. 3) (Peach et al. 1990, 1994; Fleet et al. 1996; Crocket et al. 1997; Andrews and Brenan 2002a; Gannoun et al. 2004, 2007; Fonseca et al. 2009; Mungall and Brenan 2014). At least some of this variation is likely to relate to compositional variations of sulfide and silicate, or the conditions under which equilibration occurred. Values at the low end of the range are usually found in natural occurrences of glass and sulfide (e.g., Gannoun et al. 2004, 2007), while the highest values are indirect estimates based on alloy–sulfide and alloy–silicate partitioning (e.g., Fonseca et al. 2009). A particular problem with the “indirect” estimates of alloy–silicate partitioning (Fonseca et al. 2009) is that they were determined for Fe and S-free compositions, precluding the possible formation of metal-sulfide complexes (e.g., Gaetani and Grove 1997). Moreover, the solubility of at least some of the HSE is enhanced in sulfur-bearing experiments, relative to sulfur-free experiments (Laurenz et al. 2013), bringing partition coefficients into the range of other experimental estimates (Andrews and Brenan 2002a; Mungall and Brenan 2014). While the differences in partition coefficients that remain still span up to three orders of magnitude, estimates based on individual experiments or natural coexisting sulfide–silicate show significantly less variation. These data indicate that the PGE (Os, Ir, Ru, Pt, and Pd) partition similarly into sulfide, with only Re showing a significant difference to the other HSE.
During mantle melting, sulfide will be removed in the silicate melt, as a function of temperature, fO2, pressure, and the iron content of the melt (Wallace and Carmichael 1992; Mavrogenes and O’Neill 1999; O’Neill and Mavrogenes 2002). Given the estimated sulfur content of both the primitive mantle (~ 250 μg g−1 S; Lorand 1990; O’Neill 1991; Palme and O’Neill 2003) and the depleted mantle (~ 120–150 μg g−1 S; Salters and Stracke 2004), and the relatively low degrees of partial melting required to produce most basalts, it is likely that they leave their source sulfide saturated (that is, sulfide remains as a stable mantle mineral). For example, the low HSE content of some low-degree alkali basalt partial melts can be explained by the presence of residual sulfide in the mantle source, while the high-HSE content of high-degree mantle melts, such as komatiites, can be explained by exhaustion of sulfide in the source. However, sulfide behavior alone cannot account for the systematic depletion of HSE seen in mantle rocks, or the variable HSE content and high Re abundances seen in MORB.
Silicate and oxides
Rhenium not only partitions into sulfide, but also into other mantle phases including clinopyroxene, orthopyroxene, garnet, and spinel (Hart and Ravizza 1996; Righter and Hauri 1998; Burton et al. 1999, 2000, 2002; Mallman and O’Neill 2007), particularly under reducing conditions (Mallman and O’Neill 2007). The relatively low partition coefficients for Re between silicate phases and melt, and the much lower coefficient for its partitioning between sulfide and silicate melt compared to other HSE, makes this element moderately incompatible during terrestrial partial melting (Fig. 3). The partitioning of Re into silicate then raises the question of to what degree the HSE may also be incorporated into silicates or oxides in mantle rocks. Overall, natural and experimental data suggest that silicate or oxide phases in the mantle do not exert a strong control on the behavior of HSE during partial melting. Taking estimates of the proportion of silicate phases present in the upper mantle (e.g., Workman and Hart 2005), partial melting of a sulfide-free mantle would yield melts that are slightly depleted in Os, Ir, and Ru, relative to their source. Such a pattern is consistent with that seen for high-degree melts, such as komatiites. Nevertheless, silicate and oxide behavior cannot account for the fractionation of the HSE, particularly the low Os, Ir, and Ru contents, observed in basaltic rocks.
Empirical estimates of partitioning derived from mineral separates suggest that Os, Ru, and Ir are highly compatible in Cr-bearing spinel, with partition coefficients of up to 150, while Pt and Pd are moderately compatible (Hart and Ravizza 1996; Puchtel and Humayun 2001). Experimental work on spinel–silicate melt partitioning at moderate to high fO2 suggests that for Fe-bearing spinels Ru, Rh, and Ir are all highly compatible with partition coefficients of 20 to > 1000, whereas Pd is barely compatible (Capobianco and Drake 1990; Capobianco et al. 1994, Righter et al. 2004). More recently it has been shown that the partition coefficients for Ir, Rh, and Ru are strongly controlled by the ferric-iron content of the spinels. For Cr-bearing spinels, in which Fe3+ is replaced by Cr3+, partition coefficients for Ir and Rh are much lower, and Pt and Pd are highly incompatible (Brenan et al. 2012).
Some of the first empirical data for olivine mineral separates were taken to indicate that Os may be compatible in olivine with an inferred olivine–silicate partition coefficient of ~ 20 (Hart and Ravizza 1996). However, other work on separated olivine suggested that Os is highly incompatible in olivine (Burton et al. 1999, 2000, 2002; Walker et al. 1999; Harvey et al. 2010, 2011). At this stage it is not clear whether these variations reflect compositional differences between samples, or simply the presence of micro-nuggets of sulfide or PGM in the separated silicate phase. Experimental work, however, suggests that many of the HSE are weakly compatible or only slightly incompatible, particularly under reducing conditions (Brenan et al. 2003, 2005).
Orthopyroxene and clinopyroxene
Empirical constraints from Hart and Ravizza (1996) suggest that Os may be compatible in orthopyroxene and clinopyroxene (Fig. 4), but other studies yield much lower Os concentrations for these phases (relative to coexisting sulfide or olivine) (e.g., Burton et al. 1999, 2000; Harvey et al. 2010, 2011). Experimental work indicates that Re may be mildly compatible in orthopyroxene and clinopyroxene under reducing conditions (e.g., Mallman and O’Neill 2007), but is incompatible under more oxidizing conditions (e.g., Watson et al. 1987; Righter and Hauri 1998; Righter et al. 2004; Mallman and O’Neill 2007). Platinum and Pd appear to be mildly compatible in clinopyroxene (Hill et al. 2000; Righter et al. 2004).
Overall, natural samples and experimental data suggest that silicate or oxide phases in the mantle do not exert a strong control on the behavior of the HSE during partial melting. Taking estimates of the proportion of silicate phases present in the upper mantle (e.g., Workman and Hart, 2005) partial melting of a sulfide-free mantle would yield melts that are slightly depleted in Os, Ir, and Ru, relative to their source. Such a pattern is consistent with that seen for high-degree melts, such as komatiites (Fig. 1a). Nevertheless, silicate and oxide behavior cannot account for the observed fractionations of the HSE, and in particular the low Os, Ir, and Ru contents, in basaltic rocks (Fig. 1a).
Refractory mantle sulfide
For natural magmatic and experimentally produced sulfide the data suggests that while the HSE are strongly partitioned into this phase there is little fractionation between the elements (with the exception of Re). Mantle sulfides, however, dominantly comprise refractory monosulfide solid solution (MSS) and Cu-rich sulfides, which together control much of the HSE budget of the upper mantle (e.g., Alard et al. 2000). Petrographic observations suggest that MSS often occurs as inclusions trapped in silicate phases, and is characterized by high Os, Ir, and Ru abundances, whereas the interstitial Cu-rich sulfides possess lower Os, Ir, and Ru contents (Fig. 5). The silicate hosted MSS sulfides were interpreted to be the refractory residues of partial melting, and the interstitial sulfides as having crystallized from a sulfide-bearing melt. On the basis of these observations it has been argued that the fractionation of the HSE during mantle melting might be accomplished by partitioning between refractory “solid” MSS and liquid sulfide (Bockrath et al. 2004a). However, at mantle temperatures of 1300–1400 °C and pressures of 5–16 kbar—that is, those appropriate for the generation of MORB (e.g., Klein and Langmuir 1987)—any refractory sulfide is likely to be completely molten well before the peridotitic silicate and oxide phases start to melt (Rhyzenko and Kennedy 1973; Hart and Gaetani 2006). Consequently, two phases of sulfide are unlikely to be stable during the melting that produces MORB, consistent with modeled depletion of mantle peridotites where MSS–sulfide melt partitioning cannot explain the observed variations in Pd, Pt, and Au (Fisher-Gödde et al. 2011). However, under conditions of melting at lower temperatures, for example, due to the presence of volatiles such as H2O and at fO2 lower than that at which sulfide is oxidized to sulfate, MSS fractionation may play a role in generating melts with low Os, Ir, and Ru contents (Mungall 2002; Mungall et al. 2006; Dale et al. 2012b; Botcharnikov et al. 2013).
Os–Ir–Ru metallic alloys
Osmium, Ir, and Ru (the IPGE) are not only strongly concentrated in refractory MSS, but also in platinum-group minerals (PGM), which encompass alloys and sulfides where Ru, Os, and/or Ir are the major metallic elements. It is clear from the distribution and absolute concentration of the IPGE in PGM (Fig. 6) that precipitation and accumulation of such phases will have a profound effect on IPGE/PPGE fractionation (Brenan and Andrews 2001). In addition, Pt-rich PGM, such as Pt–Ir alloys, have also been found in upper mantle lithologies (Luguet et al. 2007; Lorand et al. 2010). Palladium-rich PGM also exist, which may contain Pt, but typically Pd combines with bismuth and/or tellurium to form bismuthotellurides which are thought to be indicators of refertilization rather than being residual to melting (e.g., Lorand et al., 2010). Thus, Os–Ir–Ru and, to a lesser extent, Pt can all be retained by PGM during melting, while Pd is not. Some have argued that these alloys may represent material that was once part of the core, either as a result of incomplete segregation of metal to the core, or due to the entrainment of outer core material into the mantle at the core mantle boundary (Bird and Weathers 1975; Bird and Bassett 1980; Bird et al. 1999). However, recent experimental data suggests that metal originating in the outer core would possess similar concentrations of Os, Pt, and Re, rather than show an enrichment in Ru, Os, and Ir (Van Orman et al. 2008; Hayashi et al. 2009). The solubility of Os, Ir, and Ru is extremely low in silicate melts (e.g., Borisov and Palme 2000; Brenan et al. 2005). Therefore, it has been argued that Os–Ir–Ru-rich PGM may precipitate directly from a silicate melt, through nucleation on nanoclusters of HSE molecules (Tredoux et al. 1995). Furthermore, on the basis of the high solubility of Ir and Ru in sulfide melts it has been proposed that crystallization of Ru–Ir–Os alloys in the presence of a sulfide liquid is unlikely (Brenan and Andrews 2001). Rather it has been argued that such alloys can only precipitate from a melt that is sulfide-undersaturated (Brenan and Andrews 2001; Andrews and Brenan 20002b; Bockrath et al. 2004b; Barnes and Fiorentini 2008).
Together, these observations have been taken to suggest that the relationship between Os–Ir–Ru alloys and refractory sulfides in the mantle is key to understanding the behavior of the HSE during higher degrees of partial melting (e.g., Fonseca et al. 2012), where the removal of sulfur in silicate melts leads to a decrease in the proportion of sulfide in the source. All the while that sulfide remains present the HSE are quantitatively retained, and can reach wt% levels in sulfide. However, as soon as sulfide has been completely dissolved, Os–Ir–Ru–Pt alloys form in response to lowering of fS2 and diminished metal-sulfide complexation in the silicate melt (Fonseca et al. 2012). Effectively, much of the HSE budget of the mantle, with the exception of Re, remains in the mantle until sulfide has been completely removed, after which time Os–Ir–Ru and Pt are hosted by PGM phases rather than being liberated in a silicate melt. This model is consistent with an increasing number of petrographic observations indicating the presence of alloy phases in melt-depleted mantle peridotite (Luguet et al. 2003, 2007; Pearson et al. 2004; Brandon et al. 2006; Kogiso et al. 2008, Lorand et al. 2010, 2013; Fisher-Gödde et al. 2012)
The degree of partial melting needed to trigger PGM formation will depend on how much sulfur there is in the mantle source at the onset of melting, and is also a result of the solubility of S being inversely proportional to pressure (Mavrogenes and O’Neill 1999). The mantle that melts to produce MORB is already significantly depleted (e.g., Hofmann 1997), and the melting occurs at relatively shallow levels (e.g., Klein and Langmuir 1987). However, there is considerable uncertainty as to the amount of sulfur in the depleted mantle, with estimates ranging down to ~120 μg g−1 (Salters and Stracke 2004) compared to the concentration in primitive “fertile” (unmelted) mantle at 250 μg g−1 (Lorand 1990; O’Neill 1991; Palme and O’Neill 2003). Taking the S content of the MORB source mantle to be 120 μg g−1, then 15% melt extraction is needed to exhaust sulfide from the source, and thereby allow the generation of alloys in the mantle residue (Fonseca et al. 2011, 2012; Mungall and Brenan 2014). While these calculations indicate that even the depleted mantle requires significant degrees of melting to remove sulfide, such melt proportions are well within the range of estimates for the generation of MORB (e.g., Klein and Langmuir 1987). In this case PGM formation in the upper mantle may be a potential cause for the characteristic depletion of Os, Ir, Ru, and Rh, relative to Pt and Pd observed in MORB. The absence of significant fractionation of the HSE in komatiites, considered to represent higher degrees of melting than MORB, suggests that alloys are not stable at the higher pressure and temperature conditions required for the generation of such melts (cf. Mungall and Brenan 2014).
Overall, the natural and experimental data for mantle minerals indicates that all the while sulfide is present in the mantle, the HSE are largely retained during partial melting, the exception being Re that is not as strongly incorporated into sulfide, and is relatively soluble in silicate melts. However, if sulfide is removed from the system during high degrees of melting, at the pressure–temperature conditions appropriate for MORB melting, then this will result in the formation of Os–Ir–Ru alloys and/or sulfides.
Highly siderophile element behavior accompanying fractional crystallization
The major and trace element variations preserved in MORB indicates that their composition has been extensively modified by fractional crystallization, prior to eruption on the ocean floor (e.g., Klein and Langmuir 1987). The principal silicate phases involved in the fractional crystallization that generates MORB are olivine, plagioclase, and clinopyroxene (e.g., Klein and Langmuir 1987; Grove et al. 1992). In general, the more evolved MORB (that is, those with lower MgO and Ni contents, due to the crystallization and removal of olivine) possess lower HSE contents (Fig. 7). On the basis of early empirical estimates for the partitioning of Os into olivine, this relationship has led some to suggest that the HSE are compatible in this phase and removed from the silicate melt. However, as discussed previously, with the exception of Re, there is little evidence to suggest that the HSE are strongly partitioned into olivine, plagioclase, or clinopyroxene (Fig. 4).
Most MORB are considered to be sulfur saturated (Wallace and Carmichael 1992) and sulfide is a ubiquitous phase. Nevertheless, even if MORB melts are sulfur saturated at their source, they are likely to arrive at the surface undersaturated, because the sulfur content at sulfide saturation increases dramatically at lower pressures (e.g., Mavrogenes and O’Neill 1999). In this case the only viable mechanism by which MORB melts can become sulfur saturated is through extensive fractional crystallization, driving the residual melt to higher S contents. Therefore, it seems most likely that it is the fractional crystallization of olivine, plagioclase and clinopyroxene that drives the melt to sulfur saturation, resulting in the precipitation of sulfide. Hence, the relationship between Ni (concentrated in olivine) and the HSE (concentrated in sulfide) can be attributed to the coupled crystallization of silicates and sulfide.
Sulfide may be present in relatively high proportions in MORB (up to ~ 0.23% by mode, K. Kiseeva, 2015), which strongly incorporates most of the HSE, with sulfide/silicate melt partition coefficients of between 104 and 106. In contrast, Re, while still being compatible in sulfide, has a sulfide-silicate melt partition coefficient at least two orders of magnitude lower than that of the other HSE (DRe ~ 10–103). MORB sulfides have high Os (and other HSE) contents, and low Re/Os relative to their parental melt. Consequently, the effect of sulfur saturation and sulfide crystallization will be to decrease absolute HSE abundances, and to raise Re/Os in the residual melt.
