TY - JOUR
AU - Kjarsgaard, Bruce, A
AB - Abstract The rock type most commonly associated with komatiite throughout Earth’s history is tholeiitic basalt. Despite this well-known association, the link between komatiite and basalt formation is still debated. Two models have been suggested: that tholeiitic basalts represent the products of extensive fractional crystallization of komatiite, or that basalts are formed by lower degrees of mantle melting than komatiites in the cooler portions of a zoned plume. We present major and trace element data for tholeiitic basalts (∼7·5 wt% MgO) and dunites (46–48 wt% MgO) from the Palaeoproterozoic Winnipegosis Komatiite Belt (WKB), which, along with previous data for WKB komatiites (17–26 wt% MgO), are utilized to explore the potential links between komatiite and basalt via crystallization processes. The dunites are interpreted as olivine + chromite cumulates that were pervasively serpentinized during metamorphism. Their major and immobile trace element relationships indicate that the accumulating olivine was highly magnesian (Mg# = 0·91–0·92), and that small amounts (4–7 wt% on average) of intercumulus melt were trapped during their formation. These high inferred olivine Mg# values suggest that the dunites are derived from crystallization of komatiite. The tholeiitic basalts have undergone greenschist-facies metamorphism and have typical geochemical characteristics for Palaeoproterozoic basalts, with the exception of high FeO contents. Their REE patterns are similar to Winnipegosis komatiites, although absolute concentrations are higher by a factor of ∼2·5. The ability of thermodynamic modelling with MELTS software to reproduce komatiite liquid lines of descent (LLD) is evaluated by comparison with the crystallization sequence and mineral compositions observed for Winnipegosis komatiites. With minor caveats, MELTS is able to successfully reproduce the LLD. This modelling is extended to higher pressures to simulate crystallization of komatiitic melt in an upper crustal magma chamber. All major and rare earth element characteristics of the tholeiitic basalts can be reproduced by ∼60 % crystallization of a Winnipegosis komatiite-like parental melt at pressures of ∼1·5–2·5 kbar at oxygen fugacities between QFM − 1 and QFM + 1, where QFM is the quartz–fayalite–magnetite buffer. Winnipegosis basalts have low Mg# values that are not in equilibrium with mantle peridotite. They therefore cannot represent primary mantle melts derived from cooler mantle than the komatiites, and require fractional crystallization processes in their formation. Furthermore, their trace element characteristics indicate a depth of melting indistinguishable from that of the Winnipegosis komatiites, and derivation from an identical depleted mantle source. All geochemical and geological evidence is therefore consistent with their derivation from a komatiitic melt, and the presence of a large komatiite-derived dunite body in the WKB provides evidence of extensive fractionation of komatiite in the upper crust. The observed uniform basalt compositions are interpreted as the result of a density minimum in the evolving komatiitic melt at temperatures between clinopyroxene and plagioclase saturation, with efficient extraction of melt from a mixture containing ∼60 % crystals. We conclude that the WKB basalts formed by fractional crystallization of a komatiitic parental melt, and suggest that this model may be more broadly applicable to other localities where komatiites and associated basalts show similar geochemical or isotopic characteristics. INTRODUCTION The most common igneous rock type found in association with komatiites are tholeiitic basalts, which volumetrically dominate most komatiite-bearing volcanic sequences (Storey et al., 1991; Bickle et al., 1994; Arndt et al., 2008; Campbell & Davies, 2017). Following the discovery of komatiites (Viljoen & Viljoen, 1969a, 1969b), the relationship between komatiites and associated tholeiitic basalts was an important question in komatiite research. Two major models were proposed (Fig. 1): that basalts associated with komatiites are the product of extensive fractional crystallization of a komatiitic parental melt (fractional crystallization model), or that compositional differences between basalts and komatiites are due to derivation from parental melts formed by different degrees of mantle melting (mantle melting model; Arndt & Nesbitt, 1982). Fig. 1. Open in new tabDownload slide Schematic diagrams (not to scale) of models for formation of komatiite and associated tholeiitic basalt. (a) Mantle melting type model, in which komatiites form in the hottest central portions of ascending mantle plumes, whereas basalts form in peripheral regions cooled by entrainment of, and contact with ambient mantle (Campbell et al., 1989). (b) Fractional crystallization model, in which basalts form by extensive differentiation and possible crustal contamination of komatiitic parental melts (e.g. Shimizu et al., 2005). Fig. 1. Open in new tabDownload slide Schematic diagrams (not to scale) of models for formation of komatiite and associated tholeiitic basalt. (a) Mantle melting type model, in which komatiites form in the hottest central portions of ascending mantle plumes, whereas basalts form in peripheral regions cooled by entrainment of, and contact with ambient mantle (Campbell et al., 1989). (b) Fractional crystallization model, in which basalts form by extensive differentiation and possible crustal contamination of komatiitic parental melts (e.g. Shimizu et al., 2005). In localities where a continuous compositional spectrum between komatiites and basalts exists, both fractional crystallization models and mantle melting models have been invoked. Evidence for the derivation of basalts from komatiite crystallization is provided by systematic changes in magma composition from ultramafic to mafic within individual volcanic cycles in the Abitibi greenstone belt (Arndt et al., 1977), and within ultramafic intrusions of the Cape Smith Belt (Francis & Hynes, 1979). Conversely, in the Abitibi and elsewhere, differing major, trace, and rare earth element (REE) ratios between komatiites and associated basalts were interpreted as supporting the mantle melting model (Nesbitt & Sun, 1976; Sun & Nesbitt, 1978; Stamatelopoulou-Seymour et al., 1983). It is possible that both mechanisms are in operation in nature, even within a single greenstone belt (Arndt & Nesbitt, 1982). However, distinguishing these processes has frequently been hampered by the widespread presence of a compositional gap between komatiites and less mafic rocks at ∼18 wt% MgO (Arndt & Nisbet, 1982), and the altered nature of many Archaean komatiites and basalts (Arndt et al., 2008). Furthermore, many geochemical tools provide non-unique solutions to this problem. For example, elevated light REE/middle REE (LREE/MREE) in basalts compared with komatiites (Sun & Nesbitt, 1978) could represent their derivation from lower degree melts or less depleted mantle sources, or could suggest that significant crustal assimilation accompanied fractionation of a basalt’s komatiitic parent. Similarly, although basalts and komatiites from a number of localities, such as the ∼3·5 Ga Komati formation (Hamilton et al., 1979), ∼3·0 Ga Lumby Lake greenstone belt (Hollings & Wyman, 1999), and ∼2·7 Ga Abitibi greenstone belt (Blichert-Toft & Arndt, 1999; Debaille et al., 2013) show identical initial Nd isotopic compositions within each greenstone belt, this could be interpreted to reflect either a fractional crystallization link, or different parental melts formed by different degrees of melting of the same mantle source. Due to the issues outlined above, and in part due to the influential work of Campbell et al. (1989), later studies have tended to favour mantle melting type models. Campbell et al. (1989) suggested that komatiites formed through melting of the hottest, central portion of Archaean plumes, whereas basalts formed by melting in the cooler head. This model has been broadly adopted even for suites with very similar trace element characteristics (Rollinson, 1999; Kerr, 2005; Manikyamba et al., 2008; Schneider et al., 2019), and few recent researchers have advocated fractional crystallization models (Shimizu et al., 2005; Mole et al., 2018; Ghosh et al., 2019). Despite this, the composition of typical Archaean tholeiitic basalts provides evidence for at least some role of fractional crystallization in their generation. The modal composition of basalts in Archaean cratons is ∼6 wt% MgO (Kamber & Tomlinson, 2019), significantly lower than present-day mid-ocean ridge basalt (MORB) primary magmas (9–13 wt% MgO; Kamenetsky, 1996; Sobolev, 1996; Sobolev & Chaussidon, 1996; Laubier et al., 2007), produced in a cooler mantle. An MgO content of ∼6 wt% cannot be in equilibrium with mantle peridotite (e.g. Herzberg & O’Hara, 2002), given typical tholeiitic basalt Fe contents. This discrepancy is exacerbated given that primary magmas generated from hotter Archaean and Palaeoproterozoic ambient mantle are expected to be more MgO rich (up to 18–24 wt% MgO; Herzberg et al., 2010) than present-day magmas. Instead, the low modal MgO of basalts from Archaean cratons bears a closer resemblance to modern continental flood basalts (CFBs), widely interpreted as the products of extensive magmatic differentiation (Farmer, 2007). Therefore, the difference between fractional crystallization and mantle melting models is subtle, with mantle melting models still probably requiring extensive fractional crystallization of a higher MgO parental melt, such as a picrite or high-Mg basalt, albeit not as MgO-rich as a komatiite. The Palaeoproterozoic Winnipegosis Komatiite Belt (WKB) is an ideal location to investigate the relationship between basalts and komatiites. Both komatiitic and basaltic volcanism are abundant, and the Winnipegosis komatiites are very well preserved (Waterton et al., 2017), providing a high-fidelity bulk-rock geochemical and mineral chemistry record. Additionally, the presence of large bodies of ultramafic cumulates suggests the possibility that there was extensive komatiite fractionation in the upper crust. This study presents new major and trace element data for basalts and dunites from the WKB. These are combined with previously obtained data for Winnipegosis komatiites (Waterton et al., 2017) to test the hypothesis that the basalts formed through extensive crystallization of a komatiitic parental magma. Thermodynamic modelling of komatiite crystallization is undertaken using Rhyolite-MELTS software (Gualda et al., 2012; Ghiorso & Gualda, 2015), and the results of the models are compared with well-constrained observations of Winnipegosis komatiite crystallization (Waterton et al., 2017). The modelling indicates that fractional crystallization of komatiitic parental melts is a viable mechanism for producing tholeiitic basalts, and geological constraints indicate that it was most probably in operation during formation of the WKB. GEOLOGICAL SETTING AND SAMPLES Igneous rocks in the Winnipegosis Komatiite Belt The geology and regional tectonic setting at the time of WKB formation has been described by Waterton et al. (2017). Briefly, the 1870 ± 7 Ma WKB is an ∼150 km × 30 km greenstone belt deposited unconformably on the NW margin of the Archaean Superior Craton, in Manitoba, Canada (Fig. 2; McGregor, 2011; Waterton et al., 2017). The WKB is dominated by tholeiitic basalt and komatiite, and forms part of the Circum-Superior Belt, an ∼3000 km long series of volcano-sedimentary belts that wrap around the northern margin of the Superior Craton (Baragar & Scoates, 1981), with associated dykes found in the margins and interior of the craton (Minifie et al., 2013). The Circum-Superior Belt erupted during a relatively narrow time span at c. 1·88 Ga (Heaman et al., 2009; Ciborowski et al., 2017) during closure of the Palaeoproterozoic Manikewan Ocean, which culminated in the Trans-Hudson orogeny at 1·83–1·80 Ga (Ansdell et al., 1995; Corrigan et al., 2009; Corrigan, 2012). High liquidus temperatures requiring mantle temperatures above those of ambient mantle at 1·88 Ga, coupled with the narrow time span of magmatism, and a variety of geochemical arguments (Minifie et al., 2013; Ciborowski et al., 2017; Waterton et al., 2017), have been used to argue for a plume origin for the Circum-Superior Belt, with the plume head impinging on the central region of the Superior Craton (Waterton et al., 2017). Fig. 2. Open in new tabDownload slide Location of the Winnipegosis Komatiite Belt in relation to the Circum-Superior Belt (CSB) and Trans-Hudson Orogen (THO), modified from Baragar & Scoates (1981), Corrigan (2012), Minifie et al. (2013) and Waterton et al. (2017). CSB exposures on the Ottawa and Sleeper Islands are circled because of their small size. CSB-related mafic dykes and carbonatite complexes are identified as thick red lines and dots, respectively; widths are not to scale. Boundaries of cratons and CSB shown are surface exposures except Sask craton and WKB, for which approximate subsurface extents are shown. All other geology is not subdivided (white areas). Blue areas indicate water bodies. WKB, Winnipegosis Komatiite Belt; TNB, Thompson Nickel Belt; FRB, Fox River Belt. Blue dotted line indicates the extent of the Reindeer Zone, the juvenile core of the THO. Fig. 2. Open in new tabDownload slide Location of the Winnipegosis Komatiite Belt in relation to the Circum-Superior Belt (CSB) and Trans-Hudson Orogen (THO), modified from Baragar & Scoates (1981), Corrigan (2012), Minifie et al. (2013) and Waterton et al. (2017). CSB exposures on the Ottawa and Sleeper Islands are circled because of their small size. CSB-related mafic dykes and carbonatite complexes are identified as thick red lines and dots, respectively; widths are not to scale. Boundaries of cratons and CSB shown are surface exposures except Sask craton and WKB, for which approximate subsurface extents are shown. All other geology is not subdivided (white areas). Blue areas indicate water bodies. WKB, Winnipegosis Komatiite Belt; TNB, Thompson Nickel Belt; FRB, Fox River Belt. Blue dotted line indicates the extent of the Reindeer Zone, the juvenile core of the THO. The WKB has no surface exposure. It was identified from geophysical signatures (Fig. 3a) and proven by drilling. Lucas et al. (1996) divided the igneous rocks of the WKB into three suites, from top to bottom: the Upper Tholeiite, Winnipegosis komatiite, and Grand Island tholeiite (Fig. 3b). However, the Upper Tholeiite is geochemically very similar to the Grand Island tholeiite, and it is not certain whether it represents a stratigraphically distinct unit or is a thrust repetition of the sequence (Burnham, 2009; McGregor, 2011). Establishing the relative proportions of the different igneous rock types found in the WKB is difficult because of a lack of surface exposure, and because borehole locations were chosen based on their geophysically identified Ni–Cu potential (McGregor, 2011), thus providing a biased record of the WKB stratigraphy. Komatiites were intersected in 10 of the 27 boreholes drilled. However, as these form linear aeromagnetic highs targeted during drilling, they may be over-represented in the borehole record. Basaltic rocks also form a significant component of the belt, with eruptive basalt sequences and/or gabbros intersected in 11 boreholes. Basalts are frequently interlayered with siltstones, argillites, and dolomitic rocks, and are generally not intersected in komatiite-dominated boreholes, although thin layers of gabbro have been described in some komatiite-dominated cores (McGregor, 2011). Estimates of the total stratigraphic thickness of basalt preserved in the WKB range between 550 m (Fig. 3c; Hulbert et al., 1994) and <1 km (McGregor, 2011). This is comparable with estimates of the total komatiite thickness in the WKB, which range between >600 m (Fig. 3c; Hulbert et al., 1994) and <1 km (McGregor, 2011). Fig. 3. Open in new tabDownload slide (a) Aeromagnetic map of the central portion of the Winnipegosis Komatiite Belt, modified from McGregor (2011). Red and yellows indicate magnetic highs, including linear magnetic highs of WKB volcanic rocks and rounded magnetic high of RP7 dunite body; blue indicates magnetic lows. Black lines divide tectono-stratigraphic units. TNB, Thompson Nickel Belt. (b) Seismic-reflection interpretation of Lithoprobe line 3b modified from Lucas et al. (1996) along blue line between A and A' in (a). Approximate positions of boreholes RP1A, RP7, and RP8 are projected along-strike onto line. Black lines represent prominent reflectors and tectono-stratigraphic unit boundaries. (c) Reconstructed stratigraphy of the Winnipegosis Komatiite Belt, modified from Waterton et al. (2017), with Upper Tholeiite thickness based on section intersected by borehole RP8 (McGregor, 2011). Dashed connection between Winnipegosis komatiite and Upper Tholeiites reflects their uncertain stratigraphic relationship. Fig. 3. Open in new tabDownload slide (a) Aeromagnetic map of the central portion of the Winnipegosis Komatiite Belt, modified from McGregor (2011). Red and yellows indicate magnetic highs, including linear magnetic highs of WKB volcanic rocks and rounded magnetic high of RP7 dunite body; blue indicates magnetic lows. Black lines divide tectono-stratigraphic units. TNB, Thompson Nickel Belt. (b) Seismic-reflection interpretation of Lithoprobe line 3b modified from Lucas et al. (1996) along blue line between A and A' in (a). Approximate positions of boreholes RP1A, RP7, and RP8 are projected along-strike onto line. Black lines represent prominent reflectors and tectono-stratigraphic unit boundaries. (c) Reconstructed stratigraphy of the Winnipegosis Komatiite Belt, modified from Waterton et al. (2017), with Upper Tholeiite thickness based on section intersected by borehole RP8 (McGregor, 2011). Dashed connection between Winnipegosis komatiite and Upper Tholeiites reflects their uncertain stratigraphic relationship. By contrast, dunitic rocks were identified more rarely during WKB drilling. A single large intrusive dunite body was intersected by boreholes RP7 and RP10. This body was identified from an ∼5 km × 2·5 km ellipse-shaped magnetic high, found ∼5 km east of the supracrustal rocks of the WKB, and has a thickness of >600 m in borehole RP10. A number of smaller peridotite–dunite bodies were also identified within the Winnipegosis stratigraphic sequence, including an ∼300 m thick intrusive peridotite–gabbro section in borehole RP18 (McGregor, 2011). However, it is not clear from core logging descriptions whether many of the smaller peridotite–dunite bodies represent intrusive bodies (such as sills) or are the cumulate portions of komatiite flows. Samples and petrography Despite the widespread occurrence of basaltic rocks in the WKB, the only samples available for this study were from borehole RP8. The RP8 metabasalts (referred to as ‘RP8 basalts’ for simplicity) are from the ‘Upper Tholeiite’ suite of the WKB. Previous studies have noted the strong geochemical similarity of these rocks to the Grand Island Tholeiite (Burnham, 2009; McGregor, 2011), and so the RP8 rocks are taken as an example of the style of basaltic magmatism in the WKB. RP8 basalts are green in core samples and pervasively altered to greenschist minerals in thin section. The samples are homogeneous, and no primary minerals or textures are preserved (Fig. 4). The greenschist mineral assemblage consists of actinolite, hornblende, epidote, quartz, and titanite, with minor chlorite and opaque oxides. With the exception of epidote, which is present in samples from nearer the top of the cored section, but absent from samples deeper in the section, the mineral assemblage is consistent throughout. However, the grain size of metamorphic minerals increases considerably with depth in the section, suggesting the deeper parts of borehole RP8 experienced higher peak metamorphic conditions. Quartz veins with minor carbonate were observed in several samples. All 10 available samples from the borehole were selected for major and trace elements. Fig. 4. Open in new tabDownload slide Photomicrographs of Winnipegosis samples in plane-polarized light. Olivine (ol), chromite (chr), clinopyroxene (cpx), devitrified glass (gl), serpentine and brucite (ser + br), magnetite (mt), actinolite (act), and quartz (qz) are