THE 187Re–187Os ISOTOPE SYSTEM AND THE FORMATION OF MID-OCEAN RIDGE BASALT (MORB)
Mid-ocean ridge basalts form by partial melting of the Earth’s upper mantle, and variations in their radiogenic isotope compositions or concentration ratios of incompatible elements are considered to reflect compositional heterogeneity in the mantle source (Tatsumoto 1966; O’Nions et al. 1977; Kay 1985; Hofmann 1997). These compositional variations occur on a variety of scales and tectonic settings, ranging from the global-scale of the so-called DUPAL anomaly (centered on the Indian ocean) (Dupré and Allègre 1983; Hart 1984; Hamelin and Allègre 1985; Hamelin et al. 1986; Michard et al. 1986; Price et al. 1986; Dosso et al. 1988; Mahoney et al. 1989, 1992; Rehkamper and Hofmann 1997; Escrig et al. 2004); to those associated with ocean-island volcanics or near-ridge seamounts (White and Schilling 1978; Zindler et al. 1984; Brandl et al. 2012); to minor pervasive variations within ridge segments of normal MORB (e.g., Hofmann 1997; Agranier et al. 2005). A number of processes have been put forward to account for these compositional variations including variable degrees of mantle depletion by prior partial melting (e.g., DePaolo and Wasserburg 1976; Zindler et al. 1984), the infiltration of silicate melts or fluids (e.g., Green 1971), or recycling of lithospheric material into the mantle (e.g., Hofmann 1997).
The 187Re–187Os isotope system, based on the long-lived b− decay of 187Re to 187Os, potentially provides an exceptional tracer of recycled lithosphere in Earth’s mantle. This is because both oceanic and continental crust possess exceptionally high Re/Os (parent/daughter ratios), and develop radiogenic Os isotope compositions over time (e.g., Pegram and Allègre 1992; Shirey and Walker 1998; Hauri 2002). In contrast, portions of the lithosphere have low Re/Os, and evolve to unradiogenic Os isotope compositions relative to that of the primitive upper mantle (PUM) (Walker et al. 1989; Pearson et al. 1995). These distinctive isotope signatures can be readily traced as recycled material if mixed back into the convective mantle. For example, the 187Os/188Os variations seen in HIMU ( = high μ = elevated 238U/206Pb) ocean island basalts indicate the presence of material that has evolved over a long-time period with a high Re/Os, consistent with models indicating recycled oceanic lithosphere in the source of these volcanic rocks (Day et al. 2010b; Day 2013).
Some of the earliest measurements of 187Os/188Os in MORB also yielded isotope compositions more radiogenic than estimates for the primitive upper mantle (e.g., Martin 1991; Roy-Barman and Allègre 1994) and these were attributed either to contamination by seawater derived Os, or melting of a heterogeneous mantle (e.g., Martin 1991; Roy-Barman and Allègre 1994). The work of Schiano et al (1997) on normal MORB, however, not only indicated relatively radiogenic Os isotope compositions but also that these compositions appeared to co-vary with the Sr–Nd and Pb isotopes of the same samples. For the DUPAL anomaly, radiogenic Os isotope compositions were taken to indicate the presence of mafic continental crust in the mantle source (Escrig et al. 2004). On the other hand, radiogenic 187Os/188Os compositions for MORB from the south Atlantic were attributed to metasomatism of the asthenospheric mantle, and local effects from plume–ridge interaction (Escrig et al. 2005a). At first sight, the data from these studies might be taken to suggest that the Os isotope variations reflect those of the MORB mantle source, rather than a secondary process, and that Os isotopes do indeed act as a sensitive tracer of different recycled or enriched material in the mantle. However, these data also indicate a covariation between the Os isotope composition and the Os elemental abundance in these samples (Schiano et al. 1997; Escrig et al. 2005a). Covariations between Os, Ni, and Mg contents in MORB are most readily explained by fractional crystallization (e.g., Burton et al. 2002), but in this case it is then difficult to attribute the Os isotope variations to a mantle source, leading some to propose that the radiogenic Os isotope ratios reported by these studies must result from seawater derived contamination (e.g., Shirey and Walker 1998; Hart et al. 1999; Standish et al. 2002; Peucker-Ehrenbrink et al. 2003). Subsequent work demonstrated that many of the MORB previously analyzed (Schiano et al. 1997; Escrig et al. 2004, 2005a) had been affected by an analytical artefact (Gannoun et al. 2007), nevertheless a number of MORB samples still possessed relatively radiogenic isotope compositions (Gannoun et al. 2004, 2007; Yang et al. 2013).
Despite the potential utility of the Re–Os isotope system, in particular for tracing the presence of recycled material in MORB, these studies highlight the particular difficulties of both the measurement and the interpretation of 187Re–187Os isotope data in MORB. Mid-ocean ridge basalts possess extremely low Os concentrations, usually less than 10 parts per trillion (pg g−1) which, not only makes their accurate measurement challenging, but also renders MORB highly susceptible to effects that are rarely seen in lithophile elements isotope systems (such as Rb-Sr or Sm-Nd). Such effects include; (i) radiogenic ingrowth of 187Os, produced from the decay of 187Re over very short periods of time (< 10 kyr), (ii) seawater contamination, both directly, on the sea floor, or indirectly in the magmatic plumbing system, and (iii) sample heterogeneity, due to variable contamination in glass or amongst coexisting magmatic phases or through sulfide nugget effects.
Osmium has seven naturally occurring isotopes, two of which 187Os and 186Os are the decay products of long lived radioactive isotopes, 187Re and 190Pt. Of these two decay schemes, the Re–Os method has been used as dating tool and geochemical tracer for over four decades (Shirey and Walker 1998). Despite its great potential as a geochemical tool, analytical difficulties initially limited the application of the osmium isotope method, mainly because of the high ionization potential of Os (ca. 9eV). The discovery that a solid Os sample could yield negative molecular ions by conventional thermal ionization (Creaser et al. 1991; Volkening et al. 1991) rendered largely obsolete all the excitation methods for atomic osmium used before (Hirt et al. 1963; Luck and Allègre 1982; Walker and Fasset 1986). In the negative thermal ionization mass spectrometry (N-TIMS) method Os is measured as osmium trioxide (OsO3−) via heating on platinum filaments with an electron donor. A Ba–Na emitter solution is employed to lower the work function of the filament, which enhances the emission of negative ions. The formation of the Os oxide species is also advantaged by bleeding oxygen into the source (Walczyk et al. 1991). The ionization efficiency increases significantly with decreasing Os loads and can reach above 30% at the pg Os level (Birck 2001; Gannoun and Burton 2014).
Another major problem with Re–Os isotopic analysis has been the chemical behavior of Os in solution because of the numerous oxidation states including the volatile tetraoxide species (OsO4). At present, no single technique is equally applicable to all matrices particularly when organic matter and/or refractory mineral phases are present because the variable oxidation states may inhibit the complete homogenization of Os between sample and spike.
High temperature (~ 250 °C) oxidizing digestions using either Carius tubes (Shirey and Walker 1995) or high-pressure asher (HPA) digestion vessels (Meisel et al. 2003) have the merit of dissolving acid-resistant phases such as chromite and noble metal alloys. These methods have been supplemented by employing HF digestion after Carius tube/HPA digestion (e.g., Ishikawa et al. 2014), but with mixed results (Day et al. 2015). However, such techniques can potentially yield high total analytical blanks that can contaminate low-HSE abundance samples, such as MORB. Mid-ocean ridge basalt glass possesses low Os abundances, with some samples in the range of 0.2 and 3 ppt, in which refractory minerals are usually absent. For these reasons low-temperature digestion techniques have been used in preference to other approaches when analyzing Os in MORB. These use HF and HBr in sealed Teflon vessels at temperatures of ≤ 140 °C, followed by extraction of Os in liquid bromine (Birck et al. 1997). Extremely low blanks of < 50 fg of Os have been achieved with this method (Gannoun et al. 2004, 2007). Furthermore, MORB glasses are likely to be completely dissolved in HF–HBr acids mixtures even at room temperature.
Mid-ocean ridge basalt sulfide grains can be extracted directly using a magnet and handpicked under a binocular microscope (Gannoun et al. 2004, 2007; Harvey et al. 2006) or removed from hand-polished slabs using a diamond scribe to etch around and under the grains (Warren and Shirey 2012). The grains are weighed, spiked with 185Re–190Os and dissolved with high purity HBr. The Os fraction is then purified using microdistillation (Birck et al. 1997; Harvey et al. 2006; Gannoun et al. 2007). It is also possible to undertake dissolution simultaneously with microdistillation (Warren and Shirey 2012). The purified Re and Os are analyzed by N-TIMS following the method described by Pearson et al. (1998). Osmium analysis in sulfides can also be achieved using in situ laser ablation techniques. The strength of this technique lies in the ability to relate Os isotope information from individual sulfide to their precise spatial and textural setting in the rock (Pearson et al. 2002). However, single-sulfide Os data analyzed by the N-TIMS technique are typically of a much higher precision than in situ analysis (cf. Pearson et al. 1998; Harvey et al. 2006; Gannoun et al. 2007) even for sulfide with low Os contents (i.e., less than 10 μg g−1). Moreover, for in situ analysis, because of the isobaric interference of 187Re on 187Os accurate measurement of 187Os/188Os is only possible for sulfides with low 187Re/188Os (Pearson et al. 2002). Such conditions are typically only met in the case of mantle sulfides.
Rhenium–Osmium elemental variations in MORB glass
The fractionation of Re and Os accompanying the generation of MORB is one of the key processes controlling the distribution of these elements between Earth’s mantle and crust. Osmium behaves as a highly compatible element during partial melting, and is preferentially retained in the residual mantle. Consequently, MORB have much lower concentrations, ranging from 0.18 to 170 pg g−1 (with a mean of 10 pg g−1) than mantle peridotite, ranging from 800 to 13000 pg g−1 (with a mean of 3900 ng g−1). In contrast, Re is moderately incompatible during partial melting and preferentially enters the melt. Accordingly, MORB have high Re concentrations, ranging from 480 to 3000 pg g−1 (with a mean of 1023 pg g−1) compared to 10–450 pg g−1 in mantle peridotite (with a mean of 200 pg g−1) (Fig. 8).
By comparison, komatiiites have generally much higher Os concentrations, up to 10,000 pg g−1, with a similar range of Re concentrations as MORB. These high Re and Os concentrations are generally attributed to higher degrees of melting. Ocean island basalts (OIB) have Os concentrations that range from 1 to 500 pg g−1, and arc lavas from 0.1 to > 10 pg g−1. The low Os concentration of many arc lavas is likely due to extensive removal during fractional crystallization and, indeed, in cases where basaltic compositions have been sampled, Os concentrations can be greater than 50 pg g−1 (e.g., Woodland et al. 2002; Dale et al. 2012b). The relatively low Re concentration of many arc lavas and OIB was originally thought to reflect differences in the mineralogy of the mantle source or the extent of melting, but it is likely that for many of these samples the low Re concentrations result from volatile behavior during sub-aerial eruption (e.g., Lassiter 2003; Day et al. 2010b; Gannoun et al. 2015b). As outlined previously, the low Os concentration of MORB is likely to result, in part, from preferential partitioning into residual mantle sulfide and/or PGM phases and, in part, to the low solubility of Os in silicate melts. In addition, the Os composition of primitive MORB melts will be further reduced by sulfide segregation during fractional crystallization. In contrast, the relatively high Re concentrations result, in part from Re being much less strongly incorporated in mantle sulfide and PGM phases and, in part, from much of the Re budget being controlled by silicate phases, and having a much higher solubility in silicate melts. Rhenium is removed into both silicates and sulfide during fractional crystallization.
A remarkable feature of MORB, and indeed all other terrestrial basalts, is the relatively constant increasing fractionation of Re/Os with decreasing Os content. The values range from mantle Re/Os values of around 0.01 for Os concentrations of 2–7 ng g−1, to Re/Os values of ~1000 for lavas with concentrations of 0.1 pg g−1 (Fig. 9). The systematic nature of this fractionation, suggests either that it is dominantly controlled by a single process, such as mantle melting or fractional crystallization, or else that several process act to have the same effect, for example, fractionation by refractory mantle sulfide and also by sulfide segregation during fractional crystallization.
Rhenium shows a broad positive covariation with Al2O3 and sulfur consistent with the incompatibility of all these elements during mantle melting (Fig. 10). The positive Re–S covariation might be explained by the fact that both elements will be removed into sulfide during fractional crystallization, resulting in a decreasing S and MgO content during the differentiation of S-saturated MORB (Mathez 1976; Bezos et al. 2005; Ballhaus et al. 2006). Despite significant scatter, Os broadly covaries with Ni in MORB (Fig. 11), consistent with a role for olivine crystallization in Os partitioning. Although previous studies have attributed the Os–Ni covariation directly to the compatibility of Os in olivine (Brügmann et al. 1987; Hart and Ravizza 1995), natural samples and experiments indicate that Os is much less compatible. Burton et al. (2002) have shown that Os is in fact extremely incompatible in olivine. Rather it is the crystallization of olivine that drives the melt to sulfur saturation, which in turn results in sulfide precipitation (in which Os is highly compatible) that is trapped within the olivine as ‘melt inclusions’ (Walker et al. 1999; Burton et al. 2002; Brenan et al. 2003, 2005). In summary, Re and Os display similar overall behavior in MORB from the three major ocean basins. Osmium is highly compatible during melting and fractional crystallization, whereas Re is moderately incompatible
The 187Os/188Os isotope variations in MORB glass
The 187Os/188Os isotope compositions for MORB from the Pacific, Atlantic and Indian oceans are shown against the reciprocal of the concentration in Figure 12. Mid-ocean ridge basalts from the three major oceans show a similar range of 187Os/188Os isotope compositions, ranging from 0.126 to 0.148 with a mean value of 0.133 ± 0.009 (2σ st. dev.) (Gannoun et al. 2004, 2007; Yang et al. 2013). There is no overall correlation with Os concentration (cf. Schiano et al. 1997; Escrig et al. 2004), however, in general MORB glasses have Os concentrations in the following order: Indian > Atlantic > Pacific, and those samples with a higher Os concentration have a tendency to possess more radiogenic 187Os/188Os compositions. Comparison of 187Os/188Os with 187Re/188Os on a conventional isotope evolution diagram (Fig. 13) indicates that there is no systematic covariation. The data do, nevertheless, indicate that MORB glasses with lower 187Re/188Os are generally found in the Indian > Atlantic > Pacific. In addition, those samples with the lowest 187Re/188Os tend to possess the most radiogenic isotope compositions.
With regard to the long-lived radiogenic isotopes of Sr, Nd, and Pb, while the cross-linked data are limited, there are no systematic variations between 187Os/188Os and 87Sr/86Sr, 143Nd/144Nd, and 206Pb/204Pb (Fig. 14). Similarly, there is no correlation between 187Os/188Os composition and ridge bathymetry or spreading rate (Fig. 15) (using data compilation of DeMets et al. 2010 and Argus et al. 2011).
Analytical issues associated with MORB
Several studies have reported 187Os/188Os data for MORB glass (Schiano et al. 1997; Escrig et al. 2004) that could not be reproduced elsewhere, using lower blank techniques (Gannoun et al. 2007). Comparison of these data shows that for many of the relatively unradiogenic samples there is reasonably good agreement between studies (Figs. 16 and 17) but notably none of the very radiogenic values previously reported were reproduced for the same samples. Such a difference might be attributed either to the nature of the samples or the methods involved in their preparation for chemistry. The earlier studies used leaching techniques to remove any Fe–Mn oxyhydroxides that may have accumulated on the glass while on the sea floor. Iron-manganese precipitates, if present, are likely to possess a radiogenic Os isotope composition acquired from seawater (187Os/188Os = ~1), therefore they might shift the measured 187Os/188Os to more radiogenic values. However, experiments on some of the same glasses indicate that extensive leaching, with oxalic acid and HBr, yields indistinguishable results to those for the same glass samples simply rinsed in dilute HCl, ethanol and water. Another possibility is that because of the large samples sizes used in the earlier studies, between 1 and 5 g (Schiano et al. 1997; Escrig et al. 2004) compared to 300–500 mg (e.g., Gannoun et al. 2007), phenocrysts possessing radiogenic isotope compositions may have been inadvertently included in the material measured. Likewise, entrainment of included sulfides possessing radiogenic compositions may have the same affect. If the radiogenic 187Os/188Os were due to the presence of entrained silicates or sulfides, then some variation in the parent/daughter ratio might be expected (cf. Fig. 9 of Day 2013). Such heterogeneity is spectacularly displayed in two samples from the same locality in the Indian Ocean, where significant variations in the isotope and elemental composition of MORB glass can be attributed to the variable the presence of sulfide inclusions. However, duplicate and triplicate measurement of eleven of the samples showed no resolvable variation, and there is no evidence for isotope and elemental heterogeneity in any of these glass samples. Therefore, it seems more likely that the difference in measured 187Os/188Os composition is an analytical artefact. One possibility is that this is due to interference from 187ReO3− on the measured 187OsO3−, although this can be carefully monitored during N-TIMS analysis through the direct measurement of 185ReO3−. More likely is that the earlier data were under-corrected for the total procedural blank during chemical purification. The blanks of the original studies possessed a radiogenic 187Os/188Os composition, and the difference between the earlier data (Schiano et al. 1997; Escrig et al. 2004) and those samples that were re-analyzed increases with decreasing Os concentration in the sample, consistent with increasing contribution from the blank (Fig. 17). Overall, these studies highlight the analytical difficulties of obtaining accurate 187Os/188Os data for MORB glass many of which possess low Os concentrations (i.e., between 0.2 and 5 pg g−1).
The origin of Os isotope variations in MORB glass
Notwithstanding any shifts that arise from analytical problems, the data obtained thus far, for all the major oceans, indicates a resolvable variation in the 187Os/188Os isotope composition of MORB, ranging from values similar to those expected for the primitive upper mantle (e.g., Meisel et al. 1996) to radiogenic compositions akin to those found in ocean island basalts (e.g., Day 2013). It is unlikely that these data have also been compromised by analytical problems; first, because there is no covariation between the corrected 187Os/188Os and the Os concentration, as might be expected if the blank concentration was not correctly determined. Second, replicates with differing sample weights and subject to different dissolution technique yield indistinguishable 187Os/188Os values (Gannoun et al. 2007; Yang et al. 2013, Burton et al. 2015). Moreover, those samples with radiogenic 187Os/188Os compositions are actually those with the highest Os concentrations, and therefore would be less susceptible to any blank effect. Finally there is no significant covariation between Os and Sr, Nd or Pb isotopes, as might be expected if the variations were due to compositional heterogeneity in the mantle source.
Radiogenic growth of 187Os since MORB eruption
For lithophile elements, such as Sr or Nd, parent/daughter ratios in MORB glass and coexisting silicates are relatively low, consequently shifts in their radiogenic isotope composition are unlikely to have a measurable effect for timescales less than 103 million years (e.g., Hofmann 1997). Therefore variations in Sr or Nd isotope composition preserved in MORB can be attributed to compositional heterogeneity in the upper mantle source (e.g., Hofmann 1997). For the 187Re–187Os system however, silicate phases and glass possess exceptionally high 187Re/188Os (parent/daughter). This then raises the possibility that radiogenic 187Os could be produced in situ from the decay of 187Re over relatively short periods of time (that is a few hundred thousand years or less; e.g., Hauri et al. 2002, Gannoun et al. 2004, 2007). For example, MORB glass possesses 187Re/188Os with values ranging from 30 to 8000 (Gannoun et al. 2007; Yang et al. 2013), and a glass with 187Re/188Os = 4000 would produce a shift in 187Os/188Os from mantle values of 0.1296 to a value of 0.14 in less than 250 thousand years (Gannoun et al. 2007). This effect is illustrated in Figure 13, where timescales of between 50 ka and > 1 Ma could produce the range of 187Os/188Os preserved in the MORB glasses if they were simply due to the decay of Re.
One approach to determining the age of crystallization of the MORB glasses is the measurement of short-lived isotopes of Th–U and Ra in the same samples. Such Th–U–Ra data was obtained for a few MORB glasses spanning much of the observed range of 187Os/188Os compositions for the datasets in Gannoun et al. (2004, 2007). Of those samples measured, if it is assumed that they initially possessed a PUM-like composition at the time of crystallization, then between 700 kyr and 1.25 Myr would be required to generate their given 187Os/188Os isotope compositions. However, the same samples possess 230Th/232Th activity ratios greater than 1, suggesting that they must be ≤ 350 kyr old (that is, the maximum time available before all 230Th has decayed). Moreover, all but one sample has a 226Ra/230Th activity ratio that is also greater than 1, suggesting those samples must be ≤ 8 kyr old. Therefore, for these samples, at least, the radiogenic 187Os/188Os compositions cannot be explained solely as a result of in situ decay of 187Re subsequent to igneous crystallization (Gannoun et al. 2004, 2007).
An alternative approach that can be used with phenocryst-bearing MORB samples is to obtain Re–Os isotope data for the constituent phases in MORB, including sulfide, glass, spinel, olivine, clinopyroxene and spinel (Gannoun et al. 2004). If these coexisting phases are in Os isotope equilibrium, then they may yield an isochron that will give the age of crystallization, and the initial Os isotope composition defined by the best-fit line will correspond to that of the mantle source. However, if some of the phases were assimilated from previously crystallized basalts, gabbro (from deeper in the oceanic crust), or contaminated by seawater, then they may possess different isotope information to that of the host glass or other minerals (Gannoun et al. 2004). 187Re–187Os data were obtained for coexisting phases from two MORB samples from the FAMOUS region on the mid-Atlantic ridge (Figs. 18 and 19). These results illustrate the age information that can be obtained from MORB glass and coexisting phases, some of the processes involved in MORB genesis, and the 187Os/188Os composition of the MORB source. Sample ARP1974-011-018 (36.85°N; 33.25°W) is an olivine basalt containing olivine (Fo90–Fo80), plagioclase (An91–An95), and clinopyroxene (Wo44En15Fs5–Wo40En15Fs9) phenocrysts (up to 1–2 mm in diameter) and microphenocrysts in a hyalocrystalline matrix, and, in places, a glassy pillow rim (e.g., Le Roex et al. 1981). The 187Re–187Os isotope data for matrix, glass, plagioclase, and olivine yield a best-fit line corresponding to an age of 565 ± 336 ky and an initial 187Os/188Os of 0.1265 ± 0.0046 (Fig. 18). The data for clinopyroxene are distinct from this best-fit line, suggesting either an older age or a different and more radiogenic source for this phase. Sample ARP1973-010-003 (36.8372°N; 33.2482°W; 2760-m water depth) is a porphyritic, picritic basalt with abundant olivine phenocrysts (Fo91–Fo89; up to 5 mm in diameter) set in a glassy to hyalocrystalline matrix. Cr-spinel [Cr/(Cr + Al) = 48.01] phenocrysts and sulfide [~14 weight percent (wt %) Ni] blebs (up to 1 mm in diameter) occur as inclusions in olivine or discrete crystals in the groundmass. Plagioclase (An80) microlites are also common (Le Roex et al. 1981, Su and Langmuir 2003). The 187Re–187Os data for olivine, plagioclase, glass, and sulfide yield a best-fit line corresponding to an age of 2.53 ± 0.15 My and an initial 187Os/188Os ratio of 0.129 ± 0.002 (Fig. 19). Spinel, which is relatively Os-rich (Table 1 of Gannoun et al. 2004), possesses a distinct isotope composition from this best-fit line and is probably the phase responsible for the displacement of the matrix from the same line.
The simplest interpretation of these data is that the ages represent the time of igneous crystallization and the initial Os isotope composition represents that of the mantle source. The crystallization ages are, however, much older than might be expected from age-distance relations with the ridge axis that suggest ages of 5–10 kyr (Selo and Storzer 1979). They are also different to the ages inferred from the Th–U–Ra isotope composition of the glass. Glass from sample ARP1974-011-018 glass gives a 226Ra/230Th activity ratio close to 1, suggesting that the sample is ≤ 8 ky old, whereas the 230Th/232Th activity ratio is 1.273, suggesting that the sample is ≤ 350 ky old, consistent with previous 230Th data for the same sample (Condomines et al. 1981). Arguably the 187Re–187Os age of 565 ± 336 kyr is indistinguishable from the 230Th age constraints. Glass from sample ARP1973-010-003 gives 226Ra/230Th ratio of 1.3, which might at first be taken to indicate that the sample is less than 8 ky old. However, the same sample has a 234U/238U ratio of 1.043, and such elevated values are often taken to indicate seawater contamination, consistent with previously published data for this sample (Condomines et al. 1981), which raises the possibility that Ra has also been affected by the same seawater contamination. It might be argued that the best-fit lines are due to contamination by radiogenic Os from seawater, rather than having some age significance. This would require that the contamination occurred during mineral crystallization and has affected phases such as olivine and plagioclase in a systematic manner; otherwise, it is difficult to imagine how different phases would align to yield the correlations observed.
Alternatively, the data may indicate that few if any of the constituent phases crystallized in their present basalt host (i.e., they are xenocrysts not phenocrysts). There is evidence for assimilation of xenocrystic phases in samples from the FAMOUS region (e.g., Clocchiati 1977; le Roex et al. 1981; Shimizu 1998). For example, in this sample high-Al spinel is considered to be a relict from high-pressure crystallization (Sigurdsson and Schilling 1976), which suggests that spinel is not in Os isotopic equilibrium with the other phases. However, if most of the phases lie on the same best-fit line, then this interpretation demands that all such minerals are xenocrysts. For the picritic basalt, if eruption occurred about 5–10 kyr ago, then the Re–Os isotopic data indicate that original crystallization of the minerals occurred about 2.5 Myr prior to this event. In this case, the xenocrysts were assimilated from previously solidified “olivine–plagioclase” basalts, or cumulates through which the present host basalts have ascended.
Taken together, these results demonstrate that the radiogenic 187Os/188Os composition of MORB glass can be readily generated from the decay of 187Re over very short timescales (that is, a few hundred thousand years or less). Nevertheless, the ages obtained for the samples from the FAMOUS region on the mid-Atlantic ridge are much older than might be expected on the basis of their distance from the ridge axis, and this can only be explained either by seawater contamination (that occurred during the crystallization of magmatic minerals) or by the entrainment of crystals (i.e., xenocrysts) from older oceanic crust.
Extreme 187Os/188Os heterogeneity in MORB glass
Occasionally MORB itself shows significant Os isotope and elemental heterogeneity. For example, replicate measurements of the MORB sample EN026 10D-3 show significant heterogeneity, with 187Os/188Os isotope compositions that range from 0.128 to > 0.15 (Day et al. 2010b). For MORB glass this is exemplified by two samples from the same locality on the central Indian ridge, MD57 D9-1 and D9-6 (8.01°S; 68.07°E) which show 187Os/188Os compositions ranging from 0.126 to 0.254, with covariations in Os concentration (Fig. 20). Those samples with the least radiogenic 187Os/188Os composition possess unusually high Os concentrations (up to 220 pg g−1). Sulfides from the same samples possess 187Os/188Os between 0.126 to 0.132, and concentrations between 136 and 246 ng g−1. Given the presence of Os-rich sulfide in these samples, it seems most likely that this heterogeneity is due to the entrainment of this phase. If the radiogenic 187Os/188Os isotope composition of the glass is simply due to the radiogenic growth of 187Os from the decay of 187Re since the time of igneous crystallization, then the initial ratio determined from elemental or parent/daughter ratios may reflect the composition of the source (cf. Day 2013). Alternatively, if the radiogenic composition of the glass is due to seawater contamination or altered oceanic crust then the initial 187Os/188Os isotope composition determined from such covariations may have little relationship with that of the mantle source.
Seawater contamination or assimilation of altered oceanic crust
The age constraints from spreading rates, Th–U–Ra disequilibria and 187Re–187Os isotope data for MORB glass and coexisting minerals suggest that the radiogenic 187Os/188Os compositions of MORB glass cannot be solely explained by an age effect following igneous crystallization. An alternative possibility is that these radiogenic compositions could be due to seawater contamination, either occurring directly during quenching of the glass on the ocean floor or through the assimilation of hydrothermally altered oceanic crust in the magmatic plumbing system. Seawater possesses a radiogenic 187Os/188Os composition (~1.026–1.046) (e.g., Sharma et al. 2012, Gannoun and Burton 2014) and a 187Re/188Os ratio of ~3400, (calculated using the Os concentrations from Sharma et al. 2012, Gannoun and Burton 2014 and Re from Anbar et al. 1992, Colodner et al. 1993). In this case, seawater contamination could account for both the radiogenic Os isotope composition and the tendency of such samples to possess relatively low 187Re/188Os.
Trace elements that are enriched in seawater, such as Cl or B could potentially be used as indicators of seawater contamination. At first sight, however, there is no apparent covariation of either B or Cl with 187Os/188Os in the MORB glasses. Rather the variations that do exist indicate that many of the samples with radiogenic Os compositions possess low Cl and B concentrations, inconsistent with seawater contamination (Fig. 21). The difficulty in interpreting Cl and B is that both are highly incompatible elements, and therefore they are strongly affected by partial melting and fractional crystallization (Michael and Schilling 1989; Chaussidon and Jambon 1994; Jambon et al. 1995; Michael and Cornell 1998). Indeed, Cl and B for the same MORB glasses show a negative covariation with MgO suggesting that fractional crystallization has strongly influenced their abundances, thereby masking any subtle effects from seawater contamination. Like Cl and B, K also behaves as a highly incompatible element during melting and crystallization, in this case an alternative approach is to use incompatible element ratios such as B/K or Cl/K that are not significantly fractionated during crystallization to place some constraints on potential contamination by seawater. For example, mantle Cl/K ratios are low (< 0.08), whereas altered oceanic crust has a Cl/K ratio ~ 0.1, and seawater ~ 50 (Michael and Schilling 1989; Jambon et al. 1995; Michael and Cornell 1998). However, again there is no clear co-variation of Cl/K with 187Os/188Os, rather the radiogenic Os values appear to possess low Cl/K (Gannoun et al. 2007).
A more robust tracer of seawater interaction is provided by 11B/10B of the MORB glasses. The upper mantle is thought to possess a δ11B value (Chaussidon and Marty 1995) of −10‰, (where δ11B = 1000 × [(11B/10Bsample/11B/10Bstandard) − 1] relative to the borate standard NBS 951 with an 11B/10B ratio of 4.04558). In contrast, for altered oceanic crust δ11B ranges from +2 to +9‰, seawater has a δ11B = +39.5‰ (e.g., Spivak and Edmond 1987; Smith et al. 1995) and serpentinized oceanic mantle samples can range from +9‰ to +39‰ (Boschi et al. 2008; Vils et al. 2009; Harvey et al. 2014a). While melting and crystallization processes are unable to significantly fractionate boron isotopes, mixing with altered oceanic crust and mantle can account for the δ11B range of −7 to −1‰ observed in MORB (Chaussidon and Jambon 1994).
The δ11B values of MORB glasses for which 187Os/188Os data are available range from −9 to +2‰, and those samples with high δ11B values also possess radiogenic 187Os/188Os compositions (Fig. 22). The B concentration of seawater is ~4.6 μg g−1 which is some 5–10 times higher than that of unaltered MORB (<1 μg g−1), whereas the Os concentration in seawater of 10−2 pg g−1 is some 3 orders of magnitude less than that of average MORB. Direct mixing of seawater would be dominated by B at the low mixing proportions suggested in Figure 22 (that is, a horizontal vector in Os versus δ11B) indicating that the radiogenic 187Os/188Os and high δ11B values cannot be easily explained by direct contamination from seawater. Conversely, contamination from Fe-Mn crust with a seawater Os isotope composition would produce far greater shifts in 187Os/188Os than B. Rather, the co-variations are entirely consistent with the assimilation of between 5–10% of altered oceanic crust with a variable 187Os/188Os composition.
It might be argued that the relatively heavy δ11B values (> −5‰), and the radiogenic 187Os/188Os could be due to the presence of recycled oceanic crust (present as pyroxenite) in the MORB mantle source. Recycled oceanic crust can lose substantial amounts of Re during subduction (~50% or more, Becker 2000; Dale et al. 2007) but Re/Os ratios are still sufficiently elevated to produce radiogenic 187Os/188Os values with time. However, recent studies suggest that during dehydration of the subducting slab, B is preferentially partitioned into the released fluids, leaving a depleted residue (Moran et al. 1992; Bebout et al. 1993; Peacock and Hervig 1999; Nakano and Nakamura 2001; Harvey et al. 2014b). Furthermore, boron-isotope fractionation occurs during such dehydration and the residue becomes increasingly enriched in the light B isotope (10B) generating light δ11B values (You et al. 1996; Ishikawa et al. 2001; Leeman et al. 2004; Dale et al. 2007), rather than the heavy values required to generate the ranges observed in MORB.