labelled. (a) Well-preserved massive komatiite sample RP1A-111, with olivine and chromite phenocrysts set in a matrix of dendritic clinopyroxene and devitrified glass. (b) Random olivine spinifex sample RP12-306.1, with long, straight serpentinized dendritic olivine dividing patches of clinopyroxene dendrites and devitrified glass. (c) Dunite sample RP7-14. Olivine is completely altered to serpentine, brucite, magnetite, and minor chlorite; black undulating bands are magnetite. Examples of preserved chromite are circled. (d) Coarse-grained metabasalt sample RP8-10, from near base of borehole RP8. No primary textures remain. Fig. 4. Open in new tabDownload slide Photomicrographs of Winnipegosis samples in plane-polarized light. Olivine (ol), chromite (chr), clinopyroxene (cpx), devitrified glass (gl), serpentine and brucite (ser + br), magnetite (mt), actinolite (act), and quartz (qz) are labelled. (a) Well-preserved massive komatiite sample RP1A-111, with olivine and chromite phenocrysts set in a matrix of dendritic clinopyroxene and devitrified glass. (b) Random olivine spinifex sample RP12-306.1, with long, straight serpentinized dendritic olivine dividing patches of clinopyroxene dendrites and devitrified glass. (c) Dunite sample RP7-14. Olivine is completely altered to serpentine, brucite, magnetite, and minor chlorite; black undulating bands are magnetite. Examples of preserved chromite are circled. (d) Coarse-grained metabasalt sample RP8-10, from near base of borehole RP8. No primary textures remain. Sixteen dunite samples from borehole RP7 were selected for major element analysis, covering the entire depth range sampled in the borehole. Dunites from borehole RP7 are altered to a sub-greenschist (Jolly, 1982) metamorphic assemblage of serpentine, brucite, chlorite and magnetite, with no primary olivine preserved. Igneous chromite is preserved but almost ubiquitously overgrown by metamorphic magnetite. However, it can still be distinguished from metamorphic magnetite by its subhedral chromite habit and relatively even distribution in the dunites. By contrast, magnetite forms thin, undulating bands dividing large patches of serpentine and brucite, which partially outline a previous cumulate mesh texture. The major element composition of the igneous chromite cores was analysed by electron probe microanalysis (EPMA). Komatiite bulk-rock major element data from boreholes RP1A and RP12, measured by X-ray fluorescence (XRF), and olivine and chromite EPMA data, are taken from Waterton et al. (2017). For this study, new bulk-rock major, trace, and transition metal element data were measured for four massive komatiite samples. New transition metal data were also obtained for four massive komatiite and four differentiated komatiite samples (Supplementary Data are available for downloading at http://www.petrology.oxfordjournals.org), following the methods of Waterton et al. (2017). New clinopyroxene major element data from the komatiites was measured by EPMA for this study. Clinopyroxene data were obtained from one random olivine spinifex sample (RP12-306.1) and one acicular pyroxene sample (RP12-307.3) from the strongly differentiated flow described by Waterton et al. (2017), in addition to several massive komatiite flow samples. Clinopyroxene phenocrysts from the acicular pyroxene sample have the largest grain sizes (up to ∼4 mm long) and habits indicative of a closer approach to equilibrium conditions, whereas clinopyroxene from all other samples is dendritic, providing clear evidence of disequilibrium growth (see Supplementary Data). Clinopyroxene dendrites are larger in the random olivine spinifex sample (RP12-306.1) than the massive komatiite samples, which were subdivided into textural types based primarily on groundmass features (Waterton, 2018). These, in turn, show a decrease in dendrite size from the coarsest ‘type 1’ samples (RP1A-111), inferred to have formed at the lowest cooling rates, towards ‘type 2’ samples with the smallest clinopyroxene dendrite sizes (RP1A-18, RP1A-99). Sample RP1A-8 has a texture intermediate between type 1 and type 2 samples and correspondingly intermediate clinopyroxene dendrite sizes. METHODS Whole-rock major and minor elements Whole-rock major and minor element abundances were measured by XRF on lithium tetraborate (Li2B4O7) fusion disks at Franklin and Marshall College, following a method modified from Mertzman (2000; see below). This method has been used to obtain major and minor element data for mafic and ultramafic materials by a number of researchers, some of whom attempted to establish the accuracy and precision of the method by repeat analyses of a small number of certified reference materials (CRM; e.g. Puchtel et al., 2016, 2018; Waterton et al., 2017). However, this approach only constrains the ‘measurement repeatability’ of the method (JCGM, 2008); the measurements are made using the same calibrations and operating conditions over a short period of time, such as a single analytical session. To compare new basalt and dunite major and minor element data presented in this paper with komatiite data from Waterton et al. (2017), the intermediate measurement precision (JCGM, 2008) of the method (potentially including new calibrations, calibrators, etc.) must be established for a range of mafic and ultramafic compositions. In this light, a large number of major and minor element analyses collected over a period of up to 4 years are presented for a basalt CRM (BHVO-2; Wilson, 1998), a komatiite CRM (OKUM; IAG, 2015a) and a harzburgite CRM (MUH-1; IAG, 2015b), and compared with certified and recommended values (Table 1). Additionally, we present new major and minor element data from candidate reference material WGB-1 (gabbro rock; Sen Gupta 1994), and two in-house major element standards, depleted peridotite GP220 (Pearson et al., 2004) and Greenland kimberlite 474564 (Tappe et al., 2011) to establish baseline values for future measurements (Supplementary Data). All individual standard analyses are available in the Supplementary Data. Analytical procedure All rock powder to be analysed must be able to pass through an 80 mesh sieve, to provide sufficient surface area for the reaction with Li2B4O7 to progress to completion. Winnipegosis samples were first coarsely crushed in a steel jaw crusher, then powdered using ceramic ball milling equipment during initial characterization of the Winnipegosis Komatiite Belt (Hulbert et al., 1994). For this study, all sample and CRM powders were sieved; any powder that did not pass an 80 mesh sieve was re-crushed using a Spex shatter-box equipped with a ceramic lined sample container and a ceramic puck, and thoroughly mixed back into the fine powder. Loss on ignition (LOI) was determined by heating a precisely weighed aliquot of the sample powder in a clean dry ceramic crucible to 950 °C for 1·5 h, cooling to room temperature in a desiccator, and re-weighing the crucible. Previously ignited sample powder (0·4000 ± 0·0001 g) and 3·6000 ± 0·0002 g of Li2B4O7 was weighed out, transferred to a clean glass bottle with a snap cap and mixed for 10 min in a Spex mixer-mill. The resulting homogeneous mixture was transferred into a 95 % platinum–5 % gold crucible that was covered with a lid of the same composition. The crucible–lid assembly was heated with a Meeker burner until molten and hand mixed using tongs at 3, 6 and 9 min. This melt was poured from the crucible into the lid after 10–11 min to cast the glass disk. This procedure produced a glass disk that was used for XRF analysis of SiO2, TiO2, Al2O3, total iron reported as Fe2O3t, MnO, MgO, CaO, Na2O, K2O, P2O5, and in some cases analysis of Sr, Zr, V, Ni, Cr, and Co. Although minor and trace elements can be more precisely measured using a pressed powder briquette XRF method (e.g. Mertzman, 2000), this method was not used here due to the small size of Winnipegosis core samples analysed. Concentrations were analysed using a PANalytical, Inc. PW2404-R system equipped with a 4 kW rhodium X-ray tube and a 168-sample deck. Working curves for each element were determined by analysing geochemical rock standards using accepted values from Abbey (1983) and Govindaraju (1994). Between 60 and 70 data points were gathered for each working curve; various element interferences were also taken into account. Results were calculated using the Super-Q software package. Major and minor element accuracy and precision Thirty-three analyses of the MUH-1 harzburgite CRM were collected across 10 analytical sessions spanning the years 2014–2018 (Table 1). All elements with 2σrel < 20 % overlap the certified values for MUH-1 (IAG, 2015b) within two absolute standard deviations (2σabs), and are considered accurate at the stated 2σ level of precision. The measurands TiO2, Na2O, K2O, P2O5, Sr, Zr, and V, present in low concentrations in MUH-1, are considered imprecise with 2σrel > 30 %. Ni had 2σrel = 28 %, but showed an unusual flat distribution of data, skewed to higher values than the certified value, and is also considered imprecise. Table 1: Major and minor element data for MUH-1, OKUM, and BHVO-2 CRMs from this study, compared with certified values (cert.; IAG 2015) and GeoReM preferred values (GeoReM; Jochum et al., 2016) . MUH-1 (n = 33) . 2σ abs. . 2σ rel. (%) . 95 % conf. . MUH-1 cert. . OKUM (n = 32) . 2σ abs. . 2σ rel. (%) . 95 % conf. . OKUM cert. . BHVO-2 (n = 30) . 2σ abs. . 2σ rel. (%) . 95 % conf. . BHVO-2 GeoReM . SiO2 (wt%) 43·86 00·84 1·9 0·25 44·59 45·95 0·46 1·0 0·14 46·23 49·56 0·34 0·7 0·12 49·60 TiO2 n.q. 0·0380 0·379 0·012 3·1 0·00 0·3980 2·70 0·02 0·9 0·01 2·731 Al2O3 10·47 00·08 5·3 0·02 1·473 8·37 0·13 1·6 0·04 8·347 13·47 0·17 1·2 0·05 13·44 Fe2O3t 90·87 00·47 4·8 0·14 9·48 12·54 0·25 2·0 0·08 12·37 12·32 0·22 1·8 0·06 12·39 MnO 00·134 0·011 8·0 0·003 0·1302 0·187 0·004 2·1 0·001 0·1899 0·170 0·008 4·7 0·002 0·1690 MgO 42·11 0·67 1·6 0·20 42·23 22·41 0·29 1·3 0·09 22·30 7·26 0·05 0·7 0·01 7·257 CaO 1·40 0·07 4·8 0·02 1·339 8·36 0·14 1·6 0·04 8·222 11·42 0·05 0·4 0·02 11·40 Na2O n.q. 0·115 1·18 0·05 4·3 0·02 1·190 2·21 0·05 2·2 0·01 2·219 K2O n.q. 0·013 0·048 0·010 21·5 0·003 0·046 0·489 0·021 4·3 0·006 0·513 P2O5 n.q. 0·0083 0·026 0·003 11·6 0·001 0·0279 0·276 0·016 5·7 0·004 0·2685 Total (raw) 98·98 0·53 0·5 0·16 99·42 99·44 0·45 0·4 0·13 99·32 99·88 0·61 0·6 0·23 99·99 LOI 10·01 0·37 3·7 0·11 10·36 4·68 0·47 10·1 0·14 4·70 –0·51 0·13 25·8 0·05 –0·81 Sr (ppm) n.q. 9·4 n.q. 16·9 392 21 5·3 8 394·1 Zr n.q. n.a. n.q. 17·8 174 5 3·1 2 171·2 V n.q. 45 166 12 6·9 3 176 330 14 4·1 5 318·2 Ni n.q. 2323 n.q. 928 n.q. 119·8 Cr 2944 93 3·2 31 2992 2478 210 8·5 63 2576 296 12 4·2 5 287·2 Co 115 22 19·0 7 117·8 85 18 21·3 5 93·1 n.q. 44·89 . MUH-1 (n = 33) . 2σ abs. . 2σ rel. (%) . 95 % conf. . MUH-1 cert. . OKUM (n = 32) . 2σ abs. . 2σ rel. (%) . 95 % conf. . OKUM cert. . BHVO-2 (n = 30) . 2σ abs. . 2σ rel. (%) . 95 % conf. . BHVO-2 GeoReM . SiO2 (wt%) 43·86 00·84 1·9 0·25 44·59 45·95 0·46 1·0 0·14 46·23 49·56 0·34 0·7 0·12 49·60 TiO2 n.q. 0·0380 0·379 0·012 3·1 0·00 0·3980 2·70 0·02 0·9 0·01 2·731 Al2O3 10·47 00·08 5·3 0·02 1·473 8·37 0·13 1·6 0·04 8·347 13·47 0·17 1·2 0·05 13·44 Fe2O3t 90·87 00·47 4·8 0·14 9·48 12·54 0·25 2·0 0·08 12·37 12·32 0·22 1·8 0·06 12·39 MnO 00·134 0·011 8·0 0·003 0·1302 0·187 0·004 2·1 0·001 0·1899 0·170 0·008 4·7 0·002 0·1690 MgO 42·11 0·67 1·6 0·20 42·23 22·41 0·29 1·3 0·09 22·30 7·26 0·05 0·7 0·01 7·257 CaO 1·40 0·07 4·8 0·02 1·339 8·36 0·14 1·6 0·04 8·222 11·42 0·05 0·4 0·02 11·40 Na2O n.q. 0·115 1·18 0·05 4·3 0·02 1·190 2·21 0·05 2·2 0·01 2·219 K2O n.q. 0·013 0·048 0·010 21·5 0·003 0·046 0·489 0·021 4·3 0·006 0·513 P2O5 n.q. 0·0083 0·026 0·003 11·6 0·001 0·0279 0·276 0·016 5·7 0·004 0·2685 Total (raw) 98·98 0·53 0·5 0·16 99·42 99·44 0·45 0·4 0·13 99·32 99·88 0·61 0·6 0·23 99·99 LOI 10·01 0·37 3·7 0·11 10·36 4·68 0·47 10·1 0·14 4·70 –0·51 0·13 25·8 0·05 –0·81 Sr (ppm) n.q. 9·4 n.q. 16·9 392 21 5·3 8 394·1 Zr n.q. n.a. n.q. 17·8 174 5 3·1 2 171·2 V n.q. 45 166 12 6·9 3 176 330 14 4·1 5 318·2 Ni n.q. 2323 n.q. 928 n.q. 119·8 Cr 2944 93 3·2 31 2992 2478 210 8·5 63 2576 296 12 4·2 5 287·2 Co 115 22 19·0 7 117·8 85 18 21·3 5 93·1 n.q. 44·89 Data are presented on an anhydrous basis but not otherwise renormalized. 2σ abs., 2 absolute standard deviations; 2σ rel., 2 relative standard deviations; 95 % conf., 95 % confidence limits calculated using Student’s t value; n, number of analyses; n.a., not available, no certified data for this element; n.q., not quantifiable, could not be analysed or 2σ rel. greater than an arbitrary cut-off of 30 %. Open in new tab Table 1: Major and minor element data for MUH-1, OKUM, and BHVO-2 CRMs from this study, compared with certified values (cert.; IAG 2015) and GeoReM preferred values (GeoReM; Jochum et al., 2016) . MUH-1 (n = 33) . 2σ abs. . 2σ rel. (%) . 95 % conf. . MUH-1 cert. . OKUM (n = 32) . 2σ abs. . 2σ rel. (%) . 95 % conf. . OKUM cert. . BHVO-2 (n = 30) . 2σ abs. . 2σ rel. (%) . 95 % conf. . BHVO-2 GeoReM . SiO2 (wt%) 43·86 00·84 1·9 0·25 44·59 45·95 0·46 1·0 0·14 46·23 49·56 0·34 0·7 0·12 49·60 TiO2 n.q. 0·0380 0·379 0·012 3·1 0·00 0·3980 2·70 0·02 0·9 0·01 2·731 Al2O3 10·47 00·08 5·3 0·02 1·473 8·37 0·13 1·6 0·04 8·347 13·47 0·17 1·2 0·05 13·44 Fe2O3t 90·87 00·47 4·8 0·14 9·48 12·54 0·25 2·0 0·08 12·37 12·32 0·22 1·8 0·06 12·39 MnO 00·134 0·011 8·0 0·003 0·1302 0·187 0·004 2·1 0·001 0·1899 0·170 0·008 4·7 0·002 0·1690 MgO 42·11 0·67 1·6 0·20 42·23 22·41 0·29 1·3 0·09 22·30 7·26 0·05 0·7 0·01 7·257 CaO 1·40 0·07 4·8 0·02 1·339 8·36 0·14 1·6 0·04 8·222 11·42 0·05 0·4 0·02 11·40 Na2O n.q. 0·115 1·18 0·05 4·3 0·02 1·190 2·21 0·05 2·2 0·01 2·219 K2O n.q. 0·013 0·048 0·010 21·5 0·003 0·046 0·489 0·021 4·3 0·006 0·513 P2O5 n.q. 0·0083 0·026 0·003 11·6 0·001 0·0279 0·276 0·016 5·7 0·004 0·2685 Total (raw) 98·98 0·53 0·5 0·16 99·42 99·44 0·45 0·4 0·13 99·32 99·88 0·61 0·6 0·23 99·99 LOI 10·01 0·37 3·7 0·11 10·36 4·68 0·47 10·1 0·14 4·70 –0·51 0·13 25·8 0·05 –0·81 Sr (ppm) n.q. 9·4 n.q. 16·9 392 21 5·3 8 394·1 Zr n.q. n.a. n.q. 17·8 174 5 3·1 2 171·2 V n.q. 45 166 12 6·9 3 176 330 14 4·1 5 318·2 Ni n.q. 2323 n.q. 928 n.q. 119·8 Cr 2944 93 3·2 31 2992 2478 210 8·5 63 2576 296 12 4·2 5 287·2 Co 115 22 19·0 7 117·8 85 18 21·3 5 93·1 n.q. 44·89 . MUH-1 (n = 33) . 2σ abs. . 2σ rel. (%) . 95 % conf. . MUH-1 cert. . OKUM (n = 32) . 2σ abs. . 2σ rel. (%) . 95 % conf. . OKUM cert. . BHVO-2 (n = 30) . 2σ abs. . 2σ rel. (%) . 95 % conf. . BHVO-2 GeoReM . SiO2 (wt%) 43·86 00·84 1·9 0·25 44·59 45·95 0·46 1·0 0·14 46·23 49·56 0·34 0·7 0·12 49·60 TiO2 n.q. 0·0380 0·379 0·012 3·1 0·00 0·3980 2·70 0·02 0·9 0·01 2·731 Al2O3 10·47 00·08 5·3 0·02 1·473 8·37 0·13 1·6 0·04 8·347 13·47 0·17 1·2 0·05 13·44 Fe2O3t 90·87 00·47 4·8 0·14 9·48 12·54 0·25 2·0 0·08 12·37 12·32 0·22 1·8 0·06 12·39 MnO 00·134 0·011 8·0 0·003 0·1302 0·187 0·004 2·1 0·001 0·1899 0·170 0·008 4·7 0·002 0·1690 MgO 42·11 0·67 1·6 0·20 42·23 22·41 0·29 1·3 0·09 22·30 7·26 0·05 0·7 0·01 7·257 CaO 1·40 0·07 4·8 0·02 1·339 8·36 0·14 1·6 0·04 8·222 11·42 0·05 0·4 0·02 11·40 Na2O n.q. 0·115 1·18 0·05 4·3 0·02 1·190 2·21 0·05 2·2 0·01 2·219 K2O n.q. 0·013 0·048 0·010 21·5 0·003 0·046 0·489 0·021 4·3 0·006 0·513 P2O5 n.q. 0·0083 0·026 0·003 11·6 0·001 0·0279 0·276 0·016 5·7 0·004 0·2685 Total (raw) 98·98 0·53 0·5 0·16 99·42 99·44 0·45 0·4 0·13 99·32 99·88 0·61 0·6 0·23 99·99 LOI 10·01 0·37 3·7 0·11 10·36 4·68 0·47 10·1 0·14 4·70 –0·51 0·13 25·8 0·05 –0·81 Sr (ppm) n.q. 9·4 n.q. 16·9 392 21 5·3 8 394·1 Zr n.q. n.a. n.q. 17·8 174 5 3·1 2 171·2 V n.q. 45 166 12 6·9 3 176 330 14 4·1 5 318·2 Ni n.q. 2323 n.q. 928 n.q. 119·8 Cr 2944 93 3·2 31 2992 2478 210 8·5 63 2576 296 12 4·2 5 287·2 Co 115 22 19·0 7 117·8 85 18 21·3 5 93·1 n.q. 44·89 Data are presented on an anhydrous basis but not otherwise renormalized. 2σ abs., 2 absolute standard deviations; 2σ rel., 2 relative standard deviations; 95 % conf., 95 % confidence limits calculated using Student’s t value; n, number of analyses; n.a., not available, no certified data for this element; n.q., not quantifiable, could not be analysed or 2σ rel. greater than an arbitrary cut-off of 30 %. Open in new tab Thirty-two analyses of the OKUM komatiite CRM were made in three analytical sessions between 2015 and 2018 (Table 1). All elements and oxides with 2σrel < 25 %, with the exception of TiO2, overlap the certified values for OKUM (IAG, 2015a) within 2σabs, and are considered accurate at the stated 2σ level of precision. TiO2 is systematically ∼5 % lower than the OKUM certified values. Sr, Zr, and Ni are considered imprecise, with 2σrel > 30 %. Thirty analyses of the BHVO-2 basalt CRM were made in two analytical sessions between 2017 and 2018 (Table 1). All elements and oxides with 2σrel < 10 %, with the exception of TiO2 and K2O, overlap the GeoReM preferred values (Jochum et al., 2005, 2016) within 2σabs, and are considered accurate at the stated 2σ level of precision. TiO2 is again low compared with the BHVO-2 reference value, but differs from this value by only ∼1 %; the apparent inaccuracy is largely due to the high precision of the TiO2 analyses in BHVO-2. K2O in our data is systematically ∼5 % lower than the GeoReM preferred value. Ni and Co in BHVO-2 were not quantifiable via XRF analysis of fusion disks. LOI determinations of BHVO-2 showed poor precision with 2σrel ∼26 %. Unlike MUH-1 (a partially serpentinized harzburgite) and OKUM (an Mg poor meta-komatiite), BHVO-2 is from a recent, fresh basalt flow, and has a low volatile content. The small, negative LOI value is therefore dominated by mass differences arising from cation oxidation. As the BHVO-2 LOI is less negative in our analyses than the GeoReM preferred value, it is likely that variable, incomplete oxidation of redox-sensitive elements is the dominant cause of this imprecision. This effect is probably swamped in more volatile-rich samples by the much greater mass loss from volatiles. Measurement repeatability versus intermediate precision Data from MUH-1 provide the best estimate of the intermediate measurement precision of our XRF method for a number of elements (present in sufficient quantities), as the analyses were conducted over the largest number of sessions and period of time. This is illustrated by the SiO2 and Fe2O3t data for the CRMs above; despite having similar concentrations, OKUM and BHVO-2 have 2σrel at least 50 % lower than those for MUH-1 (better apparent precision by a factor of two), implying that differences between analytical sessions contribute a significant proportion of the uncertainty for these elements. We attempt to isolate the contributions of variability within a single session (repeatability) and of variability between different analytical sessions to the intermediate measurement precision of MUH-1. By comparing the standard deviations within single sessions with the standard deviation of session averages (Supplementary Data), we show that the variability between analytical sessions is typically around twice as large as the variability within a single analytical session. As the intermediate measurement precision includes uncertainty from both the variability within a single session and variability between sessions, estimates of the precision of this XRF method from the repeatability of a CRM within a single analytical session (e.g. Waterton et al., 2017) will overestimate (in terms of quality) the ‘true’ measurement precision by at least a factor of two. The long-term data for MUH-1 should provide a reasonable estimate of the intermediate measurement precision for this method as applied to our samples. We therefore recommend the following 2σ relative precisions for harzburgitic or depleted peridotite compositions: SiO2 and MgO < 2 %; Cr < 4 %; Al2O3, Fe2O3t and CaO < 6 %; MnO < 10 %; Co < 20 %. LOI in MUH-1 had a 2σ relative precision of <4 %, but this represents at best a lower bound for non-serpentinized samples. As the OKUM and BHVO-2 data were measured over only three and two analytical sessions respectively, the 2σ uncertainties listed in Table 1 probably underestimate the intermediate measurement precision for these compositions, although some portion of the uncertainty from inter-session variability is included by using data from more than one analytical session. Because of the similar concentrations in all three CRMs, we use the estimates of precision from MUH-1 for SiO2 and Fe2O3t to suggest intermediate measurement precisions of <2 % and <6 %, respectively, for basaltic and komatiitic compositions. As most other elements are present