Notwithstanding analytical difficulties, the Os isotope and elemental variations in MORB glass, the mismatch in age constraints and measured 187Os/188Os compositions, and the covariations with B isotopes suggest that assimilation of seawater-altered oceanic crust is likely to be the dominant process responsible for the radiogenic Os-isotope signal seen in many of the MORB glasses studied thus far.
SULFIDES IN MID-OCEAN RIDGE BASALTS
Petrology and chemistry
Sulfide is a ubiquitous phase in MORB glass, indicating that these melts were sulfur saturated (Wallace and Carmichael 1992). Because decompression will drive the melt away from sulfide saturation (e.g., Mavrogenes and O’Neill 1999) it might be expected that most MORB would be undersaturated when transported to lower pressures during eruption. The presence of sulfide globules in early crystallizing phases, however, clearly indicates that MORB are sulfur saturated during the initial stages of magmatic evolution (Mathez and Yeats 1976; Patten et al. 2012; Yang et al. 2014) and, as previously suggested for MORB, this sulfur saturation is most likely to result from fractional crystallization itself. In addition, MORB contain more sulfur than subaerially erupted basalt, because degassing is impaired by the overlying pressure of seawater.
Sulfides occur as spherules embedded in the walls of large vesicles (Moore and Calk 1971; Moore and Schilling 1973), as small irregular grains in microcrystalline aggregates of plagioclase and olivine (Mathez and Yeats 1976) and as well-developed spherical globules, in glass or in phenocrysts (Mathez and Yeats 1976; Czamanske and Moore 1977; Roy-Barman et al. 1998; Patten et al. 2012) (Figs. 23a,b). The globules, which range from 5 to 600 μm in diameter, have different textures that can be divided in three groups (Moore and Calk 1971; Mathez 1976; Mathez and Yeats 1976; Czamanske and Moore 1977; Peach et al. 1990; Roy-Barman et al. 1998; Patten et al. 2012, 2013). The first, comprise a fine grained micrometric intergrowth of Fe–Ni-rich and Cu–Fe-rich sulfide phases that represent quenched monosulfide solid solution (MSS) and intermediate solid solution (ISS). The second, comprise globules of coarser grained intergrowth of MSS and ISS with pentlandite and oxide (Mathez 1976; Czamanske and Moore 1977; Patten et al. 2012) and the third group comprise zoned globules that consist of two massive and distinct grains of MSS and ISS, first identified recently by Patten et al. (2012).
Pentlandite and oxide occur to a lesser extent in all types of textures. Sulfide droplets with different sizes and textures may coexist in the same MORB sample. Patten et al. (2012) have shown that sulfide droplets exhibiting all three textures may be present in the same sample separated by only few mm, (cf. Czamanske and Moore 1977). Patten et al. (2012) also observed a relationship between the size of the droplets and their textures. Below 30 μm, over 90% of the droplets have a fine grained texture and between 30 and 50 μm, 60% of the sulfide droplets are coarse-grained. In contrast, above 50 μm all the droplets are zoned.
Sulfide globules usually comprise fine-grained exsolution of Fe-Ni and Cu-rich sulfide phases. When the bulk compositions of sulfide are calculated to 100%, in order to estimate liquidus temperature of the MSS using the Ebel and Naldrett (1997) approach for O-free systems, they showed low variability in S content, moderate variability in Fe contents and high variability in Cu and Ni contents (Patten et al. 2012). Figure 24 shows the bulk composition of sulfide globules in terms of the system Fe–Ni–Cu. The limited field of such bulk compositions confirms the agreement between different studies (Czamanske and Moore 1977; Roy-Barman et al. 1998; Patten et al. 2012). The dashed lines in Figure 24 indicate the sulfide liquid at crystallization temperatures of the MSS at 1100, 1050, and 1000 °C from Ebel and Naldrett (1997). The liquidus temperature of the sulfide globules from MORB determined in this way, range from slightly above 1100 °C to 1030 °C where globules are randomly distributed over this temperature interval irrespective of their size or textures (cf. Patten et al. 2012).
Pentlandite occurs to a lesser extent than MSS and ISS in all textures of sulfides. Oxide also occurs either inside MSS, inside ISS or at their interface, comprising up to 7% of some sulfide globules. Oxides are best developed in zoned droplets and electron probe analyses reveal that they are Ti-free magnetite (Patten et al. 2012) in agreement with Czamanske and Moore (1977), who suggested that a few percent of magnetite is common in sulfide globules in MORB.
187Re–187Os behavior in MORB sulfide
If present, sulfide dominates the Os budget in MORB, where sulfide-silicate partition coefficients for Os in basaltic system are in the range ~104–106 (Roy-Barman et al. 1998; Gannoun et al. 2004, 2007). In contrast, Re while still being highly compatible in sulfide, has a partition coefficient at least two orders of magnitude lower than that of Os (~ 101–103; Roy-Barman et al. 1998; Gannoun et al. 2004, 2007), and similar to that of Cu (Peach et al. 1990; Gaetani and Grove 1997). As a result of the difference in partitioning of Re and Os, MORB sulfides have high Os concentrations (tens to a few hundreds of ng g−1) and low Re/Os relative to their coexisting glass (some three orders of magnitude lower). Consequently, sulfide is much less susceptible to the effects of seawater assimilation, or radiogenic in-growth, than coexisting silicate minerals or glass (Roy-Barman et al. 1998; Gannoun et al. 2004, 2007).
For those sulfides for which Os isotope and elemental abundances have been measured thus far, there is a clear covariation between 187Os/188Os and the Os concentration (Fig. 25). Where those sulfides with low Os concentrations (i.e., ≤ 10 ng g−1) possess 187Os/188Os compositions > 0.15, and those with high Os concentrations (i.e., ≥ 100 ng g−1) possess 187Os/188Os compositions around ~ 0.13 or less. This relationship might be taken to indicate that the sulfide globules, like their host glass have been systematically affected by contamination with material derived from altered oceanic crust. There is no clear relationship between the Os concentration of the sulfide and that of the host glass. However, with one exception, sulfides possess 187Os/188Os values that are less radiogenic than their glass host, where in general, the more radiogenic the host glass the greater the difference in 187Os/188Os with coexisting sulfide (Fig. 26). It is difficult to explain such a difference between sulfide and glass simply by radiogenic decay of 187Re. Rather it suggests that the 187Os/188Os composition of the glass has been more significantly affected by the assimilation of older oceanic crustal material than the coexisting sulfide.
If MORB sulfides preserve 187Os/188Os compositions that are systematically less radiogenic than their host silicate glass then this has some important implications for the timing of contamination relative to crystallization. If contamination of the silicate melt occurred before sulfide precipitation then the sulfide should possess an Os isotope composition that is indistinguishable from that of the melt. Therefore, the contrasting Os isotope composition of the glass and sulfide suggests that the silicate melt experienced contamination after the segregation of sulfide in the melt.
At the high temperatures of MORB eruption (~ 1200 °C) most sulfides will be present as liquid globules rather than as a solid phase, and diffusional equilibration between silicate and sulfide liquids is likely to be rapid. The time in which a sulfide globule will equilibrate its Os-isotope composition with a melt can be assessed using simple diffusion calculations. Using an implicit finite difference model (Crank 1975) and assuming a sulfide globule radius of 250 μm and a silicate-sulfide melt diffusion coefficient of 10−8 cm2 s−1 the sulfide will equilibrate with the melt in ~12 h (Gannoun et al. 2007). This is a relatively conservative estimate because cation diffusion in most basaltic melts is 10−5 to 10−6 cm2 s−1 (Watson and Baker 1975), whereas diffusion rates in pyrrhotite are likely to be faster than 10−9 cm2 s−1 at magmatic temperatures (Brenan et al. 2000). Therefore, under normal circumstances, complete equilibration between sulfide and glass would be expected, with both possessing an indistinguishable 187Os/188Os composition. However, because of the large concentration difference between the sulfide and the silicate liquid, a large amount of melt has to exchange with a small sulfide bleb before the sulfide reaches Os isotope equilibrium with the glass. It is possible to calculate the volume (and mass) of melt that is needed to equilibrate the sulfide using simple mass balance equations and the concentration and isotopic data for the glass and sulfides obtained here. Assuming initial 187Os/188Os for the sulfides of 0.125 and a sulfide globule radius of 250 μm, then sulfides will have only equilibrated with < 0.5 cm3 of melt (or less if the sulfide blebs were smaller). This suggests that the sulfides have only exchanged with the immediate melt surrounding the sulfide. Furthermore, a sulfide that contains > 200 ng g−1 Os would have to exchange with < 50 cm3 of melt in order to completely equilibrate with that melt. Thus, the absence of any Os isotope or elemental covariation between the sulfides and their host glass suggests that Os isotope exchange is likely to have been limited. These observations are consistent with Pd elemental data for MORB from the south west Indian ridge taken to suggest that segregated sulfides were poorly equilibrated with their host silicate magmas (Yang et al. 2013).
Nevertheless, many, if not all of the sulfides analyzed thus far are likely to have been modified by contamination, depending on their Os concentration. The sulfides with 187Os/188Os compositions > 0.13 have most likely been significantly modified through partial exchange with the contaminated silicate melt. Although those sulfides with a high Os concentration (> 20 ng g−1) may have also been affected by such exchange they do, however, yield the least radiogenic compositions yet observed in normal MORB samples.
The 187Os/188Os composition of the MORB mantle source
The MORB glass measured thus far preserves variations in 187Os/188Os extending from unradiogenic values as low as 0.125, comparable to estimates for the primitive upper mantle, to radiogenic values up to 0.25. There are no clear covariations with lithophile element isotopes, such as Sr or Nd, as might be expected from Os isotopic heterogeneity inherited for a mantle source. Rather, the radiogenic Os isotope compositions show a relationship with B isotopes that is most simply attributed to seawater-derived contamination that occurs during magma ascent. In this case, to a greater or lesser extent all MORB glass has been affected by seawater contamination. Individual sulfide grains appear to provide a much more robust record of the primary Os isotope signature (Roy-Barman et al. 1998; Gannoun et al. 2004, 2007) although even this phase appears to be susceptible to seawater contamination. In this case it is difficult to assess the extent to which any radiogenic signal, preserved in either glass or sulfide, is due to an age effect caused by 187Re decay following igneous crystallization, or the presence of Re-enriched material, such as recycled oceanic crust in the MORB source.
Assuming that the Os isotope information preserved by high-Os sulfide grains has been minimally affected by seawater contamination then they potentially provide some unique constraints on the nature of the MORB source. A fundamental assumption underlying the use of radiogenic isotopes, such as Sr, Nd and Os, in mantle derived basalts is that they are in equilibrium with their mantle source (e.g., Hofmann and Hart 1978). Abyssal peridotites are ultramafic rocks thought to represent the residue of the melting responsible for generating MORB (Dick et al. 1984; Johnson and Dick 1992; Brandon et al. 2000). Consequently during melting and basalt genesis the composition of long-lived isotopes of heavy elements in both MORB and residual abyssal peridotites should be the same. The average 187Os/188Os composition of a compilation of abyssal peridotites is 0.127 ± 0.015 (n = 129) (Fig. 27), however, like MORB, abyssal peridotites are also susceptible to seawater alteration during their exhumation on the sea floor, which may shift the composition towards radiogenic values. In this case individual abyssal peridotite sulfides are likely to yield a more reliable indication of their primary Os isotope composition, and these yield an average 187Os/188Os composition of 0.125 ± 0.021 (n = 63). The best estimate for the 187Os/188Os composition of the primitive upper mantle, that is a theoretical mantle composition with high Al2O3 that is considered to have experienced no depletion through melting, is 0.1296 ± 0.0009 (2σ; n = 117) (Meisel et al. 2001). By comparison, the high-Os (> 20 ng g−1) sulfides yield an average composition of 0.129 ± 0.005 (n = 31) with values as low as 0.1236 (Fig. 27). Therefore, these high-Os sulfides show no evidence for significant Re enrichment in the MORB source, as might accompany the presence of recycled oceanic crust. Rather they indicate that the upper mantle source of these samples has experienced a long-term depletion of Re, similar to that observed in abyssal peridotites, and consistent with the incompatible nature of this element during mantle melting.
LOWER OCEANIC CRUST
The oceanic crust comprises some 1–1.5 km of basalt and dolerite that is underlain by 4–5 km of gabbro. Therefore, MORB are thought to be evolved lavas formed by fractional crystallization in the lower oceanic crust, that itself comprises plutonic rocks and cumulates from primitive magmas. Given that Re is moderately incompatible while Os is compatible during mantle melting, one might expect that gabbros in the lower crust would have higher Os and lower Re concentrations and accordingly lower Re/Os ratios than evolved MORB, assuming that the phases that control solid/liquid partitioning of Re and Os during crystallization are similar to those involved during partial melting. Gabbroic lower oceanic crust should therefore dominate the HSE budget of the oceanic crust as whole.
However, the first reported siderophile element data for gabbros from Ocean Drilling Program (ODP) Site 735 (Blusztajn et al. 2000) yielded rather low HSE concentrations (Fig. 28)—even lower than average MORB (Bézos et al. 2005; Gannoun et al. 2007)—pointing to their evolved compositions. Indeed, Dick et al. (2000) and Hart et al. (1999) noted that the average composition of gabbro from ODP Site 735B is closer to that of average MORB (on the basis of major and trace element systematics). Consequently, the gabbro recovered at this site cannot be considered as the primitive complement to typical evolved MORB. More recently, Peucker-Ehrenbrink et al. (2012) have argued that all prior geochemical work on in situ upper oceanic crust such as DSDP-ODP sites 417, 418, and 504 (Bach et al. 2003; Peucker-Ehrenbrink et al. 2003), and 801 (Reisberg et al. 2008), and evolved gabbros at ODP 735 (Hart et al. 1999; Blusztajn et al. 2000), and site 894 (Lecuyer and Reynard 1996) failed to reproduce the true average for the complementary crustal reservoir to MORB lavas and therefore needs to be complemented with more detailed geochemical and petrologic studies of primitive gabbroic material from the lower crust.
In order to more accurately assess the global HSE chemistry of the whole oceanic crust Peucker-Ehrenbrink et al. (2012) obtained data for an oceanic crust section from the Oman ophiolite that includes the crust–mantle transition. The mean weighted composition of the 4680 m Oman section yielded Re 427 pg g−1, Os 55 pg g−1, Ir 182 pg g−1, Pd 2846 pg g−1, Pt 4151 pg g−1 and initial 187Os/188Os of 0.142, indicating higher PGE concentrations and lower Re concentrations than all data previously reported on partial sections of ocean crust that lack cumulate lower crust. Assuming that these data are truly representative of the lower oceanic crust, they suggest that these rocks are the main HSE reservoir in the oceanic crust as a whole and that the average Re in these gabbros is much lower than in MORB lavas (Re ~ 1070 pg g−1; Hauri and Hart 1997; Gannoun et al. 2007; Gannoun et al. 2016, this volume). The Oman gabbros are characterised by a distinct subchondritic average Os/Ir ratio of ~ 0.3 which is significantly different from the chondritic ratio or the primitive upper mantle value of ~ 1.1 (Becker et al. 2006; Lodders et al. 2009). This difference is surprising because Ir is generally viewed as a geochemical analogue of Os during magmatic processes (Becker et al. 2006; Puchtel and Humayun 2000). The Os/Ir fractionation observed in the Oman gabbros, while within the range observed in MORB (0.2–1.4, average 0.6), is the opposite of that observed in the upper crustal part from DSDP 504B (average Os/Ir of ~ 2.4; Peucker-Ehrenbrink et al. 2003). However the Os/Ir of abyssal peridotites in general and in the harzburgitic mantle section of Oman in particular, remains chondritic (Hanghøj et al. 2010). If such harzburgites are representative of the mantle source then the subchondritic Os/Ir ratio in Oman gabbros cannot reflect a source signature. Hanghøj et al. (2010) report both superchondritic and subchondritic Os/Ir ratios in Oman dunites (0.5–8.3). As Os and Ir alloys included in chromites have been observed in Oman dunites (Ahmed and Arai 2002; Ahmed et al. 2006), Peucker-Ehrenbrink et al. (2012) suggested that such a phase may be responsible for the fractionation of Os from Ir during melting, melt extraction or crystal fractionation.
Estimating the HSE inventory of the whole ocean crust remains challenging because of the discontinuous nature of field sampling and the question of how representative the samples that have been analyzed thus far actually are. Peucker-Ehrenbrink et al. (2012) used data from Site 504B for the upper oceanic crust (Peucker-Ehrenbrink et al. 2003) combined that for the Oman ophiolite for the lower oceanic crust (Peucker-Ehrenbrink et al. 2012). The weighted chemical and isotope characteristics of this “composite” oceanic crust (Fig. 28), corrected for Re decay since emplacement, are 736 pg g−1 Re, 45 pg g−1 Os, 133 pg g−1 Ir, 2122 pg g−1 Pd, 2072 pg g−1 Pt, 187Re/188Os: 80 and 187Os/188Os: 0.144. Such crust is more enriched in Re and less depleted in PGE than observed in average gabbros from ODP Hole 735D. Therefore, unless fundamentally altered during subduction, subducted oceanic crust will evolve to form a PGE-depleted, Re-rich mantle component that over time will evolve to radiogenic 187Os/188Os compositions. However, the projected ingrowth of radiogenic 187Os/188Os may be inhibited by the loss of Re from the basaltic upper part of the crust during eclogite-facies metamorphism (Becker et al. 2000; Dale et al. 2007), but this is not true for the gabbroic lower part of the crust (Dale et al. 2007).
Assimilation of gabbroic lower crust
Recent work has shown that the crystallization of gabbros, troctolites, and other plutonic rocks of the lower oceanic crust may be protracted, and that these rocks sometimes possess ages that are several million years older than predicted from the magnetic ages of the overlying basaltic crust (e.g., Schwartz et al. 2005; Grimes et al. 2008). This extended timescale for the growth of the lower oceanic crust has been attributed to the crystallization of gabbros in the mantle followed by uplift to lower crustal depths (Schwartz et al. 2005; Grimes et al. 2008). Such uplift may relate to unroofing by low-angle detachment faults, typical of asymmetrical spreading ridge segments (e.g., Lissenburg et al. 2009).
Over a timescale of several million years gabbros and troctolites, and their constituent phases, in the lower oceanic crust will rapidly evolve to radiogenic Os isotope compositions. This raises the possibility that younger melts passing through older lower crust may acquire a radiogenic Os isotope composition, either by remelting and assimilation of older material or through the physical entrainment of older crystals. Primitive xenocrysts are commonly found in MORB (e.g., Dungan and Rhodes 1978; Coogan 2014) with evidence for mixing shortly before eruption (e.g., Moore et al. 2014). Indeed, as discussed previously, the old ages of phenocryst phases in basalts that are thought to have been erupted just 5–10 kyr ago (Figs. 18 and 19), may indicate that these are xenocrysts physically entrained from previously solidified “olivine–plagioclase” bearing plutonic rocks through which the present host basalts have ascended. In this case, it is possible that some MORB glass may acquire a radiogenic Os isotope composition without interaction with seawater altered oceanic crust, or the presence of a radiogenic mantle source. For MORB glass such a signature might be distinguished by the absence of any covariation with Cl abundance or B isotopes.
HSE ABUNDANCES AND Re–Os ISOTOPE SYSTEMATICS OF INTRAPLATE VOLCANISM
The HSE and Re–Os systematics of intraplate volcanism were reviewed recently by Day (2013). The purpose of this section is to briefly summarize the likely origins of intraplate volcanism, based specifically upon HSE abundance and Re–Os isotope constraints, and to provide an update of developments in the field since 2013. In particular, and mostly as a function of the difficulties associated with producing precise 186Os/188Os data (e.g., Chatterjee and Lassiter 2015), there have been limited advances in the application of the Pt–Os isotope system to intraplate volcanism since Day (2013); the interested reader is referred to this earlier review article for an up-to-date appraisal of Pt–Os isotope systematics.
The origin of intraplate volcanism has been variously attributed to (i) mantle plumes (Wilson 1963; Morgan 1971), (ii) plumes which are not particularly “hot” (e.g., Falloon et al. 2007; Putirka et al. 2007), (iii) stress-driven processes (Anguita and Hernan 1975) or (iv) chemical heterogeneities preserved in the upper mantle (e.g., Courtillot et al. 2003; Arndt 2012). The occurrence of intraplate volcanism does not appear to be related to proximity to plate boundaries (cf. Hawai’i; Wilson et al. 1963 versus the Canary Islands; Morgan 1971) and does not occur systematically on either the continents or within oceanic basins, even spanning continental–oceanic margins (i.e., the Cameroon Line; Rehkämper et al. 1997; Gannoun et al. 2015a). Intraplate volcanism can be associated with convergent (e.g., Samoa; Wright and White 1987) and divergent (e.g., Iceland; Morgan 1971) tectonic settings.
In general, intraplate volcanism is controlled by anomalous thermo-chemical and/or tectonic conditions capable of producing large volumes of extrusive products. Many investigations into the HSE of intraplate volcanic rocks have predominantly featured primitive, high-MgO rocks, e.g., komatiites and picrites (e.g., Ireland et al. 2009; Connelly et al. 2011, respectively), because of the compatibility of the HSE during fractional crystallization, and the sensitivity of 187Os/188Os to crustal assimilation processes in more evolved magmas (e.g., Chu et al. 2013). However, evolved potassic and sodic mafic–alkaline volcanic rocks and phonolites, trachytes and rhyolites, which may have experienced extensive fractional crystallization, are also observed and have recently been investigated for their HSE abundances and Re–Os isotope systematics (e.g., Chu et al. 2013; Li et al. 2014; Wang et al. 2014). For this reason, we adopt the same definition used by Day (2013) for intraplate ‘hotspot’ volcanism, i.e., “Volcanic rocks that are unassociated with conventional plate tectonic boundary magmatic processes and that may require anomalous thermo-chemical and/or tectonic conditions to induce small- to large-scale melting.”
Mantle melting processes
The composition of the mantle source may be expressed by a variety of end-member compositions based upon its history of prior melt depletion i.e., depleted versus fertile peridotite (e.g., Niu 2004; Godard et al. 2008) and overall lithology, i.e., peridotite versus pyroxenite (Hirschmann and Stolper 1996; Yaxley 2000; Kogiso et al. 2004; Lambart et al. 2012, 2013). In addition, fertile heterogeneities in the mantle nucleate magmatic channels that focus melts up to the surface and hinder their re-equilibration with ambient peridotite (Katz and Weatherley 2012). Therefore, the chemical signature of hybrid melts of peridotite and pyroxenite can be retained in the composition of mantle-derived basalts. Day (2013) discussed the significance of the ‘shape’ that a melting regime can have, discussing two end-member geometries; (i) batch melting of a columnar (cylindrical) region (e.g., Rehkämper et al. 1999), and (ii) regions of adiabatic melting in triangular or corner-flow melting regime (e.g., Plank and Langmuir 1992). Each of these melting regimes aggregate melt pooled from over the melting volume, accounting for the overall composition of the magma generated.
Briefly, model (i) is most consistent with an upwelling ‘mantle plume-like’ melting regime. It assumes uniform melting throughout the source region and that the extraction of sulfide-hosted HSE is completely exhausted at 20–25% partial melting. This cylindrical melting model reproduces the HSE abundances of low-degree alkali basalts (e.g., Canary Island lavas; e.g., Day et al. 2009) and high-degree partial melts (e.g., komatiites; e.g., Rehkämper et al. 1999), but the HSE signature of some tholeiitic magmas generated by low degrees of partial melting are not predicted using this cylindrical melt volume (e.g., Momme et al. 2003, 2006). The triangular melting regime (model ii) assumes near-fractional melting in 1% increments with decreasing pressure, i.e., through adiabatic ascent (e.g., Rehkämper et al. 1999; Momme et al. 2003). In this melting regime, S-saturated low-degree partial melts with low HSE-concentrations mix with shallower, higher-degree (and potentially S-undersaturated) partial melt. Refinements to the two general classes of models described above have allowed distinct melt regimes in some continental flood basalt (CFB) provinces to be determined (Momme et al. 2006), whereas in the Icelandic rift zones depleted versus enriched mantle components have also been identified (Momme et al. 2003). Moreover, the use of these models has permitted the detection of a pyroxenitic component in primitive lavas from the Canary Islands (Day et al. 2009), and a similar component has been implicated in the generation some Hawaiian lavas (Lassiter et al. 2000; Sobolev et al. 2007).
Source compositional estimates become increasingly complicated when the necessity arises to account for the contributions from mixtures of source lithologies (e.g., peridotite and recycled sediment or basalt) and the complex interplay of the HSE that each of these source reservoirs may contribute to a pooled melt (e.g., Hirschmann and Stolper 1996).
Osmium isotopes as tracers of hotspot sources
Ocean island basalts
Many intraplate basalts retain HSE signatures of their mantle source region and osmium isotopes, when compared to lithophile element-based radiogenic isotopes, can offer a unique perspective on the petrogenesis of intraplate lavas. The large Re/Os fractionations generated during crust-mantle partitioning make it possible to model 187Os/188Os variations in OIB in the context of variably aged recycled crust and lithosphere (e.g., Hauri and Hart 1993; Marcantonio et al. 1995; Widom et al. 1999; Day et al. 2009; Day 2013). For example, ancient oceanic mantle lithosphere or SCLM has been implicated in the genesis of lavas from the Azores, Iceland and Jan Mayen (Skovgaard et al. 2001; Schaefer et al. 2002; Debaille et al. 2009), where measured unradiogenic 187Os/188Os values cannot be explained by melting exclusively of modern oceanic lithospheric material and thus require a mantle source or sources that have evolved in a low Re/Os environment (cf. unradiogenic abyssal peridotites reported by Snow and Reisberg 1995; Alard et al. 2005; Harvey et al. 2006; Liu et al. 2008; Warren and Shirey 2012; Lassiter et al. 2014). Intraplate basalts and specifically OIB, are generated from mantle sources with distinct long-term time-integrated parent-daughter fractionations of Sr–Nd–Pb–Hf isotopes (e.g., Zindler and Hart 1986; Hofmann 2003; White 2010), and also preserve a large range of 187Os/188Os compositions (e.g., Pegram and Allègre 1992; Hauri and Hart 1993; Reisberg et al. 1993; Marcantonio et al. 1995; Roy-Barman and Allègre 1995; Widom and Shirey 1996; Lassiter and Hauri 1998; Brandon et al. 1999, 2007; Widom et al. 1999; Schiano et al. 2001; Eisele et al. 2002; Schaefer et al. 2002; Lassiter et al. 2003; Workman et al. 2004; Escrig et al. 2005b; Class et al. 2009; Day et al. 2009, 2010b, 2015; Debaille et al. 2009; Ireland et al. 2009; Jackson and Shirey 2011). These signatures are only retained in instances where the melt produced at depth, albeit with ancient time-integrated compositions, and reflecting the recycling of material back into the convecting mantle (e.g., Zindler and Hart 1986), are not significantly contaminated or overprinted though interaction with the lithosphere through which these basalts necessarily transit en route to the surface. For example, in a recent study of the Louisville Seamount Chain, Tejada et al. (2015) demonstrated that OIB erupted along this chain of volcanoes reach the surface with negligible chemical interaction with the lithospheric mantle that underlies the South Pacific. Moreover, unlike the Hawaiian–Emperor Seamount chain, whose compositions are readily explained by heterogeneous mantle sources (see following section), osmium isotope signatures of these basalts have a very narrow range, consistent with their derivation from a primitive mantle source (cf. Meisel et al. 2001; Becker et al. 2006). Age corrected 187Os/188Os of the Louisville Seamount basalts range from 0.1245–0.1314, similar to other Pacific OIB, such as Rarotonga (0.1249–0.1285, Hauri and Hart (1993); 0.124–0.139, Hanyu et al. (2011) and some Samoan basalts (0.1230–0.1313, Hauri and Hart (1993); Jackson and Shirey (2011)). The age corrected 187Os/188Os for two aggregates of olivine phenocrysts separated from Louisville Seamount basalts (0.1272 and 0.1271–0.1275) agree with whole rocks from the same seamount (0.1253–0.1274; Tejada et al. 2015), supporting the hypothesis that early-crystallizing olivine can preserve the pristine magmatic Os isotopic compositions of their source (cf. Jackson and Shirey 2011; Hanyu et al. 2011) (Fig. 29).
Studies of HSE abundance complement and extend the knowledge of intraplate magma petrogenesis gleaned from Os isotope systematics. Only lavas with high MgO contents and > 0.05 ng g−1 Os should be considered as potentially being representative of the true HSE characteristics of intraplate magma and its mantle source. Such restrictions on the analysis of intraplate magmas mean that there is still a dearth of high quality HSE data on OIB. Much of what has been elucidated from HSE abundances in OIB comes from studies of Hawaiian lavas (Bennett et al. 2000; Crocket 2002; Jamais et al. 2008; Ireland et al. 2009; Pitcher et al. 2009). These studies support the hypothesis that, in general, high-MgO lavas preserve early-formed Os-rich (+ HSE) phases that become incorporated in early forming phenocrysts such as olivine (e.g., Brandon et al. 1999; Ireland et al. 2009). Removing the effects of mineral fractionation on HSE abundances allowed Day (2013) to directly compare the absolute and relative HSE abundances and calculated Re/Os of parent melts in addition to 187Os/188Os, of Hawaiian, Canary Island and Samoan lavas. Combined with the HIMU type 206Pb/204Pb compositions of Canary Island lavas, this led to the conclusion that, in contrast to Hawaiian and Samoan OIB, and komatiites, whose compositions suggest a relatively high proportion of peridotite in their parental melts, lavas from the Canary Islands, and specifically El Hierro and La Palma, contain recycled oceanic crust in their mantle source. Osmium isotope studies of HIMU-type OIB support and enhance Sr–Nd–Pb isotope and trace element arguments for a recycled oceanic lithosphere component in their mantle source (Hauri and Hart 1993; Marcantonio et al. 1995; Widom et al. 1999; Eisele et al. 2002; Day et al. 2010b). The observed range of 187Os/188Os and 206Pb/204Pb of HIMU basalts (e.g., Becker et al. 2000; Dale et al. 2009a; van Acken et al. 2010) could be produced by direct melting (~ 50% to 90%) of recycled oceanic crust, but would result in melts that contain too much silica and too little magnesium (e.g., Yaxley and Green 1998). Although field evidence suggests that pyroxenites account for ≤ 10% of mantle lithologies (e.g., Pearson et al. 1991; Reisberg et al. 1991), they melt disproportionately to peridotite under any P–T conditions (only 1–2% pyroxenite may generate up to 50% of the melt at low degrees of partial melting), thus producing silica-undersaturated, iron-rich melts with high MgO (e.g., Hirschmann et al. 2003). This means that direct melting of recycled oceanic crust and lithosphere is not necessary to produce HIMU OIB.
Spinel websterites have been suggested to be geochemically analogous to pyroxenites, at least in terms of their HSE systematics (Marchesi et al. 2014). The Re–Os fractionation generated as a result of peridotite versus pyroxenite (and/or spinel websterites) has been suggested as a likely contributor to the observed 186Os/187Os-rich compositions of some plume basalts (Luguet et al. 2008) previously attributed to interaction between the mantle and outer core (e.g., Walker et al. 1997; Brandon et al. 1998, 2003; Puchtel et al. 2005). Subsequent studies (e.g., Baker and Jensen 2004; Scherstén et al. 2004; Luguet et al. 2008) and, more recently, Marchesi et al. (2014) suggest that such enrichments could be attributed to processes requiring no input from the outer core. However, these models may require unreasonably high contributions from pyroxenitic / spinel websteritic lithologies in the mantle (as high as 90%; van Acken et al. 2010; Marchesi et al. 2014), as a result of the comparatively low Os concentrations in pyroxene-rich lithologies.