in very different concentrations in OKUM and BHVO-2 to MUH-1, we provide a rough estimate of the intermediate measurement precision for these CRMs by multiplying the 2σ uncertainties in Table 1 by a factor of two. For komatiitic compositions, conservative intermediate measurement precisions are estimated: Al2O3, MnO, MgO and CaO < 4 %; TiO2 < 6 %; Na2O < 10 %; P2O5, LOI, V and Cr < 25 %. It is uncertain whether elements such as K2O and Co would be sufficiently precise over the longer term for quantitative use. For basaltic compositions, conservative intermediate measurement precisions are estimated: TiO2, Al2O3, MgO, CaO and Na2O < 4 %; MnO, K2O, Zr, V and Cr < 10 %; Sr and P2O5 < 15 %. Cr shows much better precision in BHVO-2 than OKUM, despite much lower concentrations. It is therefore possible that one of the OKUM analytical sessions was a Cr outlier, in which case the precision estimated for OKUM is too pessimistic, or that the long-term precision of Cr in basaltic compositions is worse than estimated here. Whole-rock trace element data Whole-rock trace element data were analysed by solution inductively coupled plasma mass spectrometry (ICP-MS) at the Peter Hooper Geoanalytical Lab at Washington State University, using a two-step di-lithium tetraborate fusion followed by mixed acid digestion method (Knaack et al., 1994). Approximately 2 g of sample powder was mixed with an equal amount of Spectromelt A-10 Li-tetraborate flux and fused at 1000 °C in a carbon crucible for 30 min. The fusion bead produced was cooled and ground, and a 250 mg portion was dissolved in a mixture of 2 ml HNO3 (Fischer ACS plus grade), 6 ml HF (Baker ACS Reagent), and 2 ml HClO4 (Fischer trace metal grade, Teflon-distilled in-house) and evaporated to dryness at 110 °C. Samples were wetted with a small amount of >18 MΩ H2O and evaporated again at 160 °C with 2 ml HClO4, before redissolution in ∼10 ml H2O, 3 ml HNO3, five drops H2O2, and two drops HF. These solutions were diluted to a final mass of 60 g with H2O for analysis on an Agilent 4500 quadrupole ICP-MS system, and further diluted by an additional factor of 10 in the mass spectrometer’s Integrated Sample Introduction System (ISIS). Ru, In, and Re were used as internal standards to correct for instrumental drift. LREE oxide interferences on the MREE and heavy REE (HREE) were minimized by tuning to keep CeO+/Ce+ < 0·5 %, and the remaining effects were corrected for using two mixed element solutions of Ba, Pr, Nd, and Tb, Sm, Eu, Gd to calculate oxide production rates. Concentrations are calibrated against three in-house standards interspersed with the samples. Accuracy and repeatability of the analyses were assessed by analysing three replicates each of the BCR-2 CRM and WGB-1 candidate reference material (Table 2). All trace elements measured were repeatable to <5 % (2σrel) in BCR-2. Repeatabilities in WGB-1 were slightly poorer, possibly due to lower trace element concentrations. However, all elements analysed showed repeatabilities <10 % (2σrel), with the exception of U, which had a repeatability of <20 %. These data are in agreement with the long-term precision of the method, previously estimated to be better than 10 % (2σrel) for the REE, and better than 20 % for the other trace elements (Knaack et al., 1994). At this level of precision, all elements analysed in BCR-2 overlap the GeoReM preferred values for BCR-2 (Jochum et al., 2005, 2016) within two standard deviations (2σabs), and so the data are considered accurate. Although previous studies reporting high-quality trace element data are not as abundant for WGB-1, all elements analysed overlap the values of Meisel et al. (2002) where available, providing further evidence that the trace element data are accurate at the level of precision indicated by Knaack et al. (1994). Table 2: BCR-2 and WGB-1 trace element data (values in ppm) from this study compared with GeoReM preferred values (Jochum et al., 2005; accessed March 2018) and average values from Meisel et al. (2002), respectively. . BCR-2 . 2σ abs. . 2σ rel. (%) . GeoReM BCR-2 . WGB-1 . 2σ abs. . 2σ rel. (%) . Meisel et al. (2002) . Sc 32·61 0·43 2·8 33·53 39·5 2·9 5·5 Rb 44·99 0·28 2·5 46·02 17·3 0·7 4·5 Sr 334·4 8·0 2·9 337·4 108 6 3·6 Y 35·22 0·86 2·4 36·07 14·9 0·7 6·2 15·1 Zr 179·2 5·1 3·0 186·5 53·7 2·8 6·3 53·2 Nb 11·35 0·33 3·0 12·44 5·72 0·26 7·0 5·9 Cs 1·11 0·04 3·4 1·16 0·43 0·04 4·2 Ba 667 20 4·7 683·9 825 40 8·8 La 25·46 0·71 2·7 25·08 7·83 0·43 5·1 7·8 Ce 52·9 1·3 2·5 53·12 16·2 0·7 4·2 16·5 Pr 6·89 0·20 2·8 6·827 2·23 0·08 3·6 2·23 Nd 28·48 0·69 3·5 28·26 9·80 0·60 3·9 10·2 Sm 6·89 0·21 4·2 6·547 2·69 0·17 0·9 2·69 Eu 2·11 0·06 0·5 1·989 1·30 0·09 3·6 1·27 Gd 6·94 0·24 3·0 6·811 2·88 0·12 4·8 2·94 Tb 1·14 0·05 2·3 1·077 0·49 0·04 7·7 0·47 Dy 6·97 0·19 2·9 6·424 3·03 0·15 4·6 2·92 Ho 1·40 0·03 2·4 1·313 0·60 0·03 4·5 0·58 Er 3·74 0·10 3·2 3·67 1·58 0·06 6·4 1·65 Tm 0·54 0·02 1·7 0·5341 0·22 0·01 5·8 0·227 Yb 3·28 0·14 2·9 3·392 1·36 0·01 18·9 1·43 Lu 0·510 0·002 2·2 0·5049 0·21 0·01 5·7 0·219 Hf 4·84 0·16 0·6 4·972 1·53 0·10 3·9 1·61 Ta 0·78 0·01 3·3 0·785 0·35 0·02 8·6 0·38 Pb 10·17 0·23 2·4 10·59 6·97 0·40 5·9 Th 6·00 0·14 1·3 5·828 1·10 0·08 7·3 1·05 U 1·67 0·05 2·8 1·683 0·74 0·14 5·2 . BCR-2 . 2σ abs. . 2σ rel. (%) . GeoReM BCR-2 . WGB-1 . 2σ abs. . 2σ rel. (%) . Meisel et al. (2002) . Sc 32·61 0·43 2·8 33·53 39·5 2·9 5·5 Rb 44·99 0·28 2·5 46·02 17·3 0·7 4·5 Sr 334·4 8·0 2·9 337·4 108 6 3·6 Y 35·22 0·86 2·4 36·07 14·9 0·7 6·2 15·1 Zr 179·2 5·1 3·0 186·5 53·7 2·8 6·3 53·2 Nb 11·35 0·33 3·0 12·44 5·72 0·26 7·0 5·9 Cs 1·11 0·04 3·4 1·16 0·43 0·04 4·2 Ba 667 20 4·7 683·9 825 40 8·8 La 25·46 0·71 2·7 25·08 7·83 0·43 5·1 7·8 Ce 52·9 1·3 2·5 53·12 16·2 0·7 4·2 16·5 Pr 6·89 0·20 2·8 6·827 2·23 0·08 3·6 2·23 Nd 28·48 0·69 3·5 28·26 9·80 0·60 3·9 10·2 Sm 6·89 0·21 4·2 6·547 2·69 0·17 0·9 2·69 Eu 2·11 0·06 0·5 1·989 1·30 0·09 3·6 1·27 Gd 6·94 0·24 3·0 6·811 2·88 0·12 4·8 2·94 Tb 1·14 0·05 2·3 1·077 0·49 0·04 7·7 0·47 Dy 6·97 0·19 2·9 6·424 3·03 0·15 4·6 2·92 Ho 1·40 0·03 2·4 1·313 0·60 0·03 4·5 0·58 Er 3·74 0·10 3·2 3·67 1·58 0·06 6·4 1·65 Tm 0·54 0·02 1·7 0·5341 0·22 0·01 5·8 0·227 Yb 3·28 0·14 2·9 3·392 1·36 0·01 18·9 1·43 Lu 0·510 0·002 2·2 0·5049 0·21 0·01 5·7 0·219 Hf 4·84 0·16 0·6 4·972 1·53 0·10 3·9 1·61 Ta 0·78 0·01 3·3 0·785 0·35 0·02 8·6 0·38 Pb 10·17 0·23 2·4 10·59 6·97 0·40 5·9 Th 6·00 0·14 1·3 5·828 1·10 0·08 7·3 1·05 U 1·67 0·05 2·8 1·683 0·74 0·14 5·2 Open in new tab Table 2: BCR-2 and WGB-1 trace element data (values in ppm) from this study compared with GeoReM preferred values (Jochum et al., 2005; accessed March 2018) and average values from Meisel et al. (2002), respectively. . BCR-2 . 2σ abs. . 2σ rel. (%) . GeoReM BCR-2 . WGB-1 . 2σ abs. . 2σ rel. (%) . Meisel et al. (2002) . Sc 32·61 0·43 2·8 33·53 39·5 2·9 5·5 Rb 44·99 0·28 2·5 46·02 17·3 0·7 4·5 Sr 334·4 8·0 2·9 337·4 108 6 3·6 Y 35·22 0·86 2·4 36·07 14·9 0·7 6·2 15·1 Zr 179·2 5·1 3·0 186·5 53·7 2·8 6·3 53·2 Nb 11·35 0·33 3·0 12·44 5·72 0·26 7·0 5·9 Cs 1·11 0·04 3·4 1·16 0·43 0·04 4·2 Ba 667 20 4·7 683·9 825 40 8·8 La 25·46 0·71 2·7 25·08 7·83 0·43 5·1 7·8 Ce 52·9 1·3 2·5 53·12 16·2 0·7 4·2 16·5 Pr 6·89 0·20 2·8 6·827 2·23 0·08 3·6 2·23 Nd 28·48 0·69 3·5 28·26 9·80 0·60 3·9 10·2 Sm 6·89 0·21 4·2 6·547 2·69 0·17 0·9 2·69 Eu 2·11 0·06 0·5 1·989 1·30 0·09 3·6 1·27 Gd 6·94 0·24 3·0 6·811 2·88 0·12 4·8 2·94 Tb 1·14 0·05 2·3 1·077 0·49 0·04 7·7 0·47 Dy 6·97 0·19 2·9 6·424 3·03 0·15 4·6 2·92 Ho 1·40 0·03 2·4 1·313 0·60 0·03 4·5 0·58 Er 3·74 0·10 3·2 3·67 1·58 0·06 6·4 1·65 Tm 0·54 0·02 1·7 0·5341 0·22 0·01 5·8 0·227 Yb 3·28 0·14 2·9 3·392 1·36 0·01 18·9 1·43 Lu 0·510 0·002 2·2 0·5049 0·21 0·01 5·7 0·219 Hf 4·84 0·16 0·6 4·972 1·53 0·10 3·9 1·61 Ta 0·78 0·01 3·3 0·785 0·35 0·02 8·6 0·38 Pb 10·17 0·23 2·4 10·59 6·97 0·40 5·9 Th 6·00 0·14 1·3 5·828 1·10 0·08 7·3 1·05 U 1·67 0·05 2·8 1·683 0·74 0·14 5·2 . BCR-2 . 2σ abs. . 2σ rel. (%) . GeoReM BCR-2 . WGB-1 . 2σ abs. . 2σ rel. (%) . Meisel et al. (2002) . Sc 32·61 0·43 2·8 33·53 39·5 2·9 5·5 Rb 44·99 0·28 2·5 46·02 17·3 0·7 4·5 Sr 334·4 8·0 2·9 337·4 108 6 3·6 Y 35·22 0·86 2·4 36·07 14·9 0·7 6·2 15·1 Zr 179·2 5·1 3·0 186·5 53·7 2·8 6·3 53·2 Nb 11·35 0·33 3·0 12·44 5·72 0·26 7·0 5·9 Cs 1·11 0·04 3·4 1·16 0·43 0·04 4·2 Ba 667 20 4·7 683·9 825 40 8·8 La 25·46 0·71 2·7 25·08 7·83 0·43 5·1 7·8 Ce 52·9 1·3 2·5 53·12 16·2 0·7 4·2 16·5 Pr 6·89 0·20 2·8 6·827 2·23 0·08 3·6 2·23 Nd 28·48 0·69 3·5 28·26 9·80 0·60 3·9 10·2 Sm 6·89 0·21 4·2 6·547 2·69 0·17 0·9 2·69 Eu 2·11 0·06 0·5 1·989 1·30 0·09 3·6 1·27 Gd 6·94 0·24 3·0 6·811 2·88 0·12 4·8 2·94 Tb 1·14 0·05 2·3 1·077 0·49 0·04 7·7 0·47 Dy 6·97 0·19 2·9 6·424 3·03 0·15 4·6 2·92 Ho 1·40 0·03 2·4 1·313 0·60 0·03 4·5 0·58 Er 3·74 0·10 3·2 3·67 1·58 0·06 6·4 1·65 Tm 0·54 0·02 1·7 0·5341 0·22 0·01 5·8 0·227 Yb 3·28 0·14 2·9 3·392 1·36 0·01 18·9 1·43 Lu 0·510 0·002 2·2 0·5049 0·21 0·01 5·7 0·219 Hf 4·84 0·16 0·6 4·972 1·53 0·10 3·9 1·61 Ta 0·78 0·01 3·3 0·785 0·35 0·02 8·6 0·38 Pb 10·17 0·23 2·4 10·59 6·97 0·40 5·9 Th 6·00 0·14 1·3 5·828 1·10 0·08 7·3 1·05 U 1·67 0·05 2·8 1·683 0·74 0·14 5·2 Open in new tab EPMA Clinopyroxene analyses for massive komatiite samples from borehole RP1A and differentiated komatiite samples from borehole RP12, and chromite analyses for a dunite sample from borehole RP7, were carried out by electron probe microanalysis (EPMA) on a JEOL 8900R electron microprobe at the University of Alberta using the ‘routine’ analysis method described by Waterton et al. (2017). SiO2, TiO2, Al2O3, Cr2O3, MnO, FeO, CoO, NiO, MgO, CaO, Na2O, K2O, and P2O5 were analysed with 30 s on peak and 30 s on background for most elements, at a beam current of 20 nA and an accelerating voltage of 20 kV. Due to the micron-scale size of some of the clinopyroxene dendrites (particularly the smallest ‘type 2’ sample dendrites), there was a risk of the EPMA interaction volume partially sampling devitrified glass interstitial to the pyroxene dendrites. The data were therefore screened to exclude samples that contained more than two Si cations per formula unit (two Si cations per six oxygens), as these are not consistent with stoichiometric pyroxene compositions. The data were additionally screened to exclude points with totals >101 wt% or <99 wt%. RESULTS RP7 dunite whole-rock geochemistry Dunites from borehole RP7 are highly magnesian, with anhydrous MgO contents between ∼46 and 48 wt% (Fig. 5; Table 3). They contain low concentrations of oxides incompatible in olivine such as Al2O3 and CaO. LOI values are consistently >13 wt%, reflecting the high H2O contents of the metamorphic mineral assemblage. Their bulk-rock compositions fall close to an array of modelled olivine compositions on a plot of FeOt against MgO, and overlap the range of analysed olivine core Mg# from RP1A komatiites (Mg# = 0·87–0·92; Waterton et al., 2017). However, several dunite samples fall slightly off this array, towards lower MgO concentrations, and show a negative correlation towards high Al2O3 concentrations at lower MgO contents. The relationship between MgO and Cr2O3 is scattered, with Cr2O3 contents up to ∼1 wt%. Fig. 5. Open in new tabDownload slide Variations of selected major elements (a–c) and trace elements (d–f) in the RP7 dunites. Dunite data are compared with model olivine compositions with Mg# between 0·87 and 0·93 (black line and thick bars; calculated stoichiometrically assuming 1 wt% minor element content), and compositions of fresh olivines measured from the RP1A komatiites. In (d)–(f), measured olivine MgO contents are shown at an assumed trace element content of zero. Coloured lines show mixing trends between average Winnipegosis komatiite olivine and dunite chromites (ol-chromite), and between a range of possible accumulating assemblage compositions (Cumulate) and both Winnipegosis komatiite parental melt (Mix-komatiite) and Winnipegosis basalts (Mix-basalt), intended to represent the effect of trapped interstitial melt. Crosses with percentages indicate the amount of komatiitic or basaltic trapped melt in the mixture. Coloured arrows are vectors representing the geochemical effects of various processes. Fig. 5. Open in new tabDownload slide Variations of selected major elements (a–c) and trace elements (d–f) in the RP7 dunites. Dunite data are compared with model olivine compositions with Mg# between 0·87 and 0·93 (black line and thick bars; calculated stoichiometrically assuming 1 wt% minor element content), and compositions of fresh olivines measured from the RP1A komatiites. In (d)–(f), measured olivine MgO contents are shown at an assumed trace element content of zero. Coloured lines show mixing trends between average Winnipegosis komatiite olivine and dunite chromites (ol-chromite), and between a range of possible accumulating assemblage compositions (Cumulate) and both Winnipegosis komatiite parental melt (Mix-komatiite) and Winnipegosis basalts (Mix-basalt), intended to represent the effect of trapped interstitial melt. Crosses with percentages indicate the amount of komatiitic or basaltic trapped melt in the mixture. Coloured arrows are vectors representing the geochemical effects of various processes. Table 3: Bulk-rock major and minor element data for dunites from borehole RP7 . RP7-1 . RP7-3 . RP7-4 . RP7-5 . RP7-6 . RP7-7 . RP7-8 . RP7-9 . RP7-10 . RP7-11 . RP7-12 . RP7-13 . RP7-14 . RP7-15 . RP7-16 . SiO2 (wt%) 40·71 40·18 39·85 40·08 39·68 40·79 40·93 41·03 40·82 40·98 40·82 39·74 39·99 40·08 41·93 TiO2 0·04 0·04 0·03 0·03 0·02 0·03 0·03 0·02 0·03 0·03 0·03 0·03 0·03 0·05 0·06 Al2O3 1·10 0·96 0·81 0·98 0·81 1·13 1·06 0·86 1·17 0·92 0·85 0·85 1·12 1·25 1·62 FeOt 10·54 10·28 11·26 11·12 11·28 9·72 9·49 9·22 9·44 9·27 9·58 11·54 11·66 12·07 10·03 MnO 0·07 0·13 0·14 0·21 0·21 0·28 0·24 0·21 0·18 0·18 0·18 0·22 0·22 0·25 0·17 MgO 47·01 47·78 47·64 47·37 47·77 47·82 48·02 48·39 48·04 48·36 48·22 47·37 46·72 46·00 45·92 CaO 0·33 0·45 0·09 0·09 0·09 0·08 0·09 0·12 0·13 0·11 0·13 0·11 0·11 0·13 0·13 Na2O 0·15 0·15 0·16 0·11 0·12 0·12 0·12 0·13 0·18 0·12 0·16 0·12 0·11 0·15 0·12 K2O 0·03 0·02 0·02 0·01 0·01 0·02 0·02 0·01 0·01 0·03 0·03 0·02 0·02 0·02 0·02 P2O5 0·01 0·01 0·01 0·01 0·01 0·01 0·01 0·01 0·00 0·01 0·00 0·00 0·01 0·01 0·01 Total (raw) 98·20 98·37 98·67 98·46 98·72 98·55 98·43 98·32 98·27 98·60 98·55 98·76 98·93 98·47 98·45 LOI 14·71 15·03 15·14 15·02 15·40 15·27 15·25 15·36 15·44 15·36 15·53 15·28 14·97 14·72 13·61 Cr (ppm) 5320 6922 7153 6888 6799 5720 5268 6776 6678 6679 6743 6208 6063 6398 5857 Co 166 164 159 154 175 166 138 162 179 160 172 170 171 174 64 La 0·67 0·22 0·21 0·18 0·32 0·21 0·19 0·14 0·14 0·13 0·23 Ce 0·84 0·68 0·38 0·41 1·14 0·78 0·40 0·34 0·35 0·36 0·60 Pr 0·08 0·12 0·05 0·05 0·18 0·12 0·05 0·05 0·05 0·05 0·09 Nd 0·32 0·58 0·24 0·19 0·81 0·67 0·23 0·21 0·25 0·27 0·46 Sm 0·09 0·24 0·10 0·05 0·19 0·16 0·08 0·08 0·08 0·09 0·13 Eu 0·04 0·02 0·03 0·02 0·03 0·02 0·03 0·04 0·03 0·03 0·06 Gd 0·11 0·33 0·14 0·07 0·22 0·21 0·07 0·09 0·10 0·09 0·17 Tb 0·02 0·04 0·03 0·01 0·03 0·04 0·01 0·02 0·02 0·02 0·03 Dy 0·13 0·18 0·23 0·07 0·20 0·27 0·08 0·10 0·12 0·10 0·23 Ho 0·03 0·03 0·06 0·01 0·04 0·06 0·02 0·02 0·03 0·02 0·05 Er 0·08 0·08 0·17 0·04 0·10 0·18 0·05 0·07 0·07 0·06 0·14 Tm 0·01 0·01 0·02 0·01 0·01 0·02 0·01 0·01 0·01 0·01 0·02 Yb 0·08 0·07 0·13 0·06 0·08 0·14 0·06 0·07 0·07 0·07 0·13 Lu 0·02 0·01 0·02 0·01 0·01 0·02 0·01 0·01 0·01 0·01 0·02 Ba 4·78 1·06 0·70 0·58 0·74 0·62 2·63 0·96 1·84 9·48 1·24 Th 0·06 0·04 0·03 0·03 0·03 0·11 0·04 0·04 0·04 0·04 0·04 Nb 0·10 0·07 0·12 0·04 0·18 0·12 0·05 0·06 0·06 0·05 0·12 Y 0·76 1·17 1·39 0·41 1·15 1·69 0·45 0·52 0·60 0·57 1·17 Hf 0·06 0·05 0·07 0·03 0·10 0·08 0·04 0·04 0·05 0·03 0·07 Ta 0·01 0·01 0·01 0·00 0·01 0·01 0·00 0·01 0·00 0·00 0·01 U 0·02 0·01 0·02 0·01 0·03 0·02 0·01 0·01 0·01 0·01 0·01 Pb 6·82 3·54 3·17 2·79 2·12 1·92 2·45 2·46 3·42 4·39 1·74 Rb 0·61 0·17 0·13 0·13 0·15 0·29 0·14 0·17 0·24 0·32 0·28 Cs 0·03 0·01 0·01 0·01 0·01 0·03 0·01 0·01 0·01 0·03 0·02 Sr 15·09 10·84 9·53 9·85 10·05 10·97 12·31 12·66 13·39 13·44 14·79 Sc 5·94 6·17 5·81 5·66 5·82 5·65 5·45 6·02 6·38 6·59 8·11 Zr 2·32 1·70 2·16 1·10 2·49 2·15 1·34 1·53 1·86 1·25 2·60 . RP7-1 . RP7-3 . RP7-4 . RP7-5 . RP7-6 . RP7-7 . RP7-8 . RP7-9 . RP7-10 . RP7-11 . RP7-12 . RP7-13 . RP7-14 . RP7-15 . RP7-16 . SiO2 (wt%) 40·71 40·18 39·85 40·08 39·68 40·79 40·93 41·03 40·82 40·98 40·82 39·74 39·99 40·08 41·93 TiO2 0·04 0·04 0·03 0·03 0·02 0·03 0·03 0·02 0·03 0·03 0·03 0·03 0·03 0·05 0·06 Al2O3 1·10 0·96 0·81 0·98 0·81 1·13 1·06 0·86 1·17 0·92 0·85 0·85 1·12 1·25 1·62 FeOt 10·54 10·28 11·26 11·12 11·28 9·72 9·49 9·22 9·44 9·27 9·58 11·54 11·66 12·07 10·03 MnO 0·07 0·13 0·14 0·21 0·21 0·28 0·24 0·21 0·18 0·18 0·18 0·22 0·22 0·25 0·17 MgO 47·01 47·78 47·64 47·37 47·77 47·82 48·02 48·39 48·04 48·36 48·22 47·37 46·72 46·00 45·92 CaO 0·33 0·45 0·09 0·09 0·09 0·08 0·09 0·12 0·13 0·11 0·13 0·11 0·11 0·13 0·13 Na2O 0·15 0·15 0·16 0·11 0·12 0·12 0·12 0·13 0·18 0·12 0·16 0·12 0·11 0·15 0·12 K2O 0·03 0·02 0·02 0·01 0·01 0·02 0·02 0·01 0·01 0·03 0·03 0·02 0·02 0·02 0·02 P2O5 0·01 0·01 0·01 0·01 0·01 0·01 0·01 0·01 0·00 0·01 0·00 0·00 0·01 0·01 0·01 Total (raw) 98·20 98·37 98·67 98·46 98·72 98·55 98·43 98·32 98·27 98·60 98·55 98·76 98·93 98·47 98·45 LOI 14·71 15·03 15·14 15·02 15·40 15·27 15·25 15·36 15·44 15·36 15·53 15·28 14·97 14·72 13·61 Cr (ppm) 5320 6922 7153 6888 6799 5720 5268 6776 6678 6679 6743 6208 6063 6398 5857 Co 166 164 159 154 175 166 138 162 179 160 172 170 171 174 64 La 0·67 0·22 0·21 0·18 0·32 0·21 0·19 0·14 0·14 0·13 0·23 Ce 0·84 0·68 0·38 0·41 1·14 0·78 0·40 0·34 0·35 0·36 0·60 Pr 0·08 0·12 0·05 0·05 0·18 0·12 0·05 0·05 0·05 0·05 0·09 Nd 0·32 0·58 0·24 0·19 0·81 0·67 0·23 0·21 0·25 0·27 0·46 Sm 0·09 0·24 0·10 0·05 0·19 0·16 0·08 0·08 0·08 0·09 0·13 Eu 0·04 0·02 0·03 0·02 0·03 0·02 0·03 0·04 0·03 0·03 0·06 Gd 0·11 0·33 0·14 0·07 0·22 0·21 0·07 0·09 0·10 0·09 0·17 Tb 0·02 0·04 0·03 0·01 0·03 0·04 0·01 0·02 0·02 0·02 0·03 Dy 0·13 0·18 0·23 0·07 0·20 0·27 0·08 0·10 0·12 0·10 0·23 Ho 0·03 0·03 0·06 0·01 0·04 0·06 0·02 0·02 0·03 0·02 0·05 Er 0·08 0·08 0·17 0·04 0·10 0·18 0·05 0·07 0·07 0·06 0·14 Tm 0·01 0·01 0·02 0·01 0·01 0·02 0·01 0·01 0·01 0·01 0·02 Yb 0·08 0·07 0·13 0·06 0·08 0·14 0·06 0·07 0·07 0·07 0·13 Lu 0·02 0·01 0·02 0·01 0·01 0·02 0·01 0·01 0·01 0·01 0·02 Ba 4·78 1·06 0·70 0·58 0·74 0·62 2·63 0·96 1·84 9·48 1·24 Th 0·06 0·04 0·03 0·03 0·03 0·11 0·04 0·04 0·04 0·04 0·04 Nb 0·10 0·07 0·12 0·04 0·18 0·12 0·05 0·06 0·06 0·05 0·12 Y 0·76 1·17 1·39 0·41 1·15 1·69 0·45 0·52 0·60 0·57 1·17 Hf 0·06 0·05 0·07 0·03 0·10 0·08 0·04 0·04 0·05 0·03 0·07 Ta 0·01 0·01 0·01 0·00 0·01 0·01 0·00 0·01 0·00 0·00 0·01 U 0·02 0·01 0·02 0·01 0·03 0·02 0·01 0·01 0·01 0·01 0·01 Pb 6·82 3·54 3·17 2·79 2·12 1·92 2·45 2·46 3·42 4·39 1·74 Rb 0·61 0·17 0·13 0·13 0·15 0·29 0·14 0·17 0·24 0·32 0·28 Cs 0·03 0·01 0·01 0·01 0·01 0·03 0·01 0·01 0·01 0·03 0·02 Sr 15·09 10·84 9·53 9·85 10·05 10·97 12·31 12·66 13·39 13·44 14·79 Sc 5·94 6·17 5·81 5·66 5·82 5·65 5·45 6·02 6·38 6·59 8·11 Zr 2·32 1·70 2·16 1·10 2·49 2·15 1·34 1·53 1·86 1·25 2·60 Data for oxides, Cr, and Co were renormalized to 100 % totals on an anhydrous basis, with FeOt calculated assuming all Fe is present as FeO. Loss on ignition (LOI) was not renormalized. Raw totals are given as an indication of data quality. Open in new tab Table 3: Bulk-rock major and minor element data for dunites from borehole RP7 . RP7-1 . RP7-3 . RP7-4 . RP7-5 . RP7-6 . RP7-7 . RP7-8 . RP7-9 . RP7-10 . RP7-11 . RP7-12 . RP7-13 . RP7-14 . RP7-15 . RP7-16 . SiO2 (wt%) 40·71 40·18 39·85 40·08 39·68 40·79 40·93 41·03 40·82 40·98 40·82 39·74 39·99 40·08 41·93 TiO2 0·04 0·04 0·03 0·03 0·02 0·03 0·03 0·02 0·03 0·03 0·03 0·03 0·03 0·05 0·06 Al2O3 1·10 0·96 0·81 0·98 0·81 1·13 1·06 0·86 1·17 0·92 0·85 0·85 1·12 1·25 1·62 FeOt 10·54 10·28 11·26 11·12 11·28 9·72 9·49 9·22 9·44 9·27 9·58 11·54 11·66 12·07 10·03 MnO 0·07 0·13 0·14 0·21 0·21 0·28 0·24 0·21 0·18 0·18 0·18 0·22 0·22 0·25 0·17 MgO 47·01 47·78 47·64 47·37 47·77 47·82 48·02 48·39 48·04 48·36 48·22 47·37 46·72 46·00 45·92 CaO 0·33 0·45 0·09 0·09 0·09 0·08 0·09 0·12 0·13 0·11 0·13 0·11 0·11 0·13 0·13 Na2O 0·15 0·15 0·16 0·11 0·12 0·12 0·12 0·13 0·18 0·12 0·16 0·12 0·11 0·15 0·12 K2O 0·03 0·02 0·02 0·01 0·01 0·02 0·02 0·01 0·01 0·03 0·03 0·02 0·02 0·02 0·02 P2O5 0·01 0·01 0·01 0·01 0·01 0·01 0·01 0·01 0·00 0·01 0·00 0·00 0·01 0·01 0·01 Total (raw) 98·20 98·37 98·67 98·46 98·72 98·55 98·43 98·32 98·27 98·60 98·55 98·76 98·93 98·47 98·45 LOI 14·71 15·03 15·14 15·02 15·40 15·27 15·25 15·36 15·44 15·36 15·53 15·28 14·97 14·72 13·61 Cr (ppm) 5320 6922 7153 6888 6799 5720 5268 6776 6678 6679 6743 6208 6063 