The enriched mantle (EM) signatures of other OIB has been attributed to the addition of subducted sediment or metasomatized lithosphere into their mantle sources (e.g., Workman et al. 2004). EM-type OIB span a range of compositions in Sr–Nd–Pb isotope space, varying from EMI (e.g., Pitcairn; Woodhead and McCulloch 1998; and the Comores; Class et al. 2009), which exhibit a wide range of Os- and Pb-isotope compositions, but more restricted Sr isotope compositions, to EMII OIB (e.g., Samoa; Wright and White 1987; Workman et al. 2004; Jackson and Shirey 2011). These compositions are consistent with sediment, recycled oceanic crust and peridotite producing EMI-flavoured compositions with more radiogenic 187Os/188Os (Roy-Barman and Allègre 1995; Class et al. 2009), while subducted sediment mixed with ambient peridotite produces enriched EMII compositions with lower 187Os/188Os. Therefore, lithological variations in the mantle source play a key role in the composition of OIB, and HSE abundances combined with 187Re-187Os systematics are critical in the identification of the various components mixed with variably depleted asthenospheric mantle.
Continental intraplate volcanism
The heterogeneous mantle sources described above are not restricted to OIB, or oceanic settings in general. These modifiers of magma composition also influence intraplate volcanism associated with continental regions. The main differences between oceanic and continentally erupted intraplate magmas is the greater potential for the latter to be influenced by interaction with the thicker and older overlying sub-continental lithospheric mantle (SCLM) and continental crust, in addition to the potential compositional heterogeneities within the asthenospheric mantle. Recently, Sun et al. (2014) reported Re-Os systematics of ultrapotassic (> 7 wt% K2O) basalts from the Xiaogulihe area of western Heilongjiang Province, NE China. The relatively unradiogenic Os isotope ratios (187Os/188Os = 0.1187–0.1427) contrasted with the similarly potassic basalts from NE China reported by Chu et al. (2013) (187Os/188Os = 0.13–0.17) and were attributed by Sun et al. (2014) to a dominantly peridotitic source, but one that required an unusually high K2O content. In this particular setting, phlogopite-bearing garnet peridotite hosted within the lower part of the SCLM was implicated; its derivation being potassium-rich silicate melts produced by the subduction of ancient continent-derived sediments (> 1.5 Ga). The observation that lherzolite xenoliths from Keluo and Wudalianchi contain phlogopite (Zhang et al. 2000, 2011) supports the hypothesis that SCLM, metasomatized by potassium-rich melts, is present beneath the WEK volcanic field and contributes to the basalts from Xiaogulihe.
Crustal and lithospheric mantle assimilation/contamination
Oceanic intraplate volcanism is often assumed to be immune to lithospheric contamination. Compared to continental intraplate eruptions, OIB do not interact with thermo-chemically complex SCLM. The low Os contents in OIB (typically < 1 ng g−1) makes the Re–Os isotope system a particularly sensitive indicator of lithospheric contamination, and the relatively unradiogenic 187Os/188Os compositions (< 0.18) of OIB relative to local oceanic crustal reservoirs (typically 187Os/188Os > 0.4; Reisberg et al. 1993; Marcantonio et al. 1995; Peucker-Ehrenbrink et al. 1995; Widom et al. 1999) make the tracing of assimilation of crustal or lithospheric mantle materials in OIB a straightforward process (e.g., Reisberg et al. 1993; Marcantonio et al. 1995; Lassiter and Hauri 1998; Skovgaard et al. 2001; Gaffney et al. 2005). In particular, at the lowest levels of Os content, OIB are even more vulnerable to crustal contamination (Reisberg et al. 1993), while OIB with Os contents greater than 30 to 50 pg g−1 are typically assumed to be less susceptible to assimilation of lithospheric components (e.g., Reisberg et al. 1993; Eisele et al. 2002; Class et al. 2009). Crustal contamination thus rapidly drives Os isotope ratios to more radiogenic values resulting from the assimilation of oceanic crust with high Re/Os and 187Os/188Os.
A consequence of the low HSE abundances of crustal material is that the addition of crust to a primitive melt should result in the dilution of HSE abundances in the resultant magma. Ireland et al. (2009) presented such a model, illustrating the effect of crustal contamination on Hawaiian picrites. Briefly, three end-member scenarios are considered; (i) continental crust addition to komatiite; (ii) oceanic crust addition to tholeiite and, (iii) abyssal peridotite addition to alkali basalt. These models demonstrate that crustal contamination dilutes OIB HSE abundances at ≤20 % crustal or lithospheric assimilation. However, both 187Os/188Os and Re/Os can change dramatically in the evolving liquid, which has implications for the time integrated Os isotope ratio of such contaminated magmas and the effectiveness of using 187Os/188Os as a tracer for the mantle source of the magma. The effects of assimilation on HSE abundances (absolute or relative) in general, are less well-defined and where this issue has been addressed in the literature the consensus appears to be that fractional crystallization exerts a stronger influence on HSE distributions than contamination factors (e.g., Chazey and Neal 2005; Ireland et al. 2009). However, crustal contamination of continental flood basalts (CFB) can lead to a significant augmentation in the S content of a magma, sometimes resulting in S-saturation and significant HSE fractionation (e.g., Keays and Lightfoot 2007; Lorand and Alard 2010). This may also elevate concentrations of Re and the PPGE relative to Os, Ir, and Ru. Assimilation of mantle lithosphere also has pronounced effects on Re/Os, but requires large additions to generate significant effects on magma HSE abundances. Conversely, Widom et al. (1999) demonstrated that unusually unradiogenic 187Os/188Os in some Canary Island lavas was most likely the result of the assimilation of peridotite xenoliths with sub-chondritic 187Os/188Os and >1 ng g−1 Os, prior to the eruption of the basalt at the surface. More recently, a similar process was described by Gannoun et al (2015a) to account for particularly unradiogenic Os concentrations in basalts from the Cameroon Line (Fig. 30).
These simple crustal contamination models can be greatly complicated by the inclusion of fractional crystallization processes, which are often intimately associated with crustal contamination. The combination of these processes will almost inevitably result in the generation of elevated Re, Pt, and Pd abundances compared to Os, Ir, and Ru in melts and crustal rocks, compared with their corresponding mantle residues. However, direct measurement of 187Os/188Os in early formed mineral phases handpicked from intraplate magmas, such as olivine (Debaille et al. 2009; Jackson and Shirey 2011), generally yield more restricted ranges in 187Os/188Os than their associated whole-rocks, and may provide a means of seeing past bulk-rock contamination of OIB. As a result of these potential complications, a common sense approach, based upon a rigorous assessment of local potential contaminants and melt products was advocated by Day (2013) when applying thresholds for “contaminated” versus “uncontaminated” OIB. Both crustal and SCLM contamination of primitive melts have been reported in the literature (e.g., Ellam et al. 1992; Horan et al. 1995; Molzahn et al. 1996; Chesley and Ruiz 1998; Keays and Lightfoot 2007; Li et al. 2010; Chu et al. 2013). Successful modelling of SCLM or crustal assimilation is dependent upon the accurate determination of likely end-member compositions, ranging from the parental primitive melt to its possible assimilants. Day (2013) successfully demonstrated the effect of contamination of primitive parent melts using North Atlantic Igneous Province (NAIP) picrites (Schaefer et al. 2000; Kent et al. 2004; Dale et al. 2009b) and intrusive rocks from the Rum Intrusion (O’Driscoll et al. 2009).
The origin of Continental Flood Basalts (CFB) and Large Igneous Provinces (LIP)
Volcanic rocks from some CFB have been interpreted to have survived the transit from their asthenospheric source to eruption at the surface without any significant interaction with the SCLM or the crust (e.g., Schaefer et al. 2000; Zhang et al. 2008; Dale et al. 2009b; Rogers et al. 2010; Day et al. 2013). Many of these lavas are picritic in composition, have high-MgO (> 13.5 wt%), high Os concentrations, and 187Os/188Os which are, in general, unradiogenic; consistent with their derivation from primitive mantle or a depleted mantle source (e.g., Schaefer et al. 2000; Dale et al. 2009b; Rogers et al. 2010). This chemical and isotopic signature has, in turn, been used to suggest that such CFB may be modern-day equivalents of uncontaminated Archaean komatiites, albeit from a cooler mantle, (cf. Brügmann et al. 1987; Wilson et al. 2003; Puchtel et al. 2009; Connolly et al. 2011).
In contrast, several studies have highlighted the importance of an interaction between asthenosphere-derived melts, SCLM and the crust to produce the observed spectrum of CFB compositions (e.g., Ellam et al. 1992; Horan et al. 1995; Molzahn et al. 1996; Chesley and Ruiz 1998; Xu et al. 2007; Li et al. 2010; Heinonen et al. 2014) and the HSE fingerprint of some komatiites (Foster et al. 1996). Osmium isotope systematics, combined with other radiogenic isotope tracers in CFB demonstrate that the interplay between a primary magma and its potential lithospheric contaminants can be complex, as illustrated in a number of localities (e.g., Siberia–Horan et al. 1995; Ethiopia–Rogers et al. 2010; Emeishan, China,–Zhang et al. 2008). Correlations between 187Os/188Os and 87Sr/86Sr (Molzahn et al. 1996) 206Pb/204Pb (Xu et al. 2007), and possibly even 3He/4He (Dale et al. 2009b) illustrate the effects of lithospheric contamination on primary, asthenosphere-derived melts. However, observed variations in Os isotopes are not wholly consistent with SCLM or crustal contamination alone, suggesting that, like many OIB, some inherent heterogeneity within the asthenospheric source is present. For example, the 260 Ma Emeishan province (e.g., Li et al. 2010) requires a more depleted mantle source than 190 Ma Karoo CFB (Ellam et al. 1992). Some CFB provinces may therefore tap mantle sources that contain recycled material, similar to the source of some HIMU and EM flavoured OIB (e.g., Shirey 1997; Dale et al. 2009b), while others are derived from an essentially primitive mantle (see review in Day 2013).
Heterogeneity in the composition and distribution of sulfide types within a magma source region in the mantle (e.g., interstitial versus enclosed sulfides; Alard et al. 2002; see also Harvey et al. 2016, this volume) can have a profound influence on the composition of a basaltic melt (e.g., Harvey et al. 2010, 2011). The combination of source heterogeneity and degree of partial melting can therefore account for the observed differences in initial Os isotopic and HSE abundance variations in CFB provinces, that range from depleted DMM-like mantle compositions (e.g., Rogers et al. 2010) through undepleted basalts (e.g., Schaefer et al. 2000), to more radiogenic compositions, which provide strong evidence for recycled components in some CFB provinces (Shirey 1997)
Coupled with the effects of adding subducted oceanic lithosphere back into the convecting mantle, i.e., the source of CFB and LIP, and the combinations of pelagic / terrigenous sediments, variably altered oceanic crust and serpentinized peridotite (Allègre and Turcotte 1986), unravelling the sources of voluminous basaltic magmatism has somrtimes been demonstrated to be problematic, often requiring both HSE and Re–Os isotope evidence used in concert with more traditional lithophile element-based isotope systems. For example, Heinonen et al. (2014) invoked a mixture of depleted Os-rich peridotite with ~10–30% of seawater-altered and subduction-modified MORB (with a recycling age of less than 1.0 Ga) as the likely source of the distinctive isotopic fingerprint found in CFB from the Antarctic Karoo province. A specific mixed peridotite-pyroxenite-like source was required to explain the unusual combination of elevated initial 87Sr/86Sr and Pb isotopic ratios, and low initial 187Os/188Os observed in the dykes sampled from around Ahlmannryggen, western Dronning Maud Land. In other words, simple, two component mixing is often not consistent with the observed chemical and isotopic composition of CFB. In the example described by Heinonen et al. (2014), not only was a combination of mixed lithologies in the source, in addition to the inherent differences in their HSE and 187Os/188Os fingerprints required to account for the composition of the Ahlmannryggen dykes, but also a contribution from a seawater-altered subducted component was required.
A similar investigation into the nature of the Eastern North America (ENA) Central Atlantic Magmatic Province (CAMP) by Merle et al. (2014) also revealed the complex combination of chemical and isotopic fingerprints that can be preserved in large-volume basaltic eruptions. Although CAMP magmatism in general may have been produced as a result of either heat incubation under thick continental lithosphere (McHone 2000; De Minet al. 2003; Puffer 2003; McHone et al. 2005; Verati et al. 2005; Coltice et al. 2007), or by a plume head under the continental lithosphere (May 1971; Morgan 1983; White and McKenzie 1989; Hill 1991; Wilson 1997; Courtillot et al. 1999; Ernst and Buchan 2002; Cebria et al. 2003), Merle et al. (2014) proposed several increasingly complex scenarios to account for the chemical and isotopic signatures preserved in the ENA CAMP basalts, including (i) direct derivation from a mantle plume (Wilson 1997) or oceanic plateau basalt-type melts (e.g., Kerr and Mahoney 2007); (ii) magmas derived from a mantle plume but contaminated by continental crust en route to the surface (Arndt al. 1993); (iii) mixing between asthenospheric and ultra-alkaline mafic (lamproite, kimberlite, and kamfugite) melts (Arndt and Christensen 1992; Gibson et al. 2006; Heinonen et al. 2010), possibly followed by crustal contamination; (iv) ternary mixing between OIB, MORB and SCLM-related melts, possibly followed by crustal contamination; (v) direct melting of a shallow source enriched in incompatible elements such as metasomatized SCLM or the mantle wedge above subduction zones (Puffer 2001; De Min et al. 2003; Deckart et al. 2005; Dorais and Tubrett 2008). Unfeasibly large degrees of crustal contamination would be required to produce the observed 143Nd/144Nd, 206Pb/204Pb and 208Pb/204Pb isotopic compositions of the ENA CAMP basalts, and crustal contamination, assimilation (of continental crust) with fractional crystallization (DePaolo 1981) and assimilation through turbulent ascent were discounted on the strength of the Re-Os and 187Os/188Os systematics i.e., initial 187Os/188Os higher than 0·15 at Os concentrations lower than 50 ng g−1 (e.g., Widom 1997).
Merle et al. (2014) determined that mixing involving either OIB or MORB-like parental melts, followed by crustal contamination, partially reproduces the compositions of the ENA CAMP basalts, but the trends observed in the Nd–Pb and Os–Nd isotopic diagrams require the addition of up to 35% continental crust, yet the assimilation of more than 20% of continental crust is thermodynamically unrealistic (Spera and Bohrson 2001). Consequently, the hypothesis of a magma originating from mixing between OIB and SCLM-related melts and further contaminated by the continental crust was deemed unlikely. Therefore, the continental crust-like characteristics of the ENA CAMP were inferred to be present in the mantle source itself. Recent studies have suggested that such contrasting chemical characteristics may be derived from a metasomatized SCLM-type source (cf. Chu et al. 2013; Sun et al. 2014; Wang et al. 2014), where phlogopite in the SCLM was thought to be derived from the melting of subducted terrigenous sediments. To account for the measured Os isotope compositions of the ENA CAMP basalts, the Os isotopic composition of the source needed to be within the range of 187Os/188Os for off-cratonic SCLM (0·1180–0·1290; Carlson 2005), therefore the model favored by Merle et al. (2014) to explain the multi-isotope system fingerprint of the EMA CAMP basalts required a reservoir that experienced progressive incorporation of subducted sediments derived from the local continental crust into a depleted sub-arc mantle wedge above a subduction zone.
Recent work has revealed that HSE abundances can be broadly modelled as a function of fractional crystallization in CFB. Day et al. (2013) studied the 1.27 Ga Coppermine CFB in northern Canada, which represents the extrusive manifestation of the Mackenzie large igneous province (LIP), which includes the Mackenzie dyke swarm and the Muskox layered intrusion. These authors reported new HSE abundance and Re-Os isotope data for picrites and basalts from the CFB, as well as a highly unusual andesite glass flow. The glass contained high HSE contents (e.g., 3.8 ng g−1 Os) and mantle-like initial 187Os/188Os (γ1270Ma Os = +2.2), but δ18O, eNdi, and trace element abundances consistent with extensive crustal contamination, implicating a potential origin for this sample (CM19) as a magma mingling product formed within the Muskox Intrusion during chromitite genesis (cf. Day et al. 2008) and direct evidence for the processing of some CFB within upper-crustal magma chambers. These authors also modelled absolute and relative HSE abundances in CFB from the Coppermine, Parana and West Greenland, revealing that HSE concentrations decrease with increasing fractionation for melts with < 8 ± 1 wt% MgO (Fig. 31). The models reveal that significant inter-element fractionation between (Re + Pt + Pd)/(Os + Ir + Ru) are generated during magmatic differentiation in response to strongly contrasting partitioning of these two groups of elements into sulfides and/or HSE-rich alloys. Furthermore, fractional crystallization has a greater role on absolute and relative HSE abundances than crustal contamination under conditions of CFB petrogenesis due to the dilution effect of low total HSE continental crust. Day et al. (2013) found that picrites (> 13.5 wt% MgO) from CFB (n = 98; 1.97 ± 1.77 ppb) having higher Os abundances than OIB picrites (n = 75; 0.95 ± 0.86 ppb) and interpreted these differences to reflect either higher degrees of partial melting to form CFB, or incorporation of trace sulfide in CFB picrites from magmas that reached S-saturation in shallow-level magma chambers.