6398 5857 Co 166 164 159 154 175 166 138 162 179 160 172 170 171 174 64 La 0·67 0·22 0·21 0·18 0·32 0·21 0·19 0·14 0·14 0·13 0·23 Ce 0·84 0·68 0·38 0·41 1·14 0·78 0·40 0·34 0·35 0·36 0·60 Pr 0·08 0·12 0·05 0·05 0·18 0·12 0·05 0·05 0·05 0·05 0·09 Nd 0·32 0·58 0·24 0·19 0·81 0·67 0·23 0·21 0·25 0·27 0·46 Sm 0·09 0·24 0·10 0·05 0·19 0·16 0·08 0·08 0·08 0·09 0·13 Eu 0·04 0·02 0·03 0·02 0·03 0·02 0·03 0·04 0·03 0·03 0·06 Gd 0·11 0·33 0·14 0·07 0·22 0·21 0·07 0·09 0·10 0·09 0·17 Tb 0·02 0·04 0·03 0·01 0·03 0·04 0·01 0·02 0·02 0·02 0·03 Dy 0·13 0·18 0·23 0·07 0·20 0·27 0·08 0·10 0·12 0·10 0·23 Ho 0·03 0·03 0·06 0·01 0·04 0·06 0·02 0·02 0·03 0·02 0·05 Er 0·08 0·08 0·17 0·04 0·10 0·18 0·05 0·07 0·07 0·06 0·14 Tm 0·01 0·01 0·02 0·01 0·01 0·02 0·01 0·01 0·01 0·01 0·02 Yb 0·08 0·07 0·13 0·06 0·08 0·14 0·06 0·07 0·07 0·07 0·13 Lu 0·02 0·01 0·02 0·01 0·01 0·02 0·01 0·01 0·01 0·01 0·02 Ba 4·78 1·06 0·70 0·58 0·74 0·62 2·63 0·96 1·84 9·48 1·24 Th 0·06 0·04 0·03 0·03 0·03 0·11 0·04 0·04 0·04 0·04 0·04 Nb 0·10 0·07 0·12 0·04 0·18 0·12 0·05 0·06 0·06 0·05 0·12 Y 0·76 1·17 1·39 0·41 1·15 1·69 0·45 0·52 0·60 0·57 1·17 Hf 0·06 0·05 0·07 0·03 0·10 0·08 0·04 0·04 0·05 0·03 0·07 Ta 0·01 0·01 0·01 0·00 0·01 0·01 0·00 0·01 0·00 0·00 0·01 U 0·02 0·01 0·02 0·01 0·03 0·02 0·01 0·01 0·01 0·01 0·01 Pb 6·82 3·54 3·17 2·79 2·12 1·92 2·45 2·46 3·42 4·39 1·74 Rb 0·61 0·17 0·13 0·13 0·15 0·29 0·14 0·17 0·24 0·32 0·28 Cs 0·03 0·01 0·01 0·01 0·01 0·03 0·01 0·01 0·01 0·03 0·02 Sr 15·09 10·84 9·53 9·85 10·05 10·97 12·31 12·66 13·39 13·44 14·79 Sc 5·94 6·17 5·81 5·66 5·82 5·65 5·45 6·02 6·38 6·59 8·11 Zr 2·32 1·70 2·16 1·10 2·49 2·15 1·34 1·53 1·86 1·25 2·60 . RP7-1 . RP7-3 . RP7-4 . RP7-5 . RP7-6 . RP7-7 . RP7-8 . RP7-9 . RP7-10 . RP7-11 . RP7-12 . RP7-13 . RP7-14 . RP7-15 . RP7-16 . SiO2 (wt%) 40·71 40·18 39·85 40·08 39·68 40·79 40·93 41·03 40·82 40·98 40·82 39·74 39·99 40·08 41·93 TiO2 0·04 0·04 0·03 0·03 0·02 0·03 0·03 0·02 0·03 0·03 0·03 0·03 0·03 0·05 0·06 Al2O3 1·10 0·96 0·81 0·98 0·81 1·13 1·06 0·86 1·17 0·92 0·85 0·85 1·12 1·25 1·62 FeOt 10·54 10·28 11·26 11·12 11·28 9·72 9·49 9·22 9·44 9·27 9·58 11·54 11·66 12·07 10·03 MnO 0·07 0·13 0·14 0·21 0·21 0·28 0·24 0·21 0·18 0·18 0·18 0·22 0·22 0·25 0·17 MgO 47·01 47·78 47·64 47·37 47·77 47·82 48·02 48·39 48·04 48·36 48·22 47·37 46·72 46·00 45·92 CaO 0·33 0·45 0·09 0·09 0·09 0·08 0·09 0·12 0·13 0·11 0·13 0·11 0·11 0·13 0·13 Na2O 0·15 0·15 0·16 0·11 0·12 0·12 0·12 0·13 0·18 0·12 0·16 0·12 0·11 0·15 0·12 K2O 0·03 0·02 0·02 0·01 0·01 0·02 0·02 0·01 0·01 0·03 0·03 0·02 0·02 0·02 0·02 P2O5 0·01 0·01 0·01 0·01 0·01 0·01 0·01 0·01 0·00 0·01 0·00 0·00 0·01 0·01 0·01 Total (raw) 98·20 98·37 98·67 98·46 98·72 98·55 98·43 98·32 98·27 98·60 98·55 98·76 98·93 98·47 98·45 LOI 14·71 15·03 15·14 15·02 15·40 15·27 15·25 15·36 15·44 15·36 15·53 15·28 14·97 14·72 13·61 Cr (ppm) 5320 6922 7153 6888 6799 5720 5268 6776 6678 6679 6743 6208 6063 6398 5857 Co 166 164 159 154 175 166 138 162 179 160 172 170 171 174 64 La 0·67 0·22 0·21 0·18 0·32 0·21 0·19 0·14 0·14 0·13 0·23 Ce 0·84 0·68 0·38 0·41 1·14 0·78 0·40 0·34 0·35 0·36 0·60 Pr 0·08 0·12 0·05 0·05 0·18 0·12 0·05 0·05 0·05 0·05 0·09 Nd 0·32 0·58 0·24 0·19 0·81 0·67 0·23 0·21 0·25 0·27 0·46 Sm 0·09 0·24 0·10 0·05 0·19 0·16 0·08 0·08 0·08 0·09 0·13 Eu 0·04 0·02 0·03 0·02 0·03 0·02 0·03 0·04 0·03 0·03 0·06 Gd 0·11 0·33 0·14 0·07 0·22 0·21 0·07 0·09 0·10 0·09 0·17 Tb 0·02 0·04 0·03 0·01 0·03 0·04 0·01 0·02 0·02 0·02 0·03 Dy 0·13 0·18 0·23 0·07 0·20 0·27 0·08 0·10 0·12 0·10 0·23 Ho 0·03 0·03 0·06 0·01 0·04 0·06 0·02 0·02 0·03 0·02 0·05 Er 0·08 0·08 0·17 0·04 0·10 0·18 0·05 0·07 0·07 0·06 0·14 Tm 0·01 0·01 0·02 0·01 0·01 0·02 0·01 0·01 0·01 0·01 0·02 Yb 0·08 0·07 0·13 0·06 0·08 0·14 0·06 0·07 0·07 0·07 0·13 Lu 0·02 0·01 0·02 0·01 0·01 0·02 0·01 0·01 0·01 0·01 0·02 Ba 4·78 1·06 0·70 0·58 0·74 0·62 2·63 0·96 1·84 9·48 1·24 Th 0·06 0·04 0·03 0·03 0·03 0·11 0·04 0·04 0·04 0·04 0·04 Nb 0·10 0·07 0·12 0·04 0·18 0·12 0·05 0·06 0·06 0·05 0·12 Y 0·76 1·17 1·39 0·41 1·15 1·69 0·45 0·52 0·60 0·57 1·17 Hf 0·06 0·05 0·07 0·03 0·10 0·08 0·04 0·04 0·05 0·03 0·07 Ta 0·01 0·01 0·01 0·00 0·01 0·01 0·00 0·01 0·00 0·00 0·01 U 0·02 0·01 0·02 0·01 0·03 0·02 0·01 0·01 0·01 0·01 0·01 Pb 6·82 3·54 3·17 2·79 2·12 1·92 2·45 2·46 3·42 4·39 1·74 Rb 0·61 0·17 0·13 0·13 0·15 0·29 0·14 0·17 0·24 0·32 0·28 Cs 0·03 0·01 0·01 0·01 0·01 0·03 0·01 0·01 0·01 0·03 0·02 Sr 15·09 10·84 9·53 9·85 10·05 10·97 12·31 12·66 13·39 13·44 14·79 Sc 5·94 6·17 5·81 5·66 5·82 5·65 5·45 6·02 6·38 6·59 8·11 Zr 2·32 1·70 2·16 1·10 2·49 2·15 1·34 1·53 1·86 1·25 2·60 Data for oxides, Cr, and Co were renormalized to 100 % totals on an anhydrous basis, with FeOt calculated assuming all Fe is present as FeO. Loss on ignition (LOI) was not renormalized. Raw totals are given as an indication of data quality. Open in new tab Incompatible trace element concentrations in RP7 dunites are generally very low (Table 3), with concentrations of immobile elements of 0·1–1 times primitive upper mantle values (PUM; Hofmann 1988). As observed for Al2O3, many incompatible trace elements in the dunites show negative correlations with MgO content (Fig. 5). RP8 basalt whole-rock geochemistry Basalts from borehole RP8 show limited major element variability (Table 4). Compositionally, they share a number of similarities with typical MORB (Gale et al., 2013), with ∼50 wt% SiO2, ∼7·5 wt% MgO, ∼14 wt% Al2O3, and ∼10 wt% CaO. However, FeOt contents of 13–16 wt% are distinctly elevated relative to MORB (average FeOt ≈ 10 wt%), but well within the range of Phanerozoic continental flood basalts (CFB; e.g. Lightfoot et al., 1990; Wooden et al., 1993; Lange, 2002). TiO2 contents of ∼1·25 wt% are low compared with MORB compositions, but also lie within the range of Phanerozoic CFB (Lightfoot et al., 1990; Wooden et al., 1993). Na2O data show considerable scatter but average contents of ∼2·5 wt% are close to the global average for ∼1·9 Ga basalts (Keller & Schoene, 2012). Table 4: Bulk-rock major, minor, and trace element data for metabasalts from borehole RP8 . RP8-1 . RP8-2 . RP8-3 . RP8-4 . RP8-5 . RP8-6 . RP8-7 av. . RP8-7 2σ . RP8-8 . RP8-9 . RP8-10 . SiO2 (wt%) 51·03 50·18 51·29 48·92 50·33 50·22 50·50 1·59 49·86 50·43 49·53 TiO2 1·24 1·22 1·22 1·12 1·22 1·27 1·27 0·03 1·26 1·21 1·26 Al2O3 13·89 14·06 13·87 13·13 14·13 14·11 13·93 0·47 14·12 14·06 14·35 FeOt 13·14 13·23 12·92 15·89 13·06 13·37 13·36 0·28 13·36 13·22 13·73 MnO 0·177 0·203 0·193 0·244 0·195 0·233 0·204 0·007 0·224 0·193 0·225 MgO 7·52 7·19 7·58 7·71 7·62 7·41 7·41 0·39 7·33 7·75 7·87 CaO 10·48 11·77 10·81 9·95 10·55 10·10 10·36 0·38 10·78 10·17 10·26 Na2O 2·29 1·99 1·93 2·75 2·58 3·02 2·68 0·09 2·80 2·56 2·41 K2O 0·136 0·040 0·104 0·177 0·219 0·148 0·184 0·007 0·157 0·321 0·281 P2O5 0·094 0·114 0·090 0·106 0·090 0·116 0·085 0·007 0·111 0·087 0·092 Total (raw) 99·71 100·01 99·88 100·21 100·07 99·98 99·88 0·23 99·70 99·94 99·38 LOI 2·17 2·08 2·07 1·22 2·02 2·70 2·03 0·19 1·99 2·13 2·08 V (ppm) 362 358 342 364 357 380 369 17 368 362 427 Cr 179 132 158 174 171 151 163 23 150 183 215 Sc 43·6 44·1 43·5 43·0 43·7 45·4 45·6 45·3 44·6 45·2 Rb 2·02 0·56 1·39 2·89 3·50 1·99 2·41 2·26 4·84 3·46 Sr 109·5 132·8 116·0 103·1 118·3 95·7 99·0 100·6 74·7 82·4 Y 24·9 24·6 24·0 21·7 24·0 24·0 23·1 24·4 23·6 20·4 Zr 64·9 65·9 63·7 56·0 63·3 61·6 59·0 63·8 61·3 53·2 Nb 3·16 3·10 3·11 2·92 3·01 3·06 2·93 3·18 3·05 2·50 Cs 0·11 0·05 0·08 0·12 0·12 0·09 0·09 0·10 0·20 0·15 Ba 29·7 11·5 16·7 22·7 30·8 24·5 25·9 24·4 29·7 20·3 La 3·63 3·46 3·31 2·90 3·29 3·17 3·06 3·32 3·28 2·74 Ce 10·16 9·62 9·23 8·07 9·11 8·92 8·50 9·11 9·00 7·68 Pr 1·66 1·53 1·52 1·33 1·48 1·48 1·40 1·50 1·47 1·24 Nd 8·70 7·90 7·94 6·98 7·73 7·72 7·41 7·92 7·60 6·66 Sm 3·10 2·85 2·80 2·45 2·80 2·69 2·57 2·70 2·76 2·34 Eu 1·06 1·04 1·04 0·95 1·03 1·06 1·04 1·06 1·02 0·88 Gd 3·95 3·71 3·77 3·28 3·70 3·71 3·54 3·65 3·60 3·15 Tb 0·74 0·70 0·70 0·63 0·70 0·68 0·66 0·70 0·67 0·59 Dy 4·78 4·65 4·57 4·06 4·59 4·54 4·33 4·56 4·49 3·89 Ho 1·00 0·98 0·97 0·87 0·97 0·96 0·93 0·98 0·94 0·81 Er 2·73 2·75 2·75 2·40 2·67 2·67 2·57 2·70 2·55 2·25 Tm 0·40 0·40 0·40 0·36 0·39 0·39 0·37 0·40 0·38 0·33 Yb 2·48 2·49 2·48 2·19 2·42 2·39 2·33 2·46 2·38 2·03 Lu 0·39 0·40 0·39 0·34 0·37 0·37 0·36 0·39 0·38 0·33 Hf 1·88 1·91 1·85 1·66 1·84 1·80 1·72 1·89 1·76 1·56 Ta 0·22 0·22 0·21 0·20 0·22 0·20 0·20 0·21 0·21 0·18 Pb 1·08 3·66 0·69 0·41 0·95 0·68 0·68 0·70 0·68 0·45 Th 0·29 0·33 0·23 0·21 0·24 0·23 0·22 0·23 0·23 0·19 U 0·080 0·097 0·068 0·060 0·069 0·068 0·060 0·069 0·066 0·056 . RP8-1 . RP8-2 . RP8-3 . RP8-4 . RP8-5 . RP8-6 . RP8-7 av. . RP8-7 2σ . RP8-8 . RP8-9 . RP8-10 . SiO2 (wt%) 51·03 50·18 51·29 48·92 50·33 50·22 50·50 1·59 49·86 50·43 49·53 TiO2 1·24 1·22 1·22 1·12 1·22 1·27 1·27 0·03 1·26 1·21 1·26 Al2O3 13·89 14·06 13·87 13·13 14·13 14·11 13·93 0·47 14·12 14·06 14·35 FeOt 13·14 13·23 12·92 15·89 13·06 13·37 13·36 0·28 13·36 13·22 13·73 MnO 0·177 0·203 0·193 0·244 0·195 0·233 0·204 0·007 0·224 0·193 0·225 MgO 7·52 7·19 7·58 7·71 7·62 7·41 7·41 0·39 7·33 7·75 7·87 CaO 10·48 11·77 10·81 9·95 10·55 10·10 10·36 0·38 10·78 10·17 10·26 Na2O 2·29 1·99 1·93 2·75 2·58 3·02 2·68 0·09 2·80 2·56 2·41 K2O 0·136 0·040 0·104 0·177 0·219 0·148 0·184 0·007 0·157 0·321 0·281 P2O5 0·094 0·114 0·090 0·106 0·090 0·116 0·085 0·007 0·111 0·087 0·092 Total (raw) 99·71 100·01 99·88 100·21 100·07 99·98 99·88 0·23 99·70 99·94 99·38 LOI 2·17 2·08 2·07 1·22 2·02 2·70 2·03 0·19 1·99 2·13 2·08 V (ppm) 362 358 342 364 357 380 369 17 368 362 427 Cr 179 132 158 174 171 151 163 23 150 183 215 Sc 43·6 44·1 43·5 43·0 43·7 45·4 45·6 45·3 44·6 45·2 Rb 2·02 0·56 1·39 2·89 3·50 1·99 2·41 2·26 4·84 3·46 Sr 109·5 132·8 116·0 103·1 118·3 95·7 99·0 100·6 74·7 82·4 Y 24·9 24·6 24·0 21·7 24·0 24·0 23·1 24·4 23·6 20·4 Zr 64·9 65·9 63·7 56·0 63·3 61·6 59·0 63·8 61·3 53·2 Nb 3·16 3·10 3·11 2·92 3·01 3·06 2·93 3·18 3·05 2·50 Cs 0·11 0·05 0·08 0·12 0·12 0·09 0·09 0·10 0·20 0·15 Ba 29·7 11·5 16·7 22·7 30·8 24·5 25·9 24·4 29·7 20·3 La 3·63 3·46 3·31 2·90 3·29 3·17 3·06 3·32 3·28 2·74 Ce 10·16 9·62 9·23 8·07 9·11 8·92 8·50 9·11 9·00 7·68 Pr 1·66 1·53 1·52 1·33 1·48 1·48 1·40 1·50 1·47 1·24 Nd 8·70 7·90 7·94 6·98 7·73 7·72 7·41 7·92 7·60 6·66 Sm 3·10 2·85 2·80 2·45 2·80 2·69 2·57 2·70 2·76 2·34 Eu 1·06 1·04 1·04 0·95 1·03 1·06 1·04 1·06 1·02 0·88 Gd 3·95 3·71 3·77 3·28 3·70 3·71 3·54 3·65 3·60 3·15 Tb 0·74 0·70 0·70 0·63 0·70 0·68 0·66 0·70 0·67 0·59 Dy 4·78 4·65 4·57 4·06 4·59 4·54 4·33 4·56 4·49 3·89 Ho 1·00 0·98 0·97 0·87 0·97 0·96 0·93 0·98 0·94 0·81 Er 2·73 2·75 2·75 2·40 2·67 2·67 2·57 2·70 2·55 2·25 Tm 0·40 0·40 0·40 0·36 0·39 0·39 0·37 0·40 0·38 0·33 Yb 2·48 2·49 2·48 2·19 2·42 2·39 2·33 2·46 2·38 2·03 Lu 0·39 0·40 0·39 0·34 0·37 0·37 0·36 0·39 0·38 0·33 Hf 1·88 1·91 1·85 1·66 1·84 1·80 1·72 1·89 1·76 1·56 Ta 0·22 0·22 0·21 0·20 0·22 0·20 0·20 0·21 0·21 0·18 Pb 1·08 3·66 0·69 0·41 0·95 0·68 0·68 0·70 0·68 0·45 Th 0·29 0·33 0·23 0·21 0·24 0·23 0·22 0·23 0·23 0·19 U 0·080 0·097 0·068 0·060 0·069 0·068 0·060 0·069 0·066 0·056 XRF major and minor element data for oxides, V, and Cr were renormalized to 100 % totals on an anhydrous basis, with FeOt calculated assuming all Fe is present as FeO. Loss on ignition (LOI) and solution ICP-MS trace element data were not renormalized. Raw totals are given as an indication of data quality. XRF data for sample RP8-7 were measured in triplicate and are reported as the average and 2 standard deviations of these three analyses. Open in new tab Table 4: Bulk-rock major, minor, and trace element data for metabasalts from borehole RP8 . RP8-1 . RP8-2 . RP8-3 . RP8-4 . RP8-5 . RP8-6 . RP8-7 av. . RP8-7 2σ . RP8-8 . RP8-9 . RP8-10 . SiO2 (wt%) 51·03 50·18 51·29 48·92 50·33 50·22 50·50 1·59 49·86 50·43 49·53 TiO2 1·24 1·22 1·22 1·12 1·22 1·27 1·27 0·03 1·26 1·21 1·26 Al2O3 13·89 14·06 13·87 13·13 14·13 14·11 13·93 0·47 14·12 14·06 14·35 FeOt 13·14 13·23 12·92 15·89 13·06 13·37 13·36 0·28 13·36 13·22 13·73 MnO 0·177 0·203 0·193 0·244 0·195 0·233 0·204 0·007 0·224 0·193 0·225 MgO 7·52 7·19 7·58 7·71 7·62 7·41 7·41 0·39 7·33 7·75 7·87 CaO 10·48 11·77 10·81 9·95 10·55 10·10 10·36 0·38 10·78 10·17 10·26 Na2O 2·29 1·99 1·93 2·75 2·58 3·02 2·68 0·09 2·80 2·56 2·41 K2O 0·136 0·040 0·104 0·177 0·219 0·148 0·184 0·007 0·157 0·321 0·281 P2O5 0·094 0·114 0·090 0·106 0·090 0·116 0·085 0·007 0·111 0·087 0·092 Total (raw) 99·71 100·01 99·88 100·21 100·07 99·98 99·88 0·23 99·70 99·94 99·38 LOI 2·17 2·08 2·07 1·22 2·02 2·70 2·03 0·19 1·99 2·13 2·08 V (ppm) 362 358 342 364 357 380 369 17 368 362 427 Cr 179 132 158 174 171 151 163 23 150 183 215 Sc 43·6 44·1 43·5 43·0 43·7 45·4 45·6 45·3 44·6 45·2 Rb 2·02 0·56 1·39 2·89 3·50 1·99 2·41 2·26 4·84 3·46 Sr 109·5 132·8 116·0 103·1 118·3 95·7 99·0 100·6 74·7 82·4 Y 24·9 24·6 24·0 21·7 24·0 24·0 23·1 24·4 23·6 20·4 Zr 64·9 65·9 63·7 56·0 63·3 61·6 59·0 63·8 61·3 53·2 Nb 3·16 3·10 3·11 2·92 3·01 3·06 2·93 3·18 3·05 2·50 Cs 0·11 0·05 0·08 0·12 0·12 0·09 0·09 0·10 0·20 0·15 Ba 29·7 11·5 16·7 22·7 30·8 24·5 25·9 24·4 29·7 20·3 La 3·63 3·46 3·31 2·90 3·29 3·17 3·06 3·32 3·28 2·74 Ce 10·16 9·62 9·23 8·07 9·11 8·92 8·50 9·11 9·00 7·68 Pr 1·66 1·53 1·52 1·33 1·48 1·48 1·40 1·50 1·47 1·24 Nd 8·70 7·90 7·94 6·98 7·73 7·72 7·41 7·92 7·60 6·66 Sm 3·10 2·85 2·80 2·45 2·80 2·69 2·57 2·70 2·76 2·34 Eu 1·06 1·04 1·04 0·95 1·03 1·06 1·04 1·06 1·02 0·88 Gd 3·95 3·71 3·77 3·28 3·70 3·71 3·54 3·65 3·60 3·15 Tb 0·74 0·70 0·70 0·63 0·70 0·68 0·66 0·70 0·67 0·59 Dy 4·78 4·65 4·57 4·06 4·59 4·54 4·33 4·56 4·49 3·89 Ho 1·00 0·98 0·97 0·87 0·97 0·96 0·93 0·98 0·94 0·81 Er 2·73 2·75 2·75 2·40 2·67 2·67 2·57 2·70 2·55 2·25 Tm 0·40 0·40 0·40 0·36 0·39 0·39 0·37 0·40 0·38 0·33 Yb 2·48 2·49 2·48 2·19 2·42 2·39 2·33 2·46 2·38 2·03 Lu 0·39 0·40 0·39 0·34 0·37 0·37 0·36 0·39 0·38 0·33 Hf 1·88 1·91 1·85 1·66 1·84 1·80 1·72 1·89 1·76 1·56 Ta 0·22 0·22 0·21 0·20 0·22 0·20 0·20 0·21 0·21 0·18 Pb 1·08 3·66 0·69 0·41 0·95 0·68 0·68 0·70 0·68 0·45 Th 0·29 0·33 0·23 0·21 0·24 0·23 0·22 0·23 0·23 0·19 U 0·080 0·097 0·068 0·060 0·069 0·068 0·060 0·069 0·066 0·056 . RP8-1 . RP8-2 . RP8-3 . RP8-4 . RP8-5 . RP8-6 . RP8-7 av. . RP8-7 2σ . RP8-8 . RP8-9 . RP8-10 . SiO2 (wt%) 51·03 50·18 51·29 48·92 50·33 50·22 50·50 1·59 49·86 50·43 49·53 TiO2 1·24 1·22 1·22 1·12 1·22 1·27 1·27 0·03 1·26 1·21 1·26 Al2O3 13·89 14·06 13·87 13·13 14·13 14·11 13·93 0·47 14·12 14·06 14·35 FeOt 13·14 13·23 12·92 15·89 13·06 13·37 13·36 0·28 13·36 13·22 13·73 MnO 0·177 0·203 0·193 0·244 0·195 0·233 0·204 0·007 0·224 0·193 0·225 MgO 7·52 7·19 7·58 7·71 7·62 7·41 7·41 0·39 7·33 7·75 7·87 CaO 10·48 11·77 10·81 9·95 10·55 10·10 10·36 0·38 10·78 10·17 10·26 Na2O 2·29 1·99 1·93 2·75 2·58 3·02 2·68 0·09 2·80 2·56 2·41 K2O 0·136 0·040 0·104 0·177 0·219 0·148 0·184 0·007 0·157 0·321 0·281 P2O5 0·094 0·114 0·090 0·106 0·090 0·116 0·085 0·007 0·111 0·087 0·092 Total (raw) 99·71 100·01 99·88 100·21 100·07 99·98 99·88 0·23 99·70 99·94 99·38 LOI 2·17 2·08 2·07 1·22 2·02 2·70 2·03 0·19 1·99 2·13 2·08 V (ppm) 362 358 342 364 357 380 369 17 368 362 427 Cr 179 132 158 174 171 151 163 23 150 183 215 Sc 43·6 44·1 43·5 43·0 43·7 45·4 45·6 45·3 44·6 45·2 Rb 2·02 0·56 1·39 2·89 3·50 1·99 2·41 2·26 4·84 3·46 Sr 109·5 132·8 116·0 103·1 118·3 95·7 99·0 100·6 74·7 82·4 Y 24·9 24·6 24·0 21·7 24·0 24·0 23·1 24·4 23·6 20·4 Zr 64·9 65·9 63·7 56·0 63·3 61·6 59·0 63·8 61·3 53·2 Nb 3·16 3·10 3·11 2·92 3·01 3·06 2·93 3·18 3·05 2·50 Cs 0·11 0·05 0·08 0·12 0·12 0·09 0·09 0·10 0·20 0·15 Ba 29·7 11·5 16·7 22·7 30·8 24·5 25·9 24·4 29·7 20·3 La 3·63 3·46 3·31 2·90 3·29 3·17 3·06 3·32 3·28 2·74 Ce 10·16 9·62 9·23 8·07 9·11 8·92 8·50 9·11 9·00 7·68 Pr 1·66 1·53 1·52 1·33 1·48 1·48 1·40 1·50 1·47 1·24 Nd 8·70 7·90 7·94 6·98 7·73 7·72 7·41 7·92 7·60 6·66 Sm 3·10 2·85 2·80 2·45 2·80 2·69 2·57 2·70 2·76 2·34 Eu 1·06 1·04 1·04 0·95 1·03 1·06 1·04 1·06 1·02 0·88 Gd 3·95 3·71 3·77 3·28 3·70 3·71 3·54 3·65 3·60 3·15 Tb 0·74 0·70 0·70 0·63 0·70 0·68 0·66 0·70 0·67 0·59 Dy 4·78 4·65 4·57 4·06 4·59 4·54 4·33 4·56 4·49 3·89 Ho 1·00 0·98 0·97 0·87 0·97 0·96 0·93 0·98 0·94 0·81 Er 2·73 2·75 2·75 2·40 2·67 2·67 2·57 2·70 2·55 2·25 Tm 0·40 0·40 0·40 0·36 0·39 0·39 0·37 0·40 0·38 0·33 Yb 2·48 2·49 2·48 2·19 2·42 2·39 2·33 2·46 2·38 2·03 Lu 0·39 0·40 0·39 0·34 0·37 0·37 0·36 0·39 0·38 0·33 Hf 1·88 1·91 1·85 1·66 1·84 1·80 1·72 1·89 1·76 1·56 Ta 0·22 0·22 0·21 0·20 0·22 0·20 0·20 0·21 0·21 0·18 Pb 1·08 3·66 0·69 0·41 0·95 0·68 0·68 0·70 0·68 0·45 Th 0·29 0·33 0·23 0·21 0·24 0·23 0·22 0·23 0·23 0·19 U 0·080 0·097 0·068 0·060 0·069 0·068 0·060 0·069 0·066 0·056 XRF major and minor element data for oxides, V, and Cr were renormalized to 100 % totals on an anhydrous basis, with FeOt calculated assuming all Fe is present as FeO. Loss on ignition (LOI) and solution ICP-MS trace element data were not renormalized. Raw totals are given as an indication of data quality. XRF data for sample RP8-7 were measured in triplicate and are reported as the average and 2 standard deviations of these three analyses. Open in new tab Rare earth element patterns for RP8 samples (Fig. 6a) show similar ‘hump-shaped’ patterns to Winnipegosis komatiite samples (Waterton et al., 2017), with modest depletions in both the HREE and LREE relative to the MREE. REE concentrations are approximately six times PUM, and are consistently ∼2·5 times higher than calculated REE contents for the Winnipegosis komatiite parental melt. Immobile trace element spidergrams (Fig. 6b) also closely parallel those of the Winnipegosis komatiites, and RP8 basalts are similarly strongly depleted in highly incompatible trace elements. As for the Winnipegosis komatiites, large ion lithophile elements LILE in RP8 basalts are highly scattered. Fig. 6. Open in new tabDownload slide RP8 basalt trace element data compared to Winnipegosis komatiite data from borehole RP1A. (a) Rare earth element patterns, normalized to primitive upper mantle (PUM; Hofmann, 1988), showing ‘hump shape’ in both basalts and komatiites with depleted LREE and HREE. Estimated Winnipegosis komatiite parental melt composition, with 95 % confidence limits (95 % confidence; Table 6) is shown for comparison. (b) Immobile trace element spidergram, excluding the large ion lithophile elements Cs, Rb, Ba, Pb, and K. Sr shows evidence of mobility in some samples. Fig. 6. Open in new tabDownload slide RP8 basalt trace element data compared to Winnipegosis komatiite data from borehole RP1A. (a) Rare earth element patterns, normalized to primitive upper mantle (PUM; Hofmann, 1988), showing ‘hump shape’ in both basalts and komatiites with depleted LREE and HREE. Estimated Winnipegosis komatiite parental melt composition, with 95 % confidence limits (95 % confidence; Table 6) is shown for comparison. (b) Immobile trace element spidergram, excluding the large ion lithophile elements Cs, Rb, Ba, Pb, and K. Sr shows evidence of mobility in some samples. Mineral chemistry RP7 dunite chromite Chromite Fe3+ contents were determined using the method of Droop (1987). Chromite grains from dunite sample RP7-14 have lower Cr# [Cr# = Cr/(Cr + Al)] of 0·55–0·61 (Table 5), compared with Winnipegosis komatiite chromite (Cr# 0·64–0·72; Waterton et al., 2017). This is comparable with previous observations that chromite from thin komatiite flows generally has higher Cr# than chromite from related dunite sheets and channels (Barnes, 1998). Mg# of the RP7 dunite chromites ranges from 0·37 to 0·49, with a more restricted range and lower maximum Mg# than chromite from the Winnipegosis komatiites. This lower maximum Mg# probably reflects post-cumulus modification (Barnes, 1998) of the dunite chromites, with gain of Fe2+ from surrounding olivine at low temperatures (Ballhaus et al., 1991; Waterton et al., 2017). The RP7 chromites also show distinct enrichments in TiO2, indicative of extensive reaction with intercumulus melts trapped during dunite formation (Barnes, 1998). Table 5: Representative and average compositions of chromite (chr) and clinopyroxene (cpx) from Winnipegosis samples, analysed by EPMA Sample: . RP7-14 . RP7-14 . RP7-14 . RP1A-av. . RP12-307·3 . RP12-306 . RP1A-111 . RP1A-8 . RP1A-18 . RP1A-99 . Type: Dunite Dunite Dunite Mass. kom. Aci. px Ol spin. Type 1 Type 1/2 Type 2 Type 2 Grain: chr3 chr6 average-chr average-chr cpx7 cpx10 cpx9 cpx8 cpx15 cpx3 SiO2 0·10 0·06 0·08 0·22 52·75 48·23 48·85 47·57 45·55 46·00 TiO2 1·01 0·82 0·83 0·38 0·29 0·65 0·69 0·78 1·15 0·91 Al2O3 19·81 20·00 20·37 14·53 3·17 7·51 6·69 9·11 9·91 9·65 Cr2O3 39·22 41·34 40·83 49·15 0·47 0·25 0·29 0·20 0·12 0·12 Fe2O3 80·53 70·23 7·37 5·89 FeO 22·34 20·09 20·84 18·11 7·40 10·74 7·73 9·17 10·36 13·32 MnO 0·14 0·21 0·21 0·17 0·19 0·18 0·29 CoO 0·07 0·05 0·05 0·03 0·00 0·01 0·00 0·01 0·01 0·00 NiO 0·20 0·18 0·17 0·18 0·04 0·02 0·02 0·01 0·02 0·00 MgO 90·01 10·37 10·00 10·64 20·75 14·82 13·77 13·01 10·03 10·14 CaO 0·01 0·01 0·01 0·03 15·43 17·23 21·43 19·86 21·67 18·89 Na2O 0·00 0·00 0·00 0·00 0·13 0·20 0·24 0·25 0·27 0·35 K2O 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·03 P2O5 0·00 0·00 0·00 0·00 0·01 0·05 0·02 0·04 0·03 0·06 Total 100·28 100·14 100·54 99·30 100·66 99·93 99·89 100·20 99·29 99·76 Cr# 0·570 0·581 0·574 0·694 Mg# 0·418 0·479 0·461 0·506 0·833 0·711 0·761 0·717 0·633 0·576 Wo (Ca2Si2O6) 0·308 0·373 0·460 0·440 0·496 0·435 En (Mg2Si2O6) 0·576 0·446 0·411 0·401 0·319 0·325 Fs (Fe2Si2O6) 0·115 0·181 0·129 0·159 0·185 0·240 % Alz 5·1 10·7 9·6 12·0 13·9 12·9 Sample: . RP7-14 . RP7-14 . RP7-14 . RP1A-av. . RP12-307·3 . RP12-306 . RP1A-111 . RP1A-8 . RP1A-18 . RP1A-99 . Type: Dunite Dunite Dunite Mass. kom. Aci. px Ol spin. Type 1 Type 1/2 Type 2 Type 2 Grain: chr3 chr6 average-chr average-chr cpx7 cpx10 cpx9 cpx8 cpx15 cpx3 SiO2 0·10 0·06 0·08 0·22 52·75 48·23 48·85 47·57 45·55 46·00 TiO2 1·01 0·82 0·83 0·38 0·29 0·65 0·69 0·78 1·15 0·91 Al2O3 19·81 20·00 20·37 14·53 3·17 7·51 6·69 9·11 9·91 9·65 Cr2O3 39·22 41·34 40·83 49·15 0·47 0·25 0·29 0·20 0·12 0·12 Fe2O3 80·53 70·23 7·37 5·89 FeO 22·34 20·09 20·84 18·11 7·40 10·74 7·73 9·17 10·36 13·32 MnO 0·14 0·21 0·21 0·17 0·19 0·18 0·29 CoO 0·07 0·05 0·05 0·03 0·00 0·01 0·00 0·01 0·01 0·00 NiO 0·20 0·18 0·17 0·18 0·04 0·02 0·02 0·01 0·02 0·00 MgO 90·01 10·37 10·00 10·64 20·75 14·82 13·77 13·01 10·03 10·14 CaO 0·01 0·01 0·01 0·03 15·43 17·23 21·43 19·86 21·67 18·89 Na2O 0·00 0·00 0·00 0·00 0·13 0·20 0·24 0·25 0·27 0·35 K2O 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·03 P2O5 0·00 0·00 0·00 0·00 0·01 0·05 0·02 0·04 0·03 0·06 Total 100·28 100·14 100·54 99·30 100·66 99·93 99·89 100·20 99·29 99·76 Cr# 0·570 0·581 0·574 0·694 Mg# 0·418 0·479 0·461 0·506 0·833 0·711 0·761 0·717 0·633 0·576 Wo (Ca2Si2O6) 0·308 0·373 0·460 0·440 0·496 0·435 En (Mg2Si2O6) 0·576 0·446 0·411 0·401 0·319 0·325 Fs (Fe2Si2O6) 0·115 0·181 0·129 0·159 0·185 0·240 % Alz 5·1 10·7 9·6 12·0 13·9 12·9 Type indicates petrographical type; Mass. kom., massive komatiite; Aci. px, acicular pyroxene sample; Ol spin., random olivine spinifex. Wo, En, and Fs are the relative proportions of wollastonite (Ca2Si2O6), enstatite (Mg2Si2O6), and ferrosilite (Fe2Si2O6), respectively. % Alz is the stoichiometrically calculated percentage of the tetrahedral site occupied by Al. Open in new tab Table 5: Representative and average compositions of chromite (chr) and clinopyroxene (cpx) from Winnipegosis samples, analysed by EPMA Sample: . RP7-14 . RP7-14 . RP7-14 . RP1A-av. . RP12-307·3 . RP12-306 . RP1A-111 . RP1A-8 . RP1A-18 . RP1A-99 . Type: Dunite Dunite Dunite Mass. kom. Aci. px Ol spin. Type 1 Type 1/2 Type 2 Type 2 Grain: chr3 chr6 average-chr average-chr cpx7 cpx10 cpx9 cpx8 cpx15 cpx3 SiO2 0·10 0·06 0·08 0·22 52·75 48·23 48·85 47·57 45·55 46·00 TiO2 1·01 0·82 0·83 0·38 0·29 0·65 0·69 0·78 1·15 0·91 Al2O3 19·81 20·00 20·37 14·53 3·17 7·51 6·69 9·11 9·91 9·65 Cr2O3 39·22 41·34 40·83 49·15 0·47 0·25 0·29 0·20 0·12 0·12 Fe2O3 80·53 70·23 7·37 5·89 FeO 22·34 20·09 20·84 18·11 7·40 10·74 7·73 9·17 10·36 13·32 MnO 0·14 0·21 0·21 0·17 0·19 0·18 0·29 CoO 0·07 0·05 0·05 0·03 0·00 0·01 0·00 0·01 0·01 0·00 NiO 0·20 0·18 0·17 0·18 0·04 0·02 0·02 0·01 0·02 0·00 MgO 90·01 10·37 10·00 10·64 20·75 14·82 13·77 13·01 10·03 10·14 CaO 0·01 0·01 0·01 0·03 15·43 17·23 21·43 19·86 21·67 18·89 Na2O 0·00 0·00 0·00 0·00 0·13 0·20 0·24 0·25 0·27 0·35 K2O 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·03 P2O5 0·00 0·00 0·00 0·00 0·01 0·05 0·02 0·04 0·03 0·06 Total 100·28 100·14 100·54 99·30 100·66 99·93 99·89 100·20 99·29 99·76 Cr# 0·570 0·581 0·574 0·694 Mg# 0·418 0·479 0·461 0·506 0·833 0·711 0·761 0·717 0·633 0·576 Wo (Ca2Si2O6) 0·308 0·373 0·460 0·440 0·496 0·435 En (Mg2Si2O6) 0·576 0·446 0·411 0·401 0·319 0·325 Fs (Fe2Si2O6) 0·115 0·181 0·129 0·159 0·185 0·240 % Alz 5·1 10·7 9·6 12·0 13·9 12·9 Sample: . RP7-14 . RP7-14 . RP7-14 . RP1A-av. . RP12-307·3 . RP12-306 . RP1A-111 . RP1A-8 . RP1A-18 . RP1A-99 . Type: Dunite Dunite Dunite Mass. kom. Aci. px Ol spin. Type 1 Type 1/2 Type 2 Type 2 Grain: chr3 chr6 average-chr average-chr cpx7 cpx10 cpx9 cpx8 cpx15 cpx3 SiO2 0·10 0·06 0·08 0·22 52·75 48·23 48·85 47·57 45·55 46·00 TiO2 1·01 0·82 0·83 0·38 0·29 0·65 0·69 0·78 1·15 0·91 Al2O3 19·81 20·00 20·37 14·53 3·17 7·51 6·69 9·11 9·91 9·65 Cr2O3 39·22 41·34 40·83 49·15 0·47 0·25 0·29 0·20 0·12 0·12 Fe2O3 80·53 70·23 7·37 5·89 FeO 22·34 20·09 20·84 18·11 7·40 10·74 7·73 9·17 10·36 13·32 MnO 0·14 0·21 0·21 0·17 0·19 0·18 0·29 CoO 0·07 0·05 0·05 0·03 0·00 0·01 0·00 0·01 0·01 0·00 NiO 0·20 0·18 0·17 0·18 0·04 0·02 0·02 0·01 0·02 0·00 MgO 90·01 10·37 10·00 10·64 20·75 14·82 13·77 13·01 10·03 10·14 CaO 0·01 0·01 0·01 0·03 15·43 17·23 21·43 19·86 21·67 18·89 Na2O 0·00 0·00 0·00 0·00 0·13 0·20 0·24 0·25 0·27 0·35 K2O 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·00 0·03 P2O5 0·00 0·00 0·00 0·00 0·01 0·05 0·02 0·04 0·03 0·06 Total 100·28 100·14 100·54 99·30 100·66 99·93 99·89 100·20 99·29 99·76 Cr# 0·570 0·581 0·574 0·694 Mg# 0·418 0·479 0·461 0·506 0·833 0·711 0·761 0·717 0·633 0·576 Wo (Ca2Si2O6) 0·308 0·373 0·460 0·440 0·496 0·435 En (Mg2Si2O6) 0·576 0·446 0·411 0·401 0·319 0·325 Fs (Fe2Si2O6) 0·115 0·181 0·129 0·159 0·185 0·240 % Alz 5·1 10·7 9·6 12·0 13·9 12·9 Type indicates petrographical type; Mass. kom., massive komatiite; Aci. px, acicular pyroxene sample; Ol spin., random olivine spinifex. Wo, En, and Fs are the relative proportions of wollastonite (Ca2Si2O6), enstatite (Mg2Si2O6), and ferrosilite (Fe2Si2O6), respectively. % Alz is the stoichiometrically calculated percentage of the tetrahedral site occupied by Al. Open in new tab Winnipegosis komatiite clinopyroxenes Winnipegosis komatiite clinopyroxenes show compositional variations based on their size and habit. Large phenocrysts from the acicular pyroxene sample, RP12-307.3, are augitic in composition, and Mg-rich and Ca-poor compared with dendritic clinopyroxenes (Table 5, Fig. 7). The olivine spinifex sample RP12-306.1 contains the coarsest dendrites. These range in composition from augite to diopside, along a trend of increasing Ca and decreasing Mg at approximately constant Fe. Dendritic pyroxenes from the massive komatiite flows are diopside or augite in composition with very high Ca. There is a slight trend towards decreasing Mg/Fe with decreasing clinopyroxene dendrite size, from the coarsest grained type 1 sample (RP1A-111), through the transitional type 1/2 sample (RP1A-8) towards the finest grained type 2 samples (RP1A-18, RP1A-99). Al contents, and the percentage of the tetrahedral site occupied by Al (% Alz), generally increase with decreasing grain size. This probably results from metastable pyroxene growth at high cooling rates (Faure & Tissandier, 2014), with the highest Al contents in fine-grained type 2 dendrites resulting from the highest degree of undercooling, and lower Al contents in the RP12-307.3 phenocrysts resulting from growth closer to equilibrium conditions. The full EPMA dataset for chromite and clinopyroxene is available in the Supplementary Data. Fig. 7. Open in new tabDownload slide Compositions of clinopyroxenes from massive Winnipegosis komatiite samples from borehole RP1A, and differentiated komatiite samples from borehole RP12. Data are plotted on a section of the pyroxene quadrilateral (Morimoto et al., 1988), with vertices at diopside (Di; CaMgSi2O6), enstatite (En; Mg2Si2O6), and at 25 % ferrosilite (Fs; Fe2Si2O6) along the enstatite–ferrosilite and diopside–hedenbergite boundaries. Data are subdivided by sample (symbol colour) and petrographical type (symbol shape), and compared with compositions predicted by MELTS (‘MELTS’; modelled with an fO2 of QFM and pressure of 1 bar) and experimental pyroxenes of Parman et al. (1997), crystallised from a similar bulk-rock composition under anhydrous conditions. Crosses with labels indicate temperatures of pyroxene saturation (1198 °C) and lowest temperature pyroxene before MELTS predicts saturation of low-Ca pyroxene (1156 °C). Fig. 7. Open in new tabDownload slide Compositions of clinopyroxenes from massive Winnipegosis komatiite samples from borehole RP1A, and differentiated komatiite samples from borehole RP12. Data are plotted on a section of the pyroxene quadrilateral (Morimoto et al., 1988), with vertices at diopside (Di; CaMgSi2O6), enstatite (En; Mg2Si2O6), and at 25 % ferrosilite (Fs; Fe2Si2O6) along the enstatite–ferrosilite and diopside–hedenbergite boundaries. Data are subdivided by sample (symbol colour) and petrographical type (symbol shape), and compared with compositions predicted by MELTS (‘MELTS’; modelled with an fO2 of QFM and pressure of 1 bar) and experimental pyroxenes of Parman et al. (1997), crystallised from a similar bulk-rock composition under anhydrous conditions. Crosses with labels indicate temperatures of pyroxene saturation (1198 °C) and lowest temperature pyroxene before MELTS predicts saturation of low-Ca pyroxene (1156 °C). DISCUSSION—RELATIONSHIP BETWEEN DUNITES AND KOMATIITES The highly magnesian nature of the RP7 dunites, and their compositional similarity to olivine compositions measured in Winnipegosis komatiite samples (Fig. 5), suggests that the dunites may represent cumulates of komatiitic olivine and chromite. However, care must be taken with this interpretation as the RP7 dunites have been metamorphosed and do not represent pristine cumulates. Thus, compositional variations in RP7 dunite samples are further investigated with a view to understanding how they relate to the komatiites, and how their compositions might have been modified during metamorphism. Accumulating mineral assemblage The proportion of olivine and chromite in the accumulating mineral assemblage can be investigated on a plot of Cr2O3 against MgO (Fig. 5), as addition of chromite produces near-vertical trends, whereas variations in the olivine Mg# or addition of intercumulus melt produce near-horizontal trends (modelling described below). As no fresh olivine is preserved in the RP7 dunites, and the proportion of chromite calculated is relatively insensitive to olivine composition, average Winnipegosis komatiite olivine is used as a proxy for the igneous dunite olivine. All of the RP7 dunite compositions lie in the range of Cr2O3 values produced by a mixture of 1·5–2·3 wt% chromite and 98·5–97·7 wt% olivine. Use of the average chromite composition from Winnipegosis komatiite samples (Waterton et al., 2017) produces a similar trend, but only 1·2–1·8 wt% chromite is required to reproduce the RP7 Cr2O3 values, due to the higher Cr2O3 of the komatiite chromites. However, all dunite compositions lie to lower MgO values than a pure olivine–chromite mixture. This can be explained by the presence of trapped interstitial melt, accumulation of lower Mg# olivine than average Winnipegosis komatiite olivine, metamorphic loss of MgO, or some combination of these processes. Variations in the Mg# of accumulating olivine are predicted to form a negative trend in a plot of FeO against MgO (Fig. 5). Taken in isolation, this appears to explain much of the compositional variability in the RP7 samples, with dunites appearing to fall in the Mg# range 0·87–0·91. However, some RP7 samples are displaced away from olivine compositions towards lower MgO and increasing Al2O3, which cannot be explained by variations in olivine composition. These displacements also cannot be explained by differences in chromite chemistry, or in the relative proportions of chromite and olivine in the dunites due to the small modal proportion of chromite in the dunites. The RP7 dunite bulk-rock compositions therefore require the presence of some interstitial melt and/or metamorphic mobility of some elements. Effects of trapped interstitial melt and metamorphism The effect of trapped intercumulus melt on dunite bulk-rock compositions is modelled by mixing between an accumulating assemblage with 1·9 wt% average RP7 dunite chromite and 98·1 wt% olivine (the middle of the range of chromite:olivine proportions identified from the Cr2O3–MgO plot) and both a primitive and evolved intercumulus melt composition. The Winnipegosis komatiite parental melt composition (Table 6) is used as a proxy for primitive intercumulus melt, and the average Winnipegosis basalt composition from borehole RP8 (Table 4) is used to represent an evolved intercumulus melt. However, these different intercumulus melt compositions produce similar compositional vectors for both major and immobile trace elements; despite the evolved melt containing much higher incompatible element concentrations, it also lies at a much lower MgO content (Fig. 5). The main difference is that, for a given quantity of trapped intercumulus melt, an evolved intercumulus melt will have a larger effect on bulk-rock dunite compositions than a primitive one. Olivine compositions are varied between the highest Mg# olivine (Mg# = 0·92), the average Mg# olivine (Mg# = 0·91), and the lowest Mg# olivine (Mg# = 0·87), measured from Winnipegosis komatiites (Waterton et al., 2017). Table 6: Bulk-rock major, minor, and rare earth element composition calculated for the parental melt to RP1A komatiites Oxide . Parental melt . 95 % conf. . SiO2 (wt%) 46·01 0·41 TiO2 0·50 0·03 Al2O3 8·50 0·57 Cr2O3 0·32 0·03 FeOt 11·23 0·03 MnO 0·18 0·00 MgO 23·46 1·64 NiO 0·14 0·01 CoO 0·01 0·00 CaO 8·27 0·52 Na2O 1·18 0·10 K2O 0·07 0·02 P2O5 0·04 0·00 H2O 0·10 n.a. La (ppm) 1·32 0·12 Ce 3·70 0·32 Pr 0·58 0·05 Nd 3·11 0·25 Sm 1·08 0·08 Eu 0·42 0·03 Gd 1·52 0·12 Tb 0·27 0·02 Dy 1·79 0·14 Ho 0·38 0·03 Er 1·11 0·09 Yb 1·08 0·08 Lu 0·16 0·01 Oxide . Parental melt . 95 % conf. . SiO2 (wt%) 46·01 0·41 TiO2 0·50 0·03 Al2O3 8·50 0·57 Cr2O3 0·32 0·03 FeOt 11·23 0·03 MnO 0·18 0·00 MgO 23·46 1·64 NiO 0·14 0·01 CoO 0·01 0·00 CaO 8·27 0·52 Na2O 1·18 0·10 K2O 0·07 0·02 P2O5 0·04 0·00 H2O 0·10 n.a. La (ppm) 1·32 0·12 Ce 3·70 0·32 Pr 0·58 0·05 Nd 3·11 0·25 Sm 1·08 0·08 Eu 0·42 0·03 Gd 1·52 0·12 Tb 0·27 0·02 Dy 1·79 0·14 Ho 0·38 0·03 Er 1·11 0·09 Yb 1·08 0·08 Lu 0·16 0·01 Major element composition is renormalized with all Fe as FeO, including 0·1 wt% H2O. REE concentrations are not renormalized. Open in new tab Table 6: Bulk-rock major, minor, and rare earth element composition calculated for the parental melt to RP1A komatiites Oxide . Parental melt . 95 % conf. . SiO2 (wt%) 46·01 0·41 TiO2 0·50 0·03 Al2O3 8·50 0·57 Cr2O3 0·32 0·03 FeOt 11·23 0·03 MnO 0·18 0·00 MgO 23·46 1·64 NiO 0·14 0·01 CoO 0·01 0·00 CaO 8·27 0·52 Na2O 1·18 0·10 K2O 0·07 0·02 P2O5 0·04 0·00 H2O 0·10 n.a. La (ppm) 1·32 0·12 Ce 3·70 0·32 Pr 0·58 0·05 Nd 3·11 0·25 Sm 1·08 0·08 Eu 0·42 0·03 Gd 1·52 0·12 Tb 0·27 0·02 Dy 1·79 0·14 Ho 0·38 0·03 Er 1·11 0·09 Yb 1·08 0·08 Lu 0·16 0·01 Oxide . Parental melt . 95 % conf. . SiO2 (wt%) 46·01 0·41 TiO2 0·50 0·03 Al2O3 8·50 0·57 Cr2O3 0·32 0·03 FeOt 11·23 0·03 MnO 0·18 0·00 MgO 23·46 1·64 NiO 0·14 0·01 CoO 0·01 0·00 CaO 8·27 0·52 Na2O 1·18 0·10 K2O 0·07 0·02 P2O5 0·04 0·00 H2O 0·10 n.a. La (ppm) 1·32 0·12 Ce 3·70 0·32 Pr 0·58 0·05 Nd 3·11 0·25 Sm 1·08 0·08 Eu 0·42 0·03 Gd 1·52 0·12 Tb 0·27 0·02 Dy 1·79 0·14 Ho 0·38 0·03 Er 1·11 0·09 Yb 1·08 0·08 Lu 0·16 0·01 Major element composition is renormalized with all Fe as FeO, including 0·1 wt% H2O. REE concentrations are not renormalized. Open in new tab This modelling shows that all of the RP7 dunite incompatible element–MgO relations can be reproduced by addition of small amounts of intercumulus melt (∼3–9 wt% of evolved melt, or ∼5–20 % of primitive melt) to a cumulus assemblage dominated by Mg# 0·91–0·92 olivine. This implies that the high-FeO RP7 samples, which fall close to olivine compositions with Mg# ∼0·87, must have gained FeO during subsequent metamorphism. This interpretation is consistent with the presence of abundant metamorphic magnetite in the RP7 samples (Fig. 4c). We therefore interpret the RP7 dunites as cumulates of high-Mg# (∼0·91–0·92) olivine and chromite, which trapped small but variable amounts of intercumulus melt, and which gained up to a maximum of 3 wt% FeO during metamorphism. Average RP7 dunite compositions of 1·0 wt% Al2O3 imply average trapped melt contents of 7 % for a primitive intercumulus melt, and 4 % for an evolved intercumulus melt. The presence of this trapped melt is supported by elevated TiO2 contents in chromite (Barnes, 1998) from the RP7 dunites. RP7 dunites are metamorphosed komatiitic cumulates Assuming a similar bulk-rock FeO–MgO relationship and Fe2+/∑Fe to that for the Winnipegosis komatiites (FeO = 0·9FeOt; Waterton et al., 2017), and olivine–liquid Fe–Mg exchange coefficient [exchange coefficient, KD,Fe2+–Mgol–liq = (FeO/MgO)ol/(FeO/MgO)liq, hereafter KD ] = 0·345 (Matzen et al., 2011), dunites composed of olivine with Mg# 0·91–0·92 are in equilibrium with liquids containing 20–23 wt% MgO. Even if the highest FeO contents of RP7 dunites are in fact a primary feature, dunite compositions similar to olivine with an Mg# of 0·875 would be in equilibrium with a liquid with ∼15 wt% MgO. If the liquid composition is assumed to have a higher FeO content, more similar to the RP8 basalts, even higher MgO contents are calculated. It is therefore highly likely that komatiitic melts (with >18 wt% MgO) were required to form at least some of the RP7 dunites. Dunite bulk-rock compositions reflecting uniform and high average olivine Mg# values are not consistent through closed-system crystallization of a single batch of komatiitic magma, as this would be expected to produce cumulates with varying compositions due to melt evolution and crystallization of lower Mg# olivine. The magma chamber represented by the RP7 dunites therefore experienced open-system behaviour, with repeated episodes of magma recharge maintaining a high-MgO melt from which the olivines crystallized (e.g. Prægel & Holm, 2001). The average accumulating olivine composition of the dunites inferred from their major and trace element relationships (Mg# = 0·91–0·92) overlap, but extend to higher Mg# values than the average olivine composition measured in the Winnipegosis komatiites (Mg# = 0·907 ± 0·005; n = 183), and are close to the upper end of Winnipegosis komatiite olivine Mg# values (0·923 ± 0·005; Waterton et al., 2017). A possible interpretation is that the dunites formed from crystallization of an even more magnesian melt. However, the distribution of olivine Mg# in the Winnipegosis komatiites provides an alternative explanation. For olivine generated by ideal fractional crystallization of a single parental melt, a histogram of randomly sampled olivine Mg# should be almost flat (Thomson & Maclennan, 2013), with a slight fall in frequency towards more Mg#-poor compositions due to the higher density of Fe-rich olivine. Instead, Winnipegosis komatiites show a Gaussian-like distribution, with a mode in olivine Mg# around Mg# = 0·907 (Waterton et al., 2017), indicative of re-equilibration of the olivine grains with their evolving carrier melt (Thomson & Maclennan, 2013). This interpretation is supported by the presence of large embayed olivine grains in Winnipegosis komatiites (Waterton et al., 2017; Waterton, 2018), which indicate that the melt had begun to react with and partly dissolve early formed olivine. The calculated Mg# values of accumulating olivine in the dunites are not affected by these processes, as this olivine Mg# is derived from bulk-rock data. It is therefore possible, and we consider it likely, that the dunites and komatiites formed from similar high-MgO parental melts. RELATIONSHIP BETWEEN BASALTS AND KOMATIITES The similarity of REE patterns between the RP8 basalts and Winnipegosis komatiites (Fig. 6) suggests a possible genetic link between the two suites. This could reflect formation of the basalts through extensive crystallization of a komatiitic parental melt, or both volcanic suites being derived from a similar mantle source that experienced different degrees of melting (Arndt & Nesbitt, 1982). These possibilities are examined by first testing the hypothesis that fractional crystallization of a parental melt similar to the Winnipegosis komatiites could generate melts that match both the major element composition and REE systematics of the RP8 basalts. Modelling komatiite crystallization Modelling of limited degrees of crystallization within komatiitic rocks is relatively simple due to the dominant control of olivine (e.g. Waterton et al., 2017). By contrast, modelling the extensive, >50 % crystallization