Continental intraplate alkaline volcanism
Continental intraplate alkaline volcanic rocks (CIAV) comprise a wide spectrum of sodic and potassic compositions ranging from alkali basalts, picrites and basanites through to more evolved eruptive products that include nephelinites, carbonatites, melilitites, and kimberlites. The origin of some of these rock types are not unequivocal, with petrogenetic models ranging from pure incipient rift-related sources (e.g., Thompson et al. 2005), to ‘hotspot’ or ‘plume’ related origins (e.g., Haggerty 1999). Finding a likely source for these volcanic rocks is not made any less ambiguous when experimental and geochemical data are considered as many of these lavas are thought to derive from close to the boundary layer that separates the convecting and conducting mantle (e.g., Foley 1992; Day et al. 2005), i.e., both the asthenosphere and SCLM can be implicated in the genesis of these magmas. Re–Os isotope data are limited for these types of lavas, and instances where this is combined with HSE abundance data are comparatively rare. Examples from the literature when HSE and / or Re–Os isotope data are available are summarized in Day (2013)
When elevated osmium contents in basalts clearly exclude the influence of crustal contamination, radiogenic 187Os/188Os (e.g., > 0.15) is often interpreted as being derived from olivine-poor mantle heterogeneities, such as clinopyroxenites (Carlson et al. 1996; Carlson and Nowell 2001; Janney et al. 2002), primarily as a result of their time-integrated ingrowths to high 187Os/188Os (Reisberg et al. 1991; Reisberg and Lorand 1995; Kumar et al. 1996). At the onset of S-saturated melting at depth, these fertile heterogeneities with radiogenic Os isotopic compositions melt preferentially (Hirschmann et al. 2003; Rosenthal et al. 2009). Combined with the Os isotope and HSE signature associated with pyroxenite-dominated melts, high NiO and low MnO concentrations in olivine phenocrysts are also diagnostic of olivine-poor mantle domains such as phlogopite-rich pyroxenites (Prelević et al. 2013). These phlogopite-bearing pyroxenites can be derived from the reaction of peridotitic mantle wedge with melts derived from terrigenous sediments, possibly from the uppermost regions of the subducting slab (Prelević et al. 2015). As such, many CIAV appear to have non-peridotitic sources, with some sodic mafic-alkali magmas possessing radiogenic 187Os/188Os compositions, but moderately high Os contents (> 0.5 ng g−1 Os). Extreme Os isotopic compositions could reflect low degrees of partial melting and preferential sampling of more fusible mafic components, such as pyroxenite, in the asthenospheric mantle (cf. CFB above). Alternatively, melting of metasomatized lithosphere during rifting events (e.g., Carlson and Nowell 2001; Thompson et al. 2005) may also be responsible for the HSE abundances and Re–Os systematics of some CIAV, such as the Newer volcanic rocks, Australia (Vogel and Keays 1997). Similarly, carbonatites may also ultimately originate from mafic as opposed to ultramafic sources due to their close association with other ultrapotassic rocks (e.g., Gudfinnsson and Presnall 2005). For example, young (> 20 Ma) carbonatites from Fuerteventura, Canary Islands, possess low Os abundances (5–15 pg g−1) and highly radiogenic 187Os/188Os that extend to values in excess of 0.6 (Widom et al. 1999). Conversely, the high Os abundance and unradiogenic Os isotope signatures of some kimberlites and katungites are consistent with a petrogenesis involving the assimilation or derivation from the SCLM (Pearson et al. 1995; Carlson et al. 1996; Araujo et al. 2001; Carlson and Nowell 2001; Pearson et al. 2008). More recently, Chalapathi Rao et al (2013) provided strong evidence for contrasting mantle sources for kimberlites and lamproites in the Eastern Dharwar craton, southern India. Re–Os isotope of orangeites from the Bastar craton and Mesoproterozoic kimberlites and lamproites contrasted with an unradiogenic Re-depleted kimberlite sample with present-day 187Os/188Os (0.1109) and a Re–Os isotopic fingerprint characteristic of Proterozoic lithosphere, with the positive γOs (2.9–3.6; where γOs refers to the percentage deviation at the time of emplacement of the 187Os/188Os from Primitive Upper Mantle with a 187Os/188Os of 0.1296; Becker et al. 2006)) of two kimberlites from Raichur and Narayanpet (Eastern Dharwar craton) that retained both both plume and subduction-related source signatures (cf. Heinonen et al. 2014 for the petrogenesis of continental flood basalts from the Antarctic province of the Karoo). The enriched Re/Os mantle sources for the nearby Kodomali orangeite (γOs = +3) and the Krishna lamproites, with very radiogenic (γOs of + 56 to + 355), similar to those displayed by the lamproites of the Italian peninsula (Conticelli et al. 2007), suggest a subducted component for the latter ultra-potassic rocks, demonstrating the complex interplay of likely sources contributing to magma genesis around the Eastern Dharwar craton in both time and space (Chalapathi Rao et al. 2013).
The low Os concentrations of primary low-degree potassic and sodic mafic–alkali volcanic rocks, combined with the high Os abundance of mantle and crustal xenoliths in some kimberlites, alnoites and melnoites make these volcanic rocks highly susceptible to contamination as they pass through and interact with the SCLM and overlying crust. Evolved magmas of this type may also be susceptible to the effects of S-saturation prior to eruption (Vogel and Keays 1997), i.e., they may have experienced the prior precipitation of sulfide and concomitant harvesting of HSE from the S-saturated magma. Despite these caveats, some continental intraplate magmas still retain unique information on the composition of their mantle source. In particular, early Cretaceous alkaline picrites and basalts from the North China craton have petrological and Os–Sr–Nd isotope compositions consistent with contributions from recycled and foundered eclogitic lower continental crust (Gao et al. 2008). More recently, Chu et al. (2013) examined a suite of highly potassic basalts from Wudalianchi-Erkeshan, NE China and, despite the incorporation of modest amounts of continental crust and the potential of sulfide contamination derived from the SCLM, traced the source of the basalts back to the asthenosphere. Their findings suggested a complex interaction between crust and SCLM with highly potassic melts generated at least partly from SCLM containing phlogopite, itself with an ancient terrigenous sediment signature (Sun et al. 2014). In contrast to a predominantly peridotitic phlogopite-bearing source for continental volcanism reported by Sun et al. (2014), Miocene ultrapotassic rocks within the Sailipu area of the western Lhasa terrane, southern Tibet, were variously attributed to the interaction of both spinel- and garnet-lherzolite derived melt with a phlogopite-bearing pyroxenite source (Wang et al. 2014). Although the latter study postulated that the observed chemistry of the ultramafic melts could be attributed to crustal contamination, unfeasibly large-scale assimilation of continental crust would be necessary to account for the nature of the Sailipu basalts. While the lithophile element-based isotope systems are relatively insensitive to crustal contamination, mixing calculations using HSE concentrations and 187Os/188Os of primitive arc compositions (Os = 0.2 ng g−1; 187Os/188Os = 0.125; Shirey and Walker 1998; Suzuki et al. 2011), continental crust (Os = 0.01 ng g−1; 187Os/188Os = 1.10; Shirey and Walker 1998) and depleted mantle material (Os = 0.405 ng g−1; 187Os/188Os = 0.10815; Shirey and Walker 1998) demonstrated that the composition of the samples from western Lhasa (Wang et al. 2014) would require an unreasonably high degree of crustal contamination (> 80%) (Fig. 32). Two other studies of ultrapotassic rocks from Italy and the Balkans (Conticelli et al. 2007; Prelević et al. 2015, respectively) attributed a similar combination of mantle sources (as opposed to crustal contamination) as being primarily responsible for the observed chemical and isotopic compositions.
The recent study of Chu et al. (2013) also discussed the complex chemical and isotopic signatures preserved in the Wudalianchi-Erkeshan highly potassic basalts in the context of crustal and lithospheric contamination. Here, the range of 187Os/188Os in basalts (187Os/188Os = 0.1187–0.17) was partially attributed to 2–8 % crustal contamination; a degree of assimilation that otherwise would be difficult to detect using lithophile element isotope systems. In fact, Gannoun et al. (2015a) suggested that degrees of crustal assimilation of up to 15 % would have no measureable effect on Nd and Pb isotope ratios of basalts, while Li et al. (2014) commented that lithophile element-based isotope systems may be opaque to as much as 18 % crustal contamination. In the latter study, high NiO and SiO2 contents, but low MnO, CaO, MgO, and Pb contents, in addition to radiogenic 187Os/188Os, low Os abundances (5–43 ng g−1) and high, but variable, Re/Os (3–126) of intra-continental OIB-like basalts from West Qinling, central China, were attributed to crustal contamination on the strength of the sensitivity of Os isotope systematics to the incorporation of continental crust.
In contrast, the most unradiogenic Os isotope signatures observed in CIAV may have been affected by the assimilation of xenocryst-hosted primary sulfide. The often unradiogenic 187Os/188Os and high (> μg g−1) Os content of sulfides enclosed within olivine xenocrysts (Alard et al. 2002) are prime candidates for the source of a possible “nugget effect”. For example, a 20 μg mantle sulfide with an Os concentration of 20 μg g−1 (see Alard et al. 2000, 2002; Pearson et al. 2002; Harvey et al. 2006, 2010, 2011; Lorand et al. 2013; Harvey et al. 2016, this volume for typical sulfides) contains twice as much Os as 2 g of basalt with an Os concentration of 100 pg g−1. This type of nugget effect was attributed by Chu et al. (2013) as being responsible for the poor reproducibility of 187Os/188Os in two Wudalianchi-Erkeshan basalts (LHS-6 and HSS-6). In this instance, the heterogeneous distribution of a component that contains anomalously high Os (+ PGE) abundances throughout the sampled rock powder could account for the observed heterogeneities in replicate basalt analyses. A similar source of heterogeneity was suggested by Gannoun et al. (2015a) to account for comparable unradiogenic 187Os/188Os signatures in some Cameroon Line basalts.
Processes affecting the HSE compositions of sub-aerial volcanism
The previous sections demonstrate that it is essential to consider the many possible source and contamination factors that may influence the ultimate composition of intraplate magmas. Irrespective of the tectonic setting in which an erupted magma was generated, sub-aerially erupted lavas may be subject to an additional group of processes whose affects need to be assessed prior to interpretations concerning magma sources and potential contaminants. These processes, including post-emplacement alteration and magmatic degassing, were reviewed comprehensively in Day (2013). While there has been a dearth of new data in the intervening period, one study in particular merits attention; the recent examination of Os loss through magmatic degassing at Piton de la Fournaise, Réunion Island (Gannoun et al. 2015b).
Oceanic island basalts have lower Re concentrations than MORB. This is anomalous considering the incompatible behavior of Re during basalt petrogenesis (Hauri and Hart 1997). This apparent quirk has been attributed to two possible causes; (i) the presence of garnet and/or sulfide in their mantle source (Righter and Hauri 1998), or (ii) magmatic degassing of Re (Bennett et al. 2000; Lassiter 2003; Norman et al. 2004). Several lines of evidence support the idea that Re loss is a late and shallow stage process, which favors process (i) above. For example, an increase in oxygen fugacity promotes the loss of Re from Re metal (Borisov and Jones 1999), suggesting that at the oxidation state relevant to OIB (FMQ), the rate of Re loss from a magma will increase by an order of magnitude per log unit of fO2 increase. Sub-aerial eruptions from Réunion and Hawaii preserve evidence for an increase in fO2 in the lavas during emplacement, from FMQ – 1.8 close to eruption vents, to up to FMQ + 3 in lava samples that have travelled several km and cooled slowly (Rhodes and Vollinger 2005; Boivin and Bachélery 2009).
Although Re and Os have the highest elemental condensation temperature (1821 and 1812 K, respectively; Lodders 2003), these elements are commonly enriched in volcanic gas sublimates and aerosols (Crocket 2000; Yudovskaya et al. 2008; Mather et al. 2012). However, the relative and absolute volatilities of Re and Os, and hence the degree of degassing from sub-aerial lavas, are not well constrained. The propensity for an elemental species to be volatilized post-eruption can be described in terms of an emanation coefficient, (Ex), where Ex = (Ci – Cf) / Ci, (Ci = concentration of element x in the magma and Cf = concentration of element x in the magma after degassing; Gill et al. 1985; Lambert et al. 1986). The emanation coefficient of Re ranges from 0.12 (Rubin 1997) to as high as 0.74 Norman et al. (2004). The difficulties associated with the analysis of pg g−1 quantities of Os in basalts make the emanation coefficient of Os even less well known.
In their recent study, Gannoun et al. (2015b) investigated the Re–Os isotope and elemental systematics of basaltic lavas and gas condensates (a range of Na–K–Ca–Cu sulfates, Ca–Mg–Al–Fe fluorides, and native sulfur) produced during eruption and degassing at Piton de la Fournaise, Réunion Island, in order to examine the geochemical behavior of these two elements during magma degassing. High temperature (> 350 °C) deposits were enriched in Re (24–79 ng g−1), almost two order of magnitude higher than the corresponding lavas (0.130–0.137), while the Os abundances of the high temperature condensates were similar to those of the lavas (14–132 pg g−1). The highest temperature condensates (Na–K sulfates; 384 to 400 °C), yielded 187Os/188Os that were significantly lower (i.e., 0.124–0.129) than their corresponding lava. These unradiogenic osmium isotope ratios were attributed by Gannoun et al. (2015b) to the volatilization of Os originally contained in old, unradiogenic mantle sulfides. Sulfides associated with earlier volcanic eruptions at Réunion Island (< 7 Ma) were deemed too young to provide the distinctive unradiogenic Os fingerprint of the volcanic gas, leading Gannoun et al. (2015b) to infer that the observed unradiogenic Os was ultimately derived from a mantle source. In the context of osmium mantle geochemistry, loss of unradiogenic Os during magmas degassing could help to explain osmium isotope disequilibrium between lavas and melting residues.
This contrasted with the 187Re–187Os systematics of the low-to-medium temperature condensates, which contained the highest Os abundances (13–77 ng g−1) with unfractionated 187Os/188Os (0.130–0.135), which are indistinguishable from the April 2007 lava flow and the historical lavas of Piton de la Fournaise (i.e., 187Os/188Os = 0.130–0.137; Schiano et al. 2012). In addition, very high concentrations of iridium (1–8 ng g−1) reported for hieratite condensates (K2SiF6) suggested that Ir was also transported in volatile emissions as gaseous IrF6 (cf. Toutain and Meyer 1989). The selective enrichment of HSE demonstrates their potential for transport as metallic hexafluorides (Molski and Seppelt 2009; Craciun et al. 2010; Gannoun et al. 2015b; see also review in Day 2013). The absence of isotopic fractionation between gas deposits and lavas also indicates that external components (such as seawater, rainwater or air), which all possess particularly radiogenic 187Os/188Os (Levasseur et al. 1998, 1999; Gannoun et al. 2006; Chen et al. 2009) have no significant influence on the Os budget of volcanic gases.
HIGHLY SIDEROPHILE ELEMENT SYSTEMATICS OF ARCS
Highly siderophile element abundance studies have been applied to arc volcanism to understand both subduction processes and the generation of economic deposits of precious metals within arc settings. A critical question has regarded the potential mobility of the HSE in subduction zone environments and the collateral effects such processes have regarding the siderophile element budget of the mantle. Fractionation of Re and Pt from Os in subduction zone environments could have a potentially significant effect both on Os isotope signatures at arcs (e.g., Brandon et al. 1996), but also on the long-term Re/Os and Pt/Os fractionations observed in OIB, MORB and mantle rocks. In addition, the potential mobility of HSE in subduction zone environments has important implications regarding the formation of economic PGE ore deposits such as major epithermal gold deposits associated with some volcanic arcs (e.g., McInnes et al. 1999). For the purpose of this review, we focus on the petrogenetic implications of arc volcanism.