required to form a basalt from a komatiitic parent melt is complicated by the likelihood of saturation in other major phases, such as plagioclase and pyroxene, and requires prediction of the composition of these phases, at what point they become saturated in the melt, and their relative abundances. To undertake this modelling, Rhyolite-MELTS software v1.2 (Gualda et al., 2012; Ghiorso & Gualda, 2015) is used, which supersedes previous MELTS software for modelling crystallization of magmas in the pressure range 0–2 GPa. Rhyolite-MELTS (referred to simply as ‘MELTS’ below) and other variants of MELTS software have shown success in reproducing many features of crystallization of a number of magma compositions, including MORB (Ghiorso, 1997; Asimow & Longhi, 2004; Asimow et al., 2004), Martian basalts (Balta & McSween, 2013), and rhyolites (Gualda et al., 2012). However, the reliability of MELTS software in predicting komatiitic crystallization trends has not yet been established, and an attempt is first made to evaluate whether MELTS can successfully reproduce characteristics of Winnipegosis komatiite crystallization established by Waterton et al. (2017). The well-constrained fractionation trend of the Winnipegosis komatiites provides a robust test of the accuracy of MELTS predictions before modelling a more extensive fractionation trend from komatiite to basalt. Attempts were also made to model crystallization of Winnipegosis komatiites at 1 atm pressure with Petrolog (Danyushevsky & Plechov, 2011). However, Petrolog modelling consistently predicted the crystallization of significant quantities of orthopyroxene, as the third mineral to crystallize after olivine and spinel. The amount of orthopyroxene after 80 % crystallization varied from 3·3 to 24 wt% of the cumulate, primarily dependent on the olivine model used (Ford et al., 1983; Danyushevsky, 2001; Herzberg & O’Hara, 2002; Putirka, 2005), with a smaller dependence on the orthopyroxene model used (Beattie, 1993; Bolikhovskaya et al., 1995). This prediction is inconsistent with the absence of orthopyroxene in Winnipegosis komatiites, and surprising given their low SiO2 contents (46 wt% SiO2 parental melt; Table 6) and the general expectation that clinopyroxene should crystallize before orthopyroxene for low-pressure crystallization (<5 kbar) of mantle-derived primary magmas (e.g. Longhi, 1981; Campbell, 1985). To check that this expectation holds for Winnipegosis komatiites, their compositions were plotted on a ol–cpx–qtz ternary (Fig. 8) along with 1 atm liquidus relations from Kinzler & Grove (1985). All Winnipegosis komatiites plot close to the olivine–clinopyroxene join, in the field for which augite crystallization precedes pigeonite or orthopyroxene crystallization. The Petrolog prediction of extensive, early orthopyroxene crystallization therefore cannot be reconciled with observations from Winnipegosis komatiites (see below) or predictions from experimental petrology. Thus, Petrolog was not used for further modelling. Fig. 8. Open in new tabDownload slide Winnipegosis komatiite compositions from boreholes RP1A and RP12 plotted in an Olivine (ol)–Clinopyroxene (cpx)–Quartz (Qtz) ternary following the projection scheme of Grove et al. (1982) and Grove & Baker (1983). One atmosphere liquidus phase relations (continuous black lines) are shown from Kinzler & Grove (1985). Dashed tie-lines are from olivine separate compositions that will crystallize olivine followed by augite (aug), olivine followed by pigeonite (pig), and olivine followed by orthopyroxene (opx). All Winnipegosis komatiites have bulk compositions in the field for which augite crystallizes after olivine (ol → aug). Fig. 8. Open in new tabDownload slide Winnipegosis komatiite compositions from boreholes RP1A and RP12 plotted in an Olivine (ol)–Clinopyroxene (cpx)–Quartz (Qtz) ternary following the projection scheme of Grove et al. (1982) and Grove & Baker (1983). One atmosphere liquidus phase relations (continuous black lines) are shown from Kinzler & Grove (1985). Dashed tie-lines are from olivine separate compositions that will crystallize olivine followed by augite (aug), olivine followed by pigeonite (pig), and olivine followed by orthopyroxene (opx). All Winnipegosis komatiites have bulk compositions in the field for which augite crystallizes after olivine (ol → aug). Parental melt composition Bulk-rock compositions for most elements in the Winnipegosis komatiites are arrayed along olivine control lines (Waterton et al., 2017). The komatiite parental melt composition for this modelling can therefore be estimated by interpolating the regressions of various elements against MgO for the best preserved Winnipegosis komatiites from borehole RP1A to the parental melt MgO content of 23·6 ± 1·6 wt%, calculated from olivine–melt Fe–Mg exchange and a maximum olivine Mg# of 0·923 (Waterton et al., 2017). Uncertainties are calculated as the maximum and minimum values of the 95 % confidence intervals about the regression lines in the interval 23·6 ± 1·6 wt% MgO. For elements with tight regression lines against MgO, these uncertainties are dominated by the ±1·6 wt% uncertainty in MgO, leading to similar relative uncertainties and errors in parental melt composition that are correlated with MgO along olivine control lines. Parental melt compositions are summarized in Table 6, REE concentrations are plotted in Fig. 6, and major element compositions are indicated in Fig. 9 and in figures later in the paper. Fig. 9. Open in new tabDownload slide (a) Olivine–liquid Fe–Mg exchange coefficient, KD, as a function of liquid MgO. Values predicted by MELTS with initial fO2 ranging between QFM – 1 and QFM + 1, under fixed fO2 (fixed) and freely varying fO2 (var.), are compared with previously published KD values derived from experiments (Herzberg & O’Hara, 2002; Matzen et al., 2011). MELTS underestimates KD, particularly at high MgO. (b) Comparison of the FeO–MgO LLD predicted by MELTS, with liquid lines of descent calculated using incremental olivine subtraction from the same komatiite parental melt composition (parental), and experimentally derived KD values. Tick marks and percentages indicate wt% of liquid that has crystallized. MELTS liquid lines of descent end at the temperature step immediately before cpx saturation (∼1200 °C, ∼35–37 wt% of liquid crystallized); liquid lines of descent calculated by olivine subtraction end at ∼36 wt% liquid crystallized. Liquid lines of descent for MELTS models with varying fO2 (not shown) fall within the range defined by the fixed fO2 models. The underestimation of KD by MELTS leads to increasing errors in liquid FeO content as crystallization progresses. ‘QFM no spn’ indicates MELTS model with spinel crystallization suppressed. Fig. 9. Open in new tabDownload slide (a) Olivine–liquid Fe–Mg exchange coefficient, KD, as a function of liquid MgO. Values predicted by MELTS with initial fO2 ranging between QFM – 1 and QFM + 1, under fixed fO2 (fixed) and freely varying fO2 (var.), are compared with previously published KD values derived from experiments (Herzberg & O’Hara, 2002; Matzen et al., 2011). MELTS underestimates KD, particularly at high MgO. (b) Comparison of the FeO–MgO LLD predicted by MELTS, with liquid lines of descent calculated using incremental olivine subtraction from the same komatiite parental melt composition (parental), and experimentally derived KD values. Tick marks and percentages indicate wt% of liquid that has crystallized. MELTS liquid lines of descent end at the temperature step immediately before cpx saturation (∼1200 °C, ∼35–37 wt% of liquid crystallized); liquid lines of descent calculated by olivine subtraction end at ∼36 wt% liquid crystallized. Liquid lines of descent for MELTS models with varying fO2 (not shown) fall within the range defined by the fixed fO2 models. The underestimation of KD by MELTS leads to increasing errors in liquid FeO content as crystallization progresses. ‘QFM no spn’ indicates MELTS model with spinel crystallization suppressed. It is important to note that the komatiite bulk-rock data do not represent liquid compositions. Instead, the olivine control lines reflect mixing between olivine and chromite phenocrysts and a melt residual to their crystallization, possibly shortly before or during the komatiite eruption (Waterton et al., 2017). This method provides a good estimate of the parental melt composition for elements that are incompatible during olivine + chromite crystallization, such as TiO2 and Al2O3, as the phenocryst–melt mixing lines should closely approximate the liquid line of descent (LLD; Francis, 1985; Larsen & Pedersen, 2000; Waterton et al., 2017). However, as the Winnipegosis komatiites underwent small amounts of crustal contamination during their ascent through the crust (Waterton et al., 2017), parental melt compositions are necessarily post-contamination (see Supplementary Data for details). In contrast, elements compatible or mildly incompatible in olivine, such as FeO, SiO2, MnO, Ni, and Co are predicted to have curved LLD due to variations in the composition of the melt and variations in olivine–melt partitioning during crystallization. As a result, the straight olivine control mixing trends in bulk-rock data for these elements against MgO do not closely approximate the komatiite LLD (Nebel et al., 2014; Waterton et al., 2017). However, interpolations of these mixing lines to the parental melt MgO can produce a reasonable estimate of the parental melt concentrations for these elements, provided there was no net loss of olivine and no net change in the composition of the olivine (e.g. loss of Mg-rich olivine and gain of Fe-rich olivine; Larsen & Pedersen 2000) before the melt–phenocryst mixing process occurred. The calculated parental melt content of 23·6 ± 1·6 wt% MgO lies within error of average bulk-rock MgO content (22·2 ± 2·2 wt%; Waterton et al., 2017), suggesting no significant net loss of olivine phenocrysts. Models of olivine crystallization terminated at the petrographically observed average phenocryst content in Winnipegosis komatiite lavas produce an average olivine composition identical to the average crystallizing composition established from major element relationships (Supplementary Data). This indicates that there was no net change in the composition of olivine before the melt–phenocryst mixing process occurred, and that this method of calculating parental melt compositions can be applied to elements compatible in olivine. The exception to this is Cr2O3, as chromite exerts a significant control on its whole-rock budget in the komatiites. The method described above is valid only if there was no fractionation between olivine and chromite (i.e. no preferential incorporation of either phase into the erupted mixture) when the phenocrysts and melt were finally mixed. It is unclear whether or not this criterion was met, and it should be noted that Cr2O3 contents calculated for the parental melt are necessarily approximate. Modelling conditions MELTS crystallization models for the Winnipegosis komatiite were run at a fixed pressure of 1 bar from a temperature above the liquidus (1550 °C) to ∼1000 °C in 1 °C temperature steps, by which point the majority (>90 wt%) of the liquid had crystallized. Although crystallization of olivine and chromite phenocrysts is believed to have predominantly occurred before komatiite eruption (Waterton et al., 2017), low (atmospheric) pressure crystallization is assumed as a starting point for the modelling; the effect of higher crystallization pressures is considered below. Because of differences in the oxygen fugacity of the komatiites calculated using different methods, several models were run, with initial fO2 varying from QFM – 1 to QFM + 1 where QFM is the quartz–fayalite–magnetite buffer (Waterton et al., 2017; Waterton, 2018; Nicklas et al., 2019). Models were run under both fixed and freely varying fO2 conditions. The parental melt composition was assumed to contain 0·1 wt% H2O, the maximum value calculated assuming fixed H2O/Ce (Dixon et al., 2002; Waterton et al., 2017). However, results of the 0·1 wt% hydrous model were indistinguishable from those using an anhydrous composition. During the modelling, it was found that removal of small mass increments of minerals, particularly chromite, led to a large overestimation of their abundances and an associated gain in total mass during crystallization. As such, all models were run with the parental melt wt% composition scaled to a mass of 10 000 g. Komatiite MELTS crystallization models MELTS crystallization sequence Based on a combination of petrographical observations, and comparisons between olivine–liquid Fe–Mg exchange and Al-in-olivine thermometry, Waterton et al., (2017) determined the crystallization sequence for the Winnipegosis komatiites to be as follows: olivine as the liquidus phase, then olivine + chromite, followed by a gap in temperature before the crystallization of clinopyroxene (augite/diopside), and finally clinopyroxene + plagioclase (as observed in the groundmass of some differentiated flow samples). Most of the MELTS models were consistent with this crystallization sequence, and predicted olivine as the liquidus phase. However, the stability of Cr-spinel is predicted to increase with increasing fO2, and spinel was predicted to be the liquidus phase in the most oxidizing model, with an fO2 of QFM + 1. The MELTS models predict that olivine crystallization effectively ceases at the onset of clinopyroxene crystallization, which is consistent with the petrographical observation that olivine never terminates on clinopyroxene (Waterton et al., 2017). Although clinopyroxene was predicted to crystallize before plagioclase in all models, the temperature interval between clinopyroxene and plagioclase saturation is predicted to decrease with decreasing fO2. MELTS liquidus temperature and olivine compositions The temperature of olivine saturation varied from 1498 °C at QFM + 1 to 1508 °C at QFM − 1, in extremely good agreement with the olivine liquidus temperature of 1501 ± 32 °C calculated from Fe–Mg olivine–melt partitioning (Waterton et al., 2017). Despite this agreement, liquidus olivine compositions are more magnesian than observed in Winnipegosis komatiites, varying from an Mg# of 0·944 at QFM + 1 to Mg# 0·932 at QFM – 1. This mismatch is a result of a previously documented underestimation of the olivine–liquid Fe–Mg exchange coefficient (KD) by MELTS (Balta & McSween, 2013). Higher Mg# values are predicted under more oxidizing conditions due to both lower Fe2+/∑Fe in the melt and MELTS predicting lower KD values at higher fO2 (Fig. 9a). The effect of this KD underestimation on the LLD is investigated by comparing the liquid lines of descent predicted by MELTS with those calculated using incremental olivine subtraction from the same parental melt composition (Albarede, 1992; Larsen & Pedersen, 2000; Waterton et al., 2017) and KD values from experimental studies (Herzberg & O’Hara, 2002; Matzen et al., 2011). Although the olivine subtraction calculations do not include the effect of chromite crystallization, MELTS models run without permitting spinel crystallization indicate that this should have an effect of <0·1 wt% FeOt on the FeO–MgO LLD, due to the small proportion of chromite in the crystallizing assemblage. Liquid lines of descent predicted by MELTS have systematically high FeOt contents relative to those calculated using experimentally derived KD values, with an increasing mismatch in FeOt as crystallization progresses. However, this has a relatively small effect on the LLD, with FeOt overestimated by a maximum of ∼1 wt% FeOt after 35 wt% olivine crystallization under the most oxidizing conditions modelled (Fig. 9b). MELTS chromite Chromite saturation temperatures in the MELTS models vary from 1457 °C at QFM − 1 to 1512 °C at QFM + 1. These temperatures are significantly higher than the highest measured Al-in-olivine temperatures (1424 ± 25 °C; Waterton et al., 2017), which indicate the onset of co-crystallization of olivine and chromite. We stress that Al-in-olivine temperatures are not liquidus temperatures in the Winnipegosis komatiites due to an extensive temperature interval of olivine crystallization before chromite saturation (Waterton et al., 2017). MELTS also overestimates the ratio of chromite:olivine crystallizing relative to the average proportions of olivine and chromite phenocrysts calculated from Winnipegosis komatiite mixing lines (1·1 ± 0·1 wt% chromite; Fig. 10a; Waterton et al., 2017) under all modelled conditions. These observations suggest that the MELTS models overestimate the stability of chromite (Hirschmann et al., 1998; Matzen et al., 2011; Balta & McSween, 2013). Fig. 10. Open in new tabDownload slide (a) Proportion of chromite in the crystallizing olivine + chromite mixture predicted by MELTS as a function of temperature, compared with an average of 1·1 ± 0·1 wt% chromite in phenocryst assemblage in temperature interval ∼1501–1321 °C (liquidus to estimated eruption temperature), derived from komatiite bulk-rock data (black line and grey field; Waterton et al. 2017). (b) Comparison of chromite compositions predicted by MELTS as a function of temperature, with chromites included in olivine analysed from Winnipegosis komatiite samples (kom. chr.), and chromites from RP7 dunites (dun. chr.). Chromite that had not gained Fe2+ during low-temperature equilibration is highlighted (ol-chr_4). Fig. 10. Open in new tabDownload slide (a) Proportion of chromite in the crystallizing olivine + chromite mixture predicted by MELTS as a function of temperature, compared with an average of 1·1 ± 0·1 wt% chromite in phenocryst assemblage in temperature interval ∼1501–1321 °C (liquidus to estimated eruption temperature), derived from komatiite bulk-rock data (black line and grey field; Waterton et al. 2017). (b) Comparison of chromite compositions predicted by MELTS as a function of temperature, with chromites included in olivine analysed from Winnipegosis komatiite samples (kom. chr.), and chromites from RP7 dunites (dun. chr.). Chromite that had not gained Fe2+ during low-temperature equilibration is highlighted (ol-chr_4). Chromite compositions predicted by MELTS in the temperature range 1350 °C < T < 1500 °C have similar Cr# values [Cr# = Cr/(Cr + Al)] to Winnipegosis chromites used for Al-in-olivine analyses (Fig. 10b). However, similar to the predicted olivine compositions, they show elevated Mg# [Mg# = Mg/(Mg + Fe2+)] relative to all measured compositions. Although some chromites are likely to have gained Fe2+ during low-temperature equilibration with their host olivine, the MELTS chromite Mg# values are high even relative to chromite from an olivine–chromite pair that showed magmatic Al-in-olivine and Fe–Mg equilibration temperatures of ∼1400 °C (RP1A-111_ol-chr_4), which had not undergone low-temperature Fe gain (Waterton et al., 2017). Despite these errors in estimated Mg#, this is expected to have no appreciable effect on the LLD for MgO and FeO as chromite forms a negligible proportion of the bulk-rock FeO and MgO budget. Because of the overestimation of chromite stability, it might be expected that Cr2O3 would fall too rapidly in the MELTS calculations relative to the true LLD. However, a comparison of the MELTS Cr2O3–MgO LLD with experimentally derived Cr solubilities (Fig. 11a) shows that MELTS provides a good approximation to the expected LLD. This is because MELTS does not allow for Cr2O3 incorporation into olivine, which is a major contributor to the whole-rock Cr2O3 budget due to its volumetric dominance, despite low Cr2O3 concentrations (Supplementary Data). The MELTS calculated LLD can therefore still plausibly generate the observed bulk-rock Cr2O3–MgO mixing lines (Fig. 11b). Fig. 11. Open in new tabDownload slide (a) Comparison of the Cr–MgO LLD predicted by MELTS with experimentally derived Cr solubility of Murck & Campbell (1986) (M & C 1986), using the best-fit curve of Barnes (1998). Compositions of measured highest Mg# olivine (measured ol) and liquidus olivine predicted by MELTS (MELTS ol) are shown, with tie-lines to the komatiite parental melt (parental) and arrows indicating the effect on melt composition by removal of these olivines. Lack of Cr in MELTS olivine causes a steeper LLD, and Cr solubility is reached after less olivine crystallization (at higher temperatures). (b) Cr2O3–MgO liquid lines of descent predicted by MELTS, compared with komatiite bulk-rock data from borehole RP1A. The bulk-rock data approximately fall along mixing lines (Mix QFM-1, Mix QFM) between the MELTS liquid lines of descent at the estimated eruption temperature (1321 °C) and the composition of the olivine + chromite phenocryst assemblage calculated for the komatiites (Waterton et al., 2017). Fig. 11. Open in new tabDownload slide (a) Comparison of the Cr–MgO LLD predicted by MELTS with experimentally derived Cr solubility of Murck & Campbell (1986) (M & C 1986), using the best-fit curve of Barnes (1998). Compositions of measured highest Mg# olivine (measured ol) and liquidus olivine predicted by MELTS (MELTS ol) are shown, with tie-lines to the komatiite parental melt (parental) and arrows indicating the effect on melt composition by removal of these olivines. Lack of Cr in MELTS olivine causes a steeper LLD, and Cr solubility is reached after less olivine crystallization (at higher temperatures). (b) Cr2O3–MgO liquid lines of descent predicted by MELTS, compared with komatiite bulk-rock data from borehole RP1A. The bulk-rock data approximately fall along mixing lines (Mix QFM-1, Mix QFM) between the MELTS liquid lines of descent at the estimated eruption temperature (1321 °C) and the composition of the olivine + chromite phenocryst assemblage calculated for the komatiites (Waterton et al., 2017). MELTS clinopyroxene crystallization Clinopyroxene crystallization is predicted to commence between 1203 and 1193 °C, at liquid MgO contents of 9·9–8·6 wt%, with lower temperatures and liquid MgO contents at clinopyroxene saturation favoured by lower fO2. Although the exact temperature and MgO content at which clinopyroxene began to crystallize in Winnipegosis komatiites is difficult to establish, a number of observations are consistent with the MELTS modelling. As clinopyroxene dendrites always terminate on olivine dendrites and phenocrysts in the massive komatiite flows, the lowest Mg# olivine (Mg# = 0·87) and lowest Al-in-olivine temperatures (1321 ± 25 °C) provide an upper limit on the onset of clinopyroxene crystallization at liquids containing 14·6 ± 0·7 wt% MgO and 14·2 ± 1·3 wt% MgO, respectively (Waterton et al., 2017). This upper limit is confirmed by the observation that no pyroxene phenocrysts are observed in chill margins, which, correcting for the petrographically observed olivine phenocryst contents, have liquid MgO contents as low as 14·1 wt% MgO. Additionally, all samples adhere to olivine control lines for elements compatible in clinopyroxene, such as Sc, and Pearce element ratio plots indicate that no clinopyroxene was fractionated across the entire range of bulk compositions represented (Fig. 12). As no clinopyroxene was fractionated even to produce the acicular pyroxene samples in differentiated flows (10·6 wt% MgO), pyroxene crystallization must have commenced at liquid compositions with ≤10·6 wt% MgO. Fig. 12. Open in new tabDownload slide Investigating clinopyroxene fractionation in Winnipegosis komatiite samples. (a) Plot of Sc against MgO for all Winnipegosis komatiite samples from boreholes RP1A and RP12, with linear regressions of RP1A data and 95 % confidence limits (Waterton et al., 2017). An average value of 5·8 ppm Sc measured from LA-ICP-MS analyses of Winnipegosis komatiite olivine (P. Waterton, unpublished data) is plotted at the average olivine composition of 49 wt% MgO. Sc, compatible in clinopyroxene, is correlated with MgO, with all samples falling on an olivine control line. This indicates that clinopyroxene is not an important fractionating phase. (b) Pearce element ratio plot (Pearce, 1968; Russell et al., 1990), showing molar element ratios for Winnipegosis komatiites. Olivine fractionation produces arrays with a gradient of one on this plot; augite fractionation has a gradient of 0·33, based on the average composition of acicular pyroxene phenocrysts in sample RP12-307.3; plagioclase fractionation produces horizontal arrays; and chromite fractionation produces vertical arrays. Winnipegosis samples form an array with slope = 0·99 (Regression), indistinguishable from an olivine control line (Olivine). Fig. 12. Open in new tabDownload slide Investigating clinopyroxene fractionation in Winnipegosis komatiite samples. (a) Plot of Sc against MgO for all Winnipegosis komatiite samples from boreholes RP1A and RP12, with linear regressions of RP1A data and 95 % confidence limits (Waterton et al., 2017). An average value of 5·8 ppm Sc measured from LA-ICP-MS analyses of Winnipegosis komatiite olivine (P. Waterton, unpublished data) is plotted at the average olivine composition of 49 wt% MgO. Sc, compatible in clinopyroxene, is correlated with MgO, with all samples falling on an olivine control line. This indicates that clinopyroxene is not an important fractionating phase. (b) Pearce element ratio plot (Pearce, 1968; Russell et al., 1990), showing molar element ratios for Winnipegosis komatiites. Olivine fractionation produces arrays with a gradient of one on this plot; augite fractionation has a gradient of 0·33, based on the average composition of acicular pyroxene phenocrysts in sample RP12-307.3; plagioclase fractionation produces horizontal arrays; and chromite fractionation produces vertical arrays. Winnipegosis samples form an array with slope = 0·99 (Regression), indistinguishable from an olivine control line (Olivine). In all models, the first clinopyroxene to crystallize is a high-Ca augite, and its predicted composition varies insignificantly with modelled fO2; MgO, CaO, and FeOt are all roughly constant at ∼17 wt%, ∼20 wt%, and ∼7 wt%, respectively. Clinopyroxene compositions are then predicted to evolve, broadly, towards lower CaO and higher FeOt at roughly constant MgO, until saturation of a second, low-Ca clinopyroxene at 1167–1145 °C. These predicted pyroxene compositions cannot be compared with clinopyroxene in the massive komatiite flows, nor in the olivine spinifex zones of differentiated flows, as the dendritic nature of these pyroxenes indicates that they did not grow under equilibrium conditions. However, clinopyroxene phenocrysts from the acicular pyroxene zone of a differentiated flow (‘Aci. px’, sample RP12-307.3; Fig. 7; Supplementary Data) have habits indicative of conditions closer to equilibrium crystallization, and compositions close to augite grown in experiments under anhydrous equilibrium conditions (Parman et al., 1997). MELTS predicted compositions fall close to the compositions of the acicular clinopyroxene phenocrysts, although MELTS compositions are higher in CaO and lower in MgO than the observed phenocrysts, which have MgO and CaO contents of ∼18–20 and 15–18 wt% respectively (Fig. 7). This contrasts with MELTS predictions for mantle melting, where clinopyroxene compositions contain too little CaO due to an overestimation of melting temperatures by MELTS (Hirschmann et al., 1998). Therefore, following clinopyroxene saturation, the LLD predicted by MELTS will be too steep in MgO–CaO space compared with an LLD generated by removal of the observed phenocrysts; that is, CaO will fall by too much over a given interval of decreasing MgO. However, errors introduced as a result of these differences in clinopyroxene composition are likely to be small compared with the differences in the CaO–MgO LLD arising from the varying temperature of clinopyroxene saturation with pressure (Fig. 13). Fig. 13. Open in new tabDownload slide MELTS predicted liquid lines of descent for CaO and Al2O3 as a function of MgO, compared with RP8 basalt bulk-rock compositions. (a, b) Variation in predicted liquid lines of descent with pressure of crystallization varying from 1 bar to 4 kbar. fO2 was initially set to QFM at the liquidus but allowed to vary during crystallization (QFMv). Compositions at clinopyroxene saturation (Cpx sat.) are indicated in both figures, with temperatures of saturation at 1 bar, 2 kbar, and 4 kbar indicated in (a). Compositions and temperatures at plagioclase saturation (Plag sat) are indicated in (b). Higher pressures lead to a larger gap in temperature and MgO between clinopyroxene and plagioclase saturation, causing greater Al2O3 enrichment in the melt. Orthopyroxene (opx) saturation in the 4 kbar model is indicated. (c, d) Variation in the predicted LLD with fO2 at a fixed pressure of 2 kbar. Oxygen fugacity was fixed during crystallization to maintain constant fO2 differences. Lower fO2 also promotes greater Al2O3 enrichment in the melt by delaying both clinopyroxene and plagioclase saturation to lower MgO. Fig. 13. Open in new tabDownload slide MELTS predicted liquid lines of descent for CaO and Al2O3 as a function of MgO, compared with RP8 basalt bulk-rock compositions. (a, b) Variation in predicted liquid lines of descent with pressure of crystallization varying from 1 bar to 4 kbar. fO2 was initially set to QFM at the liquidus but allowed to vary during crystallization (QFMv). Compositions at clinopyroxene saturation (Cpx sat.) are indicated in both figures, with temperatures of saturation at 1 bar, 2 kbar, and 4 kbar indicated in (a). Compositions and temperatures at plagioclase saturation (Plag sat) are indicated in (b). Higher pressures lead to a larger gap in temperature and MgO between clinopyroxene and plagioclase saturation, causing greater Al2O3 enrichment in the melt. Orthopyroxene (opx) saturation in the 4 kbar model is indicated. (c, d) Variation in the predicted LLD with fO2 at a fixed pressure of 2 kbar. Oxygen fugacity was fixed during crystallization to maintain constant fO2 differences. Lower fO2 also promotes greater Al2O3 enrichment in the melt by delaying both clinopyroxene and plagioclase saturation to lower MgO. MELTS plagioclase crystallization Plagioclase was predicted to saturate at ∼1192 °C in all 1 bar crystallization models, showing no dependence on fO2. The first plagioclase to saturate in all models was Ca rich, with molar anorthite (An) contents of ∼80 %; albite contents are predicted to increase at the expense of An with continued crystallization. No plagioclase phenocrysts were observed in Winnipegosis komatiites against which to assess the MELTS predicted plagioclase compositions. However, the predicted compositions are similar to granular plagioclase compositions in coarse spinifex-textured komatiite from Gorgona Island (up to An83; Echeverria, 1980), and so the MELTS predicted plagioclase compositions appear broadly reasonable. Modelling the major element composition of RP8 basalts To summarize, MELTS appears able to adequately model the low-pressure komatiite LLD from the liquidus temperature to ∼1200 °C to a good degree of precision, with two main caveats. First, FeOt in the MELTS liquid is likely to be overestimated by up to ∼1 wt% at the onset of clinopyroxene saturation, due to errors in the MELTS predicted olivine–melt Fe–Mg exchange coefficient. Second, the prediction of clinopyroxene with higher CaO contents and lower MgO contents than observed may cause the LLD to evolve along a trajectory that is too steep in CaO–MgO space, causing slightly underestimated CaO at a given MgO after clinopyroxene saturation. With these caveats in mind, MELTS is used to model the formation of the RP8 basalts through fractional crystallization of the RP1A komatiitic parental melt. Although the effects of metamorphism on the basalt compositions have not been established because of a lack of pristine samples or primary minerals, the modelling below first investigates whether the observed compositions can be explained by igneous processes alone, without invoking metamorphic processes. Effect of pressure Extensive crystallization of a komatiitic melt to a basalt would probably require the presence of a magma chamber at depth, and so the effect of different crystallization pressures on the LLD must be considered. MELTS models run with freely varying fO2 starting from QFM show olivine saturation temperatures increasing by ∼8 °C per kbar, from 1505 °C at 1 bar to 1534 °C at 4 kbar. However, olivine compositions are almost constant between models and so this is expected to have no significant effect on the LLD. Chromite saturation temperatures are also predicted to increase with depth, and between 3 and 4 kbar Cr-spinel becomes the liquidus phase. Again, this is not expected to significantly affect the LLD for most elements except Cr2O3. The main effect of crystallization pressure on the LLD predicted by MELTS is a difference in the relative onset of clinopyroxene and plagioclase crystallization. Increased pressures cause both pyroxene and plagioclase to saturate at higher temperatures, but the effect is greater for pyroxene, leading to relatively earlier crystallization of clinopyroxene (e.g. O’Hara, 1968; Kinzler & Grove, 1992) and a larger temperature gap between pyroxene and plagioclase saturation at higher pressures. This predicted gap increases from 8 °C at a pressure of 1 bar (clinopyroxene saturation at 1201 °C, plagioclase saturation at 1193 °C) to 51 °C at a pressure of 4 kbar (clinopyroxene saturation at 1275 °C, plagioclase saturation at 1224 °C). This is manifested in the LLD as a larger gap in MgO between CaO beginning to fall and Al2O3 rising more rapidly (upon clinopyroxene saturation), and Al2O3 beginning to fall (upon plagioclase saturation; Fig. 13a and b). Notably, this causes maximum Al2O3 enrichment in the evolving melt to increase with pressure. The best fit to the basalt Al2O3–MgO data for crystallization of the komatiitic parent with initial fO2 of QFM is for a pressure of ∼2 kbar. However, this is not a unique solution becaue of a trade-off with modelled fO2; at lower fO2, clinopyroxene crystallizes at lower MgO (Fig. 13c and d). This causes similar Al2O3 enrichment at lower crystallization pressures, with a 1 log unit decrease in fO2 having approximately the same effect as a 0·5 kbar increase in pressure. Uncertainties in the parental melt Al2O3 composition of 8·50 ± 0·57 wt% might be expected to have a similar effect. However, uncertainties in Al2O3 and many other elements in the komatiites are strongly correlated with MgO uncertainties along the olivine control lines used to estimate the parental melt composition. Uncertainties in the parental melt composition therefore largely move the parental melt composition along olivine control lines, leading to almost identical liquid lines of descent (not shown). However, the amount of fractional crystallization required to produce basaltic MgO contents varies by ∼4 wt% over the range of MgO contents calculated for the parental melt (23·6 ± 1·6 wt% MgO). CaO contents in some basalt samples are slightly underestimated in the variable fO2 models (Fig. 13a and b). This may be explained by the overestimated CaO/MgO in the MELTS clinopyroxene compositions, or by the evolution of the modelled melts to higher fO2 with crystallization (from QFM at the liquidus to ∼QFM + 0·5 at basaltic compositions), which causes earlier clinopyroxene saturation. However, considering the evidence of CaO mobility in sub-greenschist RP1A komatiites (Waterton et al., 2017), and the higher greenschist metamorphic grade and presence of quartz–carbonate veins in RP8 basalts, it is also possible that CaO was mobile during metamorphism. Major element composition of RP8 basalts can be generated by extensive crystallization of a komatiite parental magma By adjusting the pressure and fO2 conditions, the majority of elements in the RP8 basalts can be matched by ∼60 wt% crystallization of an RP1A komatiite parental melt (Fig. 14), between a liquidus temperature of ∼1550 °C and the production of a basaltic composition at ∼1185 °C. The exact percentage of crystallization required varies with the pressure, fO2, uncertainties in the composition of the starting komatiite melt and the range in RP8 basalt compositions. For the average calculated RP1A komatiite parental melt (23·5 wt% MgO) crystallizing to the average MgO of RP8 basalts (7·5 wt% MgO), 57 %, 62 %, or 66 % crystallization is required for crystallization at 1·5 kbar and QFM – 1, 2 kbar and QFM, or 2·5 kbar and QFM + 1 conditions respectively. Fig. 14. Open in new tabDownload slide MELTS predicted liquid lines of descent for komatiite parental melt (parental) with initial fO2 = QFM − 1 and pressure = 1·5 kbar (QFM-1v 1·5 kb), initial fO2 = QFM and pressure = 2 kbar (QFMv 2 kb), and initial fO2 = QFM + 1 and pressure = 2·5 kbar (QFM + 1v 2·5 kb), compared with RP1A komatiite and RP8 basalt data. fO2 was allowed to vary in all models. Liquid lines of descent show an excellent match to basalt compositions for all major elements with the exception of FeOt and SiO2. Komatiite bulk-rock data follow mixing trends between the phenocryst assemblage and residual melt, which are oblique to the LLD for elements compatible or mildly incompatible in olivine or chromite (Mixing). Na2O follows trends oblique to the LLD for both the basalts and komatiites, probably due to metamorphism (arrows), although MELTS liquid lines of descent still pass through the average basalt composition. SiO2 plot shows a mixing line from the MELTS liquid lines of descent through the average RP8 basalt composition to quartz (Mix quartz). FeOt plot also shows the LLD calculated using the Matzen et al. (2011) olivine–melt KD value as MELTS underestimates KD leading to overestimated FeOt contents in the melt. Plausible liquid lines of descent accounting for this FeOt overestimation are calculated by subtracting 1 wt% FeOt from the MELTS predictions after their intersection with the Matzen et al. (2011) LLD (grey lines). Fig. 14. Open in new tabDownload slide MELTS predicted liquid lines of descent for komatiite parental melt (parental) with initial fO2 = QFM − 1 and pressure = 1·5 kbar (QFM-1v 1·5 kb), initial fO2 = QFM and pressure = 2 kbar (QFMv 2 kb), and initial fO2 = QFM + 1 and pressure = 2·5 kbar (QFM + 1v 2·5 kb), compared with RP1A komatiite and RP8 basalt data. fO2 was allowed to vary in all models. Liquid lines of descent show an excellent match to basalt compositions for all major elements with the exception of FeOt and SiO2. Komatiite bulk-rock data follow mixing trends between the phenocryst assemblage and residual melt, which are oblique to the LLD for elements compatible or mildly incompatible in olivine or chromite (Mixing). Na2O follows trends oblique to the LLD for both the basalts and komatiites, probably due to metamorphism (arrows), although MELTS liquid lines of descent still pass through the average basalt composition. SiO2 plot shows a mixing line from the MELTS liquid lines of descent through the average RP8 basalt composition to quartz (Mix quartz). FeOt plot also shows the LLD calculated using the Matzen et al. (2011) olivine–melt KD value as MELTS underestimates KD leading to overestimated FeOt contents in the melt. Plausible liquid lines of descent accounting for this FeOt overestimation are calculated by subtracting 1 wt% FeOt from the MELTS predictions after their intersection with the Matzen et al. (2011) LLD (grey lines). The exceptions to this are FeOt, which is overestimated because of the errors in KD outlined above, and SiO2, which is systematically underestimated by 2–4 wt% for conditions that replicate the other major elements. Basalt FeOt data can be matched by accounting for an ∼1 wt% overestimation of FeOt by the MELTS models (Fig. 14). However, MELTS predicted igneous mineral assemblage at basaltic compositions (dominated by clinopyroxene and plagioclase) cannot account for the enrichment in SiO2 observed in the RP8 basalts, or the steep trend of RP8 basalt SiO2 data against MgO. Estimated analytical uncertainties for basalt SiO2 measurements of ±2 % (2σ relative), or approximately ±1 wt%, are insufficient to explain this discrepancy. Three possibilities are considered to explain the elevated SiO2 relative to modelled compositions. First, the komatiitic parental melt could have been further crustally contaminated by SiO2-rich material during crystallization, beyond the ∼2 % crustal contamination calculated for Winnipegosis komatiite (Waterton et al., 2017). Second, SiO2 may have been elevated during metamorphic processes, such as formation of the petrographically observed secondary quartz veins. Third, high SiO2 may indicate that the basalts were derived from a different parental melt from the komatiites (Aitken & Echeverría, 1984). The first two of these possibilities should be distinguishable using the REE systematics of the