Arc magmatism includes tholeiitic to calc-alkaline compositions and dominantly involves the generation of basalt-andesites, andesites and more evolved magma-types. Only a limited number of arc volcanoes are known to erupt lavas approaching basaltic or picritic compositions. Because the HSE are typically compatible during mantle melting, as well as during fractional crystallization, this means that Os concentrations in arc volcanic rocks are typically very low, resulting in increased susceptibility of arc lavas to crustal contamination (e.g., Lassiter and Luhr 2001; Hart et al. 2002; Righter et al. 2002; Turner et al. 2009; Bezard et al. 2015). High 187Os/188Os in arc lavas has therefore been attributed to assimilation of arc crust during magmatic ascent, but also due to enrichment in radiogenic Os due to contamination of the mantle wedge by slab-derived fluids/melts (e.g., Alves et al. 1999, 2002; Borg et al. 2000), or a combination of these processes (Suzuki et al. 2011). In this section, we review the work done so far in arcs, using both lavas, as well as mantle-derived xenoliths erupted in association with active arcs. Since the behavior of the HSE are reviewed extensively elsewhere in this volume, the focus of this section is largely on the information that can be obtained from the HSE regarding arc processes.
HSE and 187Os/188Os in arc lavas
The majority of arc related volcanism is located around the Pacific ‘Ring of Fire’, extending from the southern tip of Chile, up much of South and North America, into the Aleutians and Kamchatka, through Japan and down as far as the Tonga Trench and New Zealand. Other significant arcs include the Lesser Antilles Arc and the Scotia Arc (Fig. 33). Despite the extensive distribution of arc volcanoes, limited work has been conducted on Re-Os isotopes in arc volcanic rocks, primarily due to the limited availability of high MgO lavas, which are normally favored for study by Os isotope and HSE abundance studies. High MgO lavas do occur in some arc settings, most notably Grenada, south Lesser Antilles Arc, and as boninite occurrences. These lavas are discussed in detail, below.
Work on arcs has shown that arc volcanic rocks typically contain between 0.00005 and 1 ng g−1 Os and 0.01–1 ng g−1 Re (Fig. 34). Rhenium concentrations generally increase with decreasing MgO in arc lavas, consistent with moderate incompatibility of Re. However, Re can also behave as a volatile element during oxidizing conditions in arc lavas, and for this reason it is likely that low concentrations could reflect loss of Re by this process (e.g., Righter et al. 2008). Positive correlation between Os and MgO is consistent with strong compatibility of Os during fractional crystallization of arc lavas. The low MgO and HSE contents in arc lavas can make them potentially highly susceptible to crustal contamination effects (cf. Lassiter and Luhr 2001). Osmium isotopic ratios in recently erupted arc lavas can span an extreme range, from high MgO lavas with 187Os/188Os (~0.1268–0.128) similar to typical mantle estimates, to andesites, rhyolites and dacites with 187Os/188Os > 1. There is an overall relationship of increasing 187Os/188Os with decreasing Os content, although more than one trend has been recognized in plots of reciprocal Os versus 187Os/188Os (Fig. 35). Alves et al. (2002) pointed out that initial Os isotopic ratios are positively and systematically correlated on 187Os/188Os versus reciprocal Os plots, reflecting binary mixing processes, with a common end-member represented by upper mantle peridotite compositions
To date, no study has found clear associations of Re or Os contents and 187Os/188Os with arc basement type, convergence rate or sediment supply (Table 1). This may be partly due to the lack of available high MgO rocks with which to make cross-comparison of ‘primary magmatic composition’. For example, Lassen Peak lavas with 8–11.1 wt% MgO have up to 0.37 ng g−1 Os and span a range of 187Os/188Os from 0.1289–0.235 (Borg et al. 2002). It has been suggested that these lavas contain a contribution of radiogenic Os from the subducting slab. Conversely, Grenada picrites and basalts (10.5–17.4 wt% MgO) contain up to 0.36 ng g−1 Os and have a slightly more restricted range of 187Os/188Os (0.1268–0.1644), yet these lavas are not considered to have a contribution from the slab, but instead have experienced various levels of crustal assimilation (Woodland et al. 2002; Bezard et al. 2015). Likewise, boninite (13 wt%) and some low MgO lavas (< 1.5 wt%) from the Tonga-Kermadec arc have 187Os/188Os of 0.1275–0.1283, indicating that more radiogenic values for lavas in this arc are consistent with localized arc contamination (Turner et al. 2009). Unique to that study is that the sample with the least radiogenic Os signature is a dacite, suggesting that evolved magmas can develop by fractionation from mantle-derived magma with minimal interaction with high Re/Os arc crust.
Contents of the HSE in arc-related lavas have been reported for Grenada basalts and picrites, Izu-Bonin lavas (Woodland et al. 2002) and Lihir lavas (McInnes et al. 1999) (Fig. 36). These generally high MgO lavas show similar Re and PPGE enrichment over the IPGE, to many intraplate tholeiites and alkali basalts (e.g., Day 2013). However, despite the picritic (MgO > 13·5 wt%) nature of Grenada lavas, they contain low concentrations of the HSE (< 0·2 ng g−1 Ir, 1–4 ng g−1 Pd) compared with lavas from other settings of similar MgO content (see section on continental flood basalts). Woodland et al. (2002) argued that this was probably due to a combination of lower degrees of partial mantle melting and early removal of PGE with cumulus phases such as olivine, magnetite and sulfide. Comparison of alkali Grenada lavas with boninitic Izu–Bonin lavas illustrates that although the major element chemistries of Grenada and Izu–Bonin are different, relative and absolute abundances of the IPGE and PPGE are similar. Rhenium, however, is markedly depleted in the Grenada picrites compared with the Izu–Bonin boninites, suggesting either retention of Re by residual garnet in the Grenada sub-arc mantle wedge (Woodland et al. 2002) or volatile-loss of Re. In both cases, their generation above a subduction zone did not appear to have any significant systematic effect on the HSE signatures of resultant lavas.
HSE and 187Os/188Os in arc xenoliths
Studies of mantle xenoliths from arc settings have provided the opportunity to document the behavior of the HSE during slab fluid-induced metasomatism of the mantle wedge, with spinel harzburgite, websterite and pyroxenite mantle xenoliths occurring in back-arc environments in a number of arcs. Relatively radiogenic Os isotope signatures in mantle xenoliths and mantle rocks from arc settings, including the Cascades, Canadian Cordillera, Japan, Lihir, Papua New Guinea, Kamchatka, and the Catalina Schist have been documented, and attributed to the mobility of Os in slab fluids (Brandon et al. 1996, 1999; McInnes et al. 1999; Peslier et al. 2000; Widom et al. 2003). For example Simcoe xenoliths, which represent fragments of mantle lithosphere from the back-arc of the Cascade arc front, have been metasomatized by silica-rich fluids or hydrous melts leading to higher fO2 leading to radiogenic Os isotopic compositions being imparted to these peridotites (Brandon et al. 1996, 1999). These features are consistent with part or the entire metasomatic agent being derived from the Juan de Fuca slab. Studies of Kamchatka peridotites also indicate metasomatism of the Kamchatka sub-arc mantle wedge by radiogenic slab-derived fluids and melts (Widom et al. 2003).
The HSE patterns of the arc-related mantle xenoliths are broadly similar to typical oceanic mantle xenoliths (Fig. 36), but the xenoliths can often exhibit elevated 187Os/188Os, with Simcoe xenoliths ranging from 0.1226–0.1566 and Kamchatka xenoliths ranging from 0.1232–0.1484. The regional variations in Re–Os isotope signatures are consistent with previous petrographic and geochemical studies of the Kamchatka mantle xenoliths that reveal multistage metasomatic histories resulting from interaction of the mantle wedge with a variety of slab-derived fluids and melts, including silicic slab–melt metasomatism associated with subduction of relatively hot, young (~15–25 Ma) oceanic crust in the northern arc front, hydrous slab-fluid metasomatism associated with subduction of colder, old (~100 Ma) oceanic crust in the southern arc front, and carbonate-rich slab-melt metasomatism in the southern segment behind the arc front, where the slab is deeper. Similar ranges of Re–Os isotope signatures in peridotites from Avachinsky, Japan and Lihir, and from Valovayam and the Cascades, respectively, suggest that the age (temperature) and depth of subducting oceanic crust influences the Re–Os composition of metasomatized sub-arc mantle.
Radiogenic Os from slab components or from crustal contamination
A continuing debate exists over the influence of slab-derived 187Os/188Os to arcs, versus the potential for crustal or seawater contamination of magmas with low Os abundances. From Lassen lavas, Borg et al. (2000) showed that crustal contamination could only explain the Re-Os isotope systematics if distribution coefficients for Re in sulfide were ~40–1100 times higher than published estimates, and instead argued for contributions from a highly radiogenic Os slab component (187Os/188Os up to 1.4). Alves et al. (2002) also favoured slab components adding radiogenic Os to arcs, citing evidence from arcs worldwide for different mixing systematics between mantle peridotite and variably radiogenic Os slab contributions. Conversely, Bezard et al. (2015) have shown that Grenada picrites with radiogenic 87Sr/86Sr (0.705) have 187Os/188Os (0.127) that overlap with the mantle range and that assimilation and fractional crystallization can explain compositions of Lesser Antilles lavas, without the requirement of a slab input (Fig. 37). Dreher et al. (2005) studied Os isotopes in Mindanao adakites, showing that the majority of these rocks had unradiogenic Os isotopes, inconsistent with the idea that adakites with high Sr/Y and low Y and heavy rare earth element concentrations, reflect melting of young subducted crust in subduction zones.
On the other hand, the range in Os isotopes in Mexican Volcanic Belt rocks, which represent subduction-related calc-alkaline and lamprophyric rocks in which high fO2 precludes sulfide fractionation, could be explained by up to 12% assimilation and fractional crystallization (Lassiter and Luhr 2001). To obviate potential issues of shallow-level crustal contamination, Suzuki et al. (2011) examined Cr-spinel from beach sands in the Bonin Islands, reasoning that Cr-spinel is an early-formed mineral in most magmas and an indicator of primitive magma Os compositions. They found unradiogenic Os in Cr-spinel with boninitic affinity, versus a potential slab component reflected in spinel with tholeiitic affinity. These authors also argued that oxidative conditions in the mantle can lead to radiogenic Os mobilization in the arc. Ultimately, the most convincing arguments for or against radiogenic Os from the slab comes from high-MgO Grenada picrites. These samples have been shown to have less-radiogenic Os signatures in more mafic lavas, with an increasing influence of crustal contamination in more evolved melts (Woodland et al. 2002; Bezard et al. 2015). Combined with evidence for the potential influence of subduction zone fluids on the composition of arc xenoliths, these results suggest that some contribution from the slab can be exhibited in arc lavas, but that the role of crustal contamination of melts within the arc itself can obfuscate original mantle-derived signatures.
Mechanical mixing processes
The debate as to whether slab-derived signatures are evident in HSE and Os isotopes within arc volcanic rocks has recently been enhanced by the recognition that mechanical mixing between peridotite mantle and recycled ocean rocks is likely an important process in modifying HSE contents at subduction zones. Studies of HSE contents and Os isotope compositions of mélange mafic metamorphic blocks at Catalina Island and the Franciscan Complex (California) and at the Samana Metamorphic Complex (Dominican Republic) have shown significant differences between block cores and block rims (Penniston-Dorland et al. 2012, 2014). In particular, while the cores of the blocks have enhanced PPGE compared with IPGE and radiogenic 187Os/188Os, mimicking patterns for evolved basaltic rocks, or some sedimentary protoliths, the rims approach HSE contents expected in some mantle peridotites, with less radiogenic 187Os/188Os than the cores (Fig. 38). Penniston-Dorland et al. (2014) have demonstrated that mélange mechanical mixing occurs across a range of temperatures (≤ 200 to ~ 600 °C) during subduction leading to a hybrid rock composition of peridotite, basaltic materials and sediments. Measurements of the HSE in arc volcanics suggest variable amounts of peridotitic mantle with radiogenic Os components (e.g., Alves et al. 1999, 2002; Borg et al. 2000) and mechanical mixing may play a major role in this process.
CONCLUSIONS AND PERSPECTIVES
The highly siderophile elements are expected to be strongly incorporated into Earth’s metallic core, but their abundance in the upper mantle appears to have been set by the late addition of meteoritic material after core formation was complete. Partial melting of the mantle since that time has resulted in a significant fractionation of the HSE. The platinum-PGE, Re and Au, can behave as moderately compatible or incompatible elements during melting, and may be variably enriched in melts,while the Iridium-PGE behave as highly compatible elements. Sulfide appears to be a major host for HSE in mantle rocks, despite its relatively low abundance (between 0.04 and 0.08%). However, sulfide cannot account for the fractionation of HSE that occurs during the melting that generates MORB, which generally possess very low Os–Ir–Ru contents, and relatively high Re–Pd and Pt. Rather this fractionation appears to result from the crystallization of Os–Ir–Ru alloy phases in refractory mantle rocks, accompanying the exhaustion of sulfide by melting. The HSE content of MORB is further modified by the segregation of sulfide during fractional crystallization in the magmatic environment, where the HSEs are quantitatively removed into sulfide, leaving the residual melt depleted in these elements.
The fractionation of Re and Os accompanying the generation of MORB, intraplate lavas and those produced at convergent margins is one of the key processes controlling the distribution of these elements between Earth’s mantle and crust. Therefore, decay of 187Re to 187Os provides an exceptional tracer of recycled crustal materials in Earth’s mantle. This is because oceanic and continental crust possess high Re/Os ratios, and develop radiogenic Os isotope compositions over time, which in turn can be readily traced as recycled material if mixed back into the convective mantle. However, while MORB glass commonly preserves a radiogenic 187Os/188Os composition, this is most readily explained by seawater-derived contamination of the melt that occurs during magma ascent through the oceanic crust. Although reliable data for MORB glass remain limited these observations suggest that to a greater or lesser extent all MORB glass has been affected seawater contamination. This then also implies that other elements may have been affected by such contamination, most likely dependent upon their relative concentration in MORB glass and seawater. Sulfide, although demonstrably affected by the same seawater contamination, provides a more reliable record of the primary 187Os/188Os isotope composition of MORB, particularly those sulfides with high Os concentrations (i.e., > 100 ppb). These high-Os sulfides preserve relatively unradiogenic 187Os/188Os isotope compositions pointing to a mantle source that has experienced long term depletion of Re, similar to abyssal peridotites, with no evidence for the presence of recycled crust.
In addition to the effects of seawater contamination observed in MORB, intraplate lavas and those generated at convergent margins may interact with sub-continental lithospheric mantle, itself variably contaminated by multiple metasomatic events since it became isolated from the convecting mantle, and incorporate additional complications from the overlying crust. At convergent margins there is the additional complication of fluxes generated as a result of the subduction of the down-going slab with the potential for overprinting pre-existing Re–Os isotope and HSE fingerprints. While the HSE and its isotope systems offer some unique perspectives on mantle processes and the generation of a wide range of magmas, their application needs to be exercised with care—the geochemical context provided by other isotope systems and trace element signatures should be considered and the specific set of local conditions, both physical and chemical, taken into account in addition to the use of these invaluable tools.
We thank Jean-Louis Birck, Olivier Alard, Christian Pin, Ivan Vlastélic, Anthony Cohen, Ali Bouhifd for valuable insight and discussions over the years. AG would like to thank Jean-Luc Devidal for assistance with the electron microprobe measurements at Blaise Pascal University. The authors thank Chris W. Dale for a review that greatly improved the manuscript. This research was partially financed by the French Government Laboratory of Excellence initiative n°ANR-10-LABX-0006, the Région Auvergne and the European Regional Development Fund. JD acknowledges the support of NSF (NSF-EAR grant 1447130). JH was supported by a NERC Advanced Research Fellowship ( NE/J017981/1), a Blaustein Visiting Professorship at Stanford University, and a visiting investigator appointment at the Carnegie Institution, Washington. This is Laboratory of Excellence ClerVolc contribution number 178.