basalts and komatiites, as potential crustal contaminants such as the tonalitic compositions invoked to explain the incompatible element systematics of the komatiites (Waterton et al., 2017) should be elevated in LREE. Modelling the REE composition of RP8 basalts To evaluate whether elevated SiO2 in RP8 basalts relative to MELTS models is a result of crustal contamination, a REE partitioning model is attached to the outputs of the MELTS crystallization models above that successfully reproduce the major element compositions of the RP8 basalts. The model is run both for fractional crystallization alone (FC), and for assimilation–fractional crystallization (AFC; Taylor, 1980; DePaolo, 1981). The AFC calculations use the local tonalitic crust through which the Winnipegosis komatiites erupted (Waterton et al., 2017) as an assimilant, assume an r value of 0·3 (where r is the ratio of mass of assimilation to mass of crystallization at each step), and assume the crystallizing assemblage is unchanged from the FC models. At each 1 °C temperature increment, REE are partitioned between the liquid and the solids fractionated in that step, and REE in the solid fraction are removed. For the AFC models, REE abundances calculated for the amount of crust assimilated in that step are added to the remaining liquid before the next step. Chromite mineral/liquid REE partition coefficients (⁠ DREEmin–liq ⁠, hereafter D) are assumed to be zero due to its minor mass and lack of partition coefficients for natural compositions, although experiments on synthetic compositions suggest that D values are low (Nagasawa et al., 1980). Augite and pigeonite D values are calculated as a function of clinopyroxene Wo content, after McKay et al., (1986). Plagioclase D values are taken from QFM, 1200 °C experimental runs of Aigner-Torres et al. (2007), to match the approximate temperature and oxygen fugacity of plagioclase crystallization in the MELTS models. Olivine D values are calculated using the lattice strain model of Sun & Liang (2013), which incorporates pressure, temperature, olivine Fo content, and melt Al content. Terminating the FC models within the range of basalt MgO (7·2–7·9 wt%), all of the FC models are able to reproduce the REE patterns of RP8 basalts (Fig. 15a). For the partition coefficients used, all REE are incompatible in olivine, augite (pyroxene) and plagioclase (Fig. 15b). Therefore, fractional crystallization alone has little effect on the shape of REE patterns, although the removal of olivine and particularly clinopyroxene (augite), which are relatively enriched in HREE, causes a slight HREE depletion in the modelled melts. Uncertainties in the komatiite parental melt composition (23·6 ± 1·6 wt% MgO) have little effect on the REE patterns generated due to the error correlation along olivine control lines described previously; a higher MgO parental melt has lower initial REE contents but must crystallize more olivine to reach basaltic compositions, driving REE contents up correspondingly. Fig. 15. Open in new tabDownload slide Models of REE concentrations based on MELTS predicted crystallization of an RP1A komatiite parental melt (Table 6). (a) REE concentrations predicted for crystallization of an RP1A komatiite parental melt to the average MgO content of RP8 basalts (7·5 wt%) MgO, compared with the calculated RP1A komatiite parental melt and average RP8 basalt. This corresponds to 57 %, 62 %, or 66 % crystallization by mass for crystallization at 1·5 kbar and QFM – 1, 2 kbar and QFM, and 2·5 kbar and QFM + 1 conditions, respectively. (b) Calculated REE abundances in the first olivine, clinopyroxene, and plagioclase to crystallize in the 2 kbar QFM model, labelled by their MELTS predicted saturation temperature. (c) Comparison of fractional crystallization (FC) and assimilation–fractional crystallization (AFC), of an RP1A komatiite parental melt after 20 %, 40 %, and 62 % crystallization. AFC models cannot reproduce the RP8 basalt REE patterns, whereas FC alone produces a very good match for 62 % crystallization, the amount of crystallization required to reach average RP8 basalt MgO of 7·5 wt%. Fig. 15. Open in new tabDownload slide Models of REE concentrations based on MELTS predicted crystallization of an RP1A komatiite parental melt (Table 6). (a) REE concentrations predicted for crystallization of an RP1A komatiite parental melt to the average MgO content of RP8 basalts (7·5 wt%) MgO, compared with the calculated RP1A komatiite parental melt and average RP8 basalt. This corresponds to 57 %, 62 %, or 66 % crystallization by mass for crystallization at 1·5 kbar and QFM – 1, 2 kbar and QFM, and 2·5 kbar and QFM + 1 conditions, respectively. (b) Calculated REE abundances in the first olivine, clinopyroxene, and plagioclase to crystallize in the 2 kbar QFM model, labelled by their MELTS predicted saturation temperature. (c) Comparison of fractional crystallization (FC) and assimilation–fractional crystallization (AFC), of an RP1A komatiite parental melt after 20 %, 40 %, and 62 % crystallization. AFC models cannot reproduce the RP8 basalt REE patterns, whereas FC alone produces a very good match for 62 % crystallization, the amount of crystallization required to reach average RP8 basalt MgO of 7·5 wt%. By contrast, AFC models cannot reproduce the REE patterns of RP8 basalts for any amount of crystallization and assimilation. Assimilation–fractional crystallization leads to elevated LREE/MREE, which are not observed in RP8 basalts. We conclude that the high SiO2 in RP8 basalts cannot be caused by crustal assimilation. Formation of the RP8 basalts To summarize, almost all analysed aspects of the RP8 basalt geochemistry can be explained by ∼60 % fractional crystallization of a komatiite with the composition of the RP1A komatiite parental melt, with a best-fit MELTS pressure of crystallization of 1·5–2·5 kbar, for fO2 in the range QFM – 1 to QFM + 1. The main exceptions to this are high SiO2 contents in RP8 basalts, with steep slopes in SiO2 against MgO. These are best explained in this model by metamorphic effects such as fluid flow and quartz veining (Fig. 14) during metamorphism. It is stressed here that these best-fit pressures and oxygen fugacities are not precisely constrained, as the success of the MELTS model in predicting the LLD at non-atmospheric pressures could not be evaluated. However, the underlying reasons for the variations in LLD (reduced plagioclase stability compared with clinopyroxene at higher pressure) and approximate fO2 both appear geologically reasonable. Could RP8 basalts have formed by lower degree melting of a similar mantle source to the komatiites? An alternative explanation for the similarity in REE patterns between RP8 basalts and the RP1A komatiites is that the basalts formed from a similarly depleted mantle source that experienced a lower degree of partial melting. This also has the potential to explain the high SiO2 in RP8 basalts relative to models of komatiite crystallization, as SiO2 is expected to increase with decreasing mantle potential temperature and associated shallower depths of melting initiation (e.g. Asimow & Longhi, 2004; Lee et al., 2009). This can be envisioned as the basalts forming through passive rifting before and/or after plume impingement generated the komatiites, forming in the cooler edges of a plume spreading beneath the lithosphere (spatial variations in plume temperature; e.g. Campbell et al., 1989), or forming during a cooler pulse of plume activity (temporal variations in plume temperature; e.g. White et al., 1995). We first address this by investigating whether the Winnipegosis RP8 basalts can represent primary mantle melts, or require the differentiation of a more mafic parental magma. RP8 basalts contain an average of 7·5 wt% MgO, which, as noted in the Introduction, is significantly lower than present-day mid-ocean ridge basalt (MORB) primary magmas (9–13 wt% MgO; Kamenetsky, 1996; Sobolev, 1996; Sobolev & Chaussidon, 1996; Laubier et al., 2007) and predicted primary melts of Palaeoproterozoic ambient mantle (up to 18–24 wt% MgO; Herzberg et al., 2010). Furthermore, RP8 basalts have an average Mg# of 0·50, and are too Fe rich and MgO poor to have formed even by near solidus melting of mantle peridotite at low pressures (Herzberg & O’Hara, 2002). Use of Primelt3 (Herzberg & Asimow, 2015) on RP8 basalts fails to produce primary melt solutions; addition of olivine alone to RP8 compositions produces unusual melts rich in FeOt and TiO2 but very poor in CaO. This is best explained by RP8 basalts forming from a parental melt that fractionated plagioclase and clinopyroxene in addition to olivine. Alternatively, the RP8 basalts could have formed from a different mantle source (e.g. pyroxenite), but this would require the close similarity in REE patterns between the basalts and komatiites to be purely coincidental, which we consider unlikely. As the RP8 basalts do not appear to be primary mantle melts, and very probably required some degree of crystal fractionation in their formation, the question of whether they formed from a lower degree melt requires a fine distinction to be made. Either the parental melt was a komatiite, which fractionated extensively to form the basalts, or lower degrees of melting (in a cooler portion or pulse of the plume, or during passive rifting) generated a picritic or high-Mg basalt parental melt, which then fractionated to form the RP8 basalts. Making this distinction is difficult for a number of reasons. The RP8 basalts lack primary minerals, and exhibit evidence of metamorphic disturbance of some elements such as SiO2 and Na2O, the latter which can be used to indicate depths and extents of melting (Klein & Langmuir, 1987). Furthermore, our MELTS modelling shows that a wide range of basalt compositions (particularly Al2O3 and CaO contents) can be produced from the same primary melt depending on the pressure and fO2 of crystallization. Therefore, ratios such as CaO/Al2O3 and Al2O3/TiO2 traditionally used to distinguish basalts and komatiites cannot be used, as these are strongly affected by the degree and the P–T–fO2 conditions of crystallization. Finally, for many incompatible elements (e.g. TiO2), there is a trade-off between increasing degree of mantle melting, which decreases their concentrations in the parental melt, and an increasing degree of fractional crystallization required to reach basaltic compositions. We therefore attempt to distinguish a komatiitic parental melt from a lower degree parental melt by considering trace element ratios that are sensitive to depth of melting. Winnipegosis komatiites are intermediate between Al-depleted and Al-undepleted komatiites, with a slight garnet signature of depleted HREE relative to MREE and subchondritic Al2O3/TiO2 (Waterton et al., 2017). Therefore, if the high SiO2 of RP8 basalts relative to MELTS models of komatiite crystallization is to be explained by shallower, lower degree melting, then RP8 basalts would be expected to show a reduced garnet signature. On a Gd/Yb against Al2O3/TiO2 plot (Fig. 16a), RP8 basalts lie at slightly higher Gd/Yb and at lower Al2O3/TiO2 than Winnipegosis komatiites. If the basalts formed by crystallization of a komatiite parental melt, these features are easily explained by fractionation of olivine and clinopyroxene sequestering HREE relative to MREE, and fractionation of plagioclase reducing Al2O3 relative to TiO2. However, if these are interpreted as primary features they would instead imply a greater garnet signature and correspondingly greater average depth of melting for the Winnipegosis basalts compared with the komatiites. On a plot of TiO2/Yb against Nb/Yb (Fig. 16b; Pearce, 2008), RP8 basalts and RP1A komatiites overlap, indicative of similar depths of melting of a similarly depleted source. In detail, the basalts lie at slightly higher TiO2/Yb and Nb/Yb, which can be explained in the komatiite crystallization model by removal of HREE during clinopyroxene fractionation. Again, if interpreted as primary differences between the basalts and komatiites, this would imply a greater depth of melting for the basalts, which cannot explain their elevated SiO2 relative to MELTS models of komatiite crystallization. We conclude that the formation of petrographically observed quartz veins is the best explanation for the elevated SiO2 in RP8 basalts and steep trends in SiO2 against MgO. Fig. 16. Open in new tabDownload slide Indicators of depth of melting for Winnipegosis komatiites and basalts. (a) Gd/Yb against Al2O3/TiO2 plot. Winnipegosis RP8 basalts lie at slightly higher Gd/Yb, and lower Al2O3/TiO2 than Winnipegosis RP1A komatiites, which can be explained by fractionation of clinopyroxene and plagioclase, respectively, from a komatiitic parental melt. (b) TiO2/Yb against Nb/Yb plot (Pearce, 2008). Winnipegosis basalts and komatiite compositions overlap in this plot, although slightly higher TiO2/Yb and Nb/Yb of Winnipegosis basalts can be explained by fractionation of olivine and clinopyroxene with relatively enriched HREE. GEOROC picrite whole-rock data were accessed 7 November 2019, additionally screened to include only samples with ICP-MS or isotope dilution data for Nb, Gd, and Yb. Fig. 16. Open in new tabDownload slide Indicators of depth of melting for Winnipegosis komatiites and basalts. (a) Gd/Yb against Al2O3/TiO2 plot. Winnipegosis RP8 basalts lie at slightly higher Gd/Yb, and lower Al2O3/TiO2 than Winnipegosis RP1A komatiites, which can be explained by fractionation of clinopyroxene and plagioclase, respectively, from a komatiitic parental melt. (b) TiO2/Yb against Nb/Yb plot (Pearce, 2008). Winnipegosis basalts and komatiite compositions overlap in this plot, although slightly higher TiO2/Yb and Nb/Yb of Winnipegosis basalts can be explained by fractionation of olivine and clinopyroxene with relatively enriched HREE. GEOROC picrite whole-rock data were accessed 7 November 2019, additionally screened to include only samples with ICP-MS or isotope dilution data for Nb, Gd, and Yb. A compilation of picrites from the GEOROC database (Sarbas & Nohl, 2008; http://georoc.mpch-mainz.gwdg.de/georoc/) shows that picrites (i.e. lower degree parental melts than komatiites) with overlapping geochemical characteristics to the RP8 basalts do exist elsewhere (Fig. 16). However, no evidence was found of a possible lower degree parental melt, such as a picrite, in the Winnipegosis Komatiite Belt. We consider it unlikely that a lower degree parental melt existed with identical trace element characteristics and identical depth of melting to the Winnipegosis komatiites. Instead, all available geochemical evidence is consistent with the formation of the RP8 basalts through extensive crystallization of a komatiite parental melt, which also erupted as un-fractionated, high-MgO, spinifex-textured rocks in the same greenstone belt. It is concluded that the RP8 basalts most probably formed through ∼60 % crystallization of a komatiitic parent melt, and subsequently experienced minor alteration during greenschist metamorphism. PLUMBING SYSTEM OF THE WINNIPEGOSIS KOMATIITE BELT Formation of the RP8 basalts via extensive crystallization of a komatiitic parental melt is also consistent with all the available geological constraints. Extensive fractionation of komatiitic magmas in the upper crust near the Winnipegosis Komatiite Belt supracrustal rocks is evidenced by the large dunite body intersected by borehole RP7, and numerous gabbro bodies intersected throughout the WKB (McGregor, 2011). Furthermore, the RP7 dunite body is at least ∼5 km × 2·5 km × 0·6 km (7·5 km3 volume), with an erosional upper contact. For the ∼60 wt% crystallization required to generate a basalt from a komatiitic precursor, a magma chamber of this size could process 12·5 km3 of komatiitic melt, generating 5 km3 of basalt. Although this alone may not be enough to generate all of the basalt in the WKB, the complete lack of surface exposure and limited understanding of the subsurface geology based on the small number of boreholes drilled in the WKB make it difficult to constrain how common magma chambers of this size might be. The RP7 dunite body demonstrates that magma chambers capable of producing cubic kilometres of basalt were present in the upper crust at the time of WKB formation. The approximate palaeo-depth of the RP7 dunite body is calculated for comparison with the best-fit pressures identified from the MELTS modelling. Flow tops in the WKB consistently dip towards the west, consistent with seismic fabrics with dips of ∼20° to the west (Fig. 3b; Lucas et al., 1996), suggesting that the entire belt was tilted westward during the Trans-Hudson Orogen. The RP7 dunite body occurs ∼5 km perpendicular to strike from the structurally lowest supracrustal rocks in the WKB, and ∼16·5 km perpendicular to strike from borehole RP8 (Fig. 3a). Assuming stratigraphic relationships in the WKB were broadly preserved through Trans-Hudson deformation, and a dip of ∼20°, the approximate depth of the RP7 dunite body was 5·6 km at the time of RP8 basalt eruption. For comparison, the MELTS best-fit pressures of 1·5–2·5 kbar, and an average crustal density of 2·88 g cm–3 in the WKB (Hosain & Bamburak, 2002), suggest that depths of 5·3–8·8 km are necessary for the basalts to have formed through differentiation of a komatiite. Therefore not only does the RP7 dunite body provide evidence of extensive fractionation of komatiitic magma in the WKB, but it could potentially represent a magma chamber in which differentiation to form the RP8 basalts occurred. It is suggested that, rather than representing melts from different portions of a temperature-zoned plume (e.g. Campbell et al., 1989), the different geochemical compositions and volcanic styles observed in the Winnipegosis Komatiite Belt simply reflect differences in low-pressure fractional crystallization of a komatiitic parental melt. Where the melt was able to rapidly ascend to the surface, minimal fractionation occurred and komatiite flows were formed. If the magma stalled in an upper crustal magma chamber, it fractionated extensively to form basaltic flows and complementary ultramafic cumulates, as previously proposed for basalts and komatiites from Gorgona Island (Revillon et al., 2000; Fig. 17). Fig. 17. Open in new tabDownload slide Schematic diagram of volcanism in the WKB. Komatiitic magma that rises rapidly to the surface forms komatiite flows, but if it stalls in a magma chamber it differentiates to produce basalt and ultramafic cumulates. Komatiitic magma experienced crustal assimilation during ascent (Waterton et al., 2017), but there is no evidence of additional crustal assimilation during fractionation of the komatiitic parental melt to form basalts. Fig. 17. Open in new tabDownload slide Schematic diagram of volcanism in the WKB. Komatiitic magma that rises rapidly to the surface forms komatiite flows, but if it stalls in a magma chamber it differentiates to produce basalt and ultramafic cumulates. Komatiitic magma experienced crustal assimilation during ascent (Waterton et al., 2017), but there is no evidence of additional crustal assimilation during fractionation of the komatiitic parental melt to form basalts. What caused eruption of homogeneous basalt? Dunite compositions in borehole RP7 indicate only small variations in accumulating olivine Mg# over a large section of cumulate, reflecting episodes of magma recharge. However, the compositions of RP8 basalts are extremely uniform. This begs the question, if the basalts are the products of extensive crystallization of multiple batches komatiitic melt, what caused them to erupt with such a uniform composition? Two previously suggested fluid dynamic models can potentially explain the uniform composition of RP8 basalts. Sparks & Huppert (1984) noted that removal of dense phases such as olivine and pyroxene decrease the density of a fractionating magma. However, just prior to plagioclase saturation (which is less dense than the magma), there is a minimum in magma density. This density minimum means that magmas that have fractionated plagioclase and more primitive magmas injected at the base of the magma chamber should have similar densities, promoting mixing. This can cause the growth of a zone of constant melt composition in the magma chamber. MELTS predicted densities are consistent with this model, as a broad density minimum exists between clinopyroxene saturation at ∼10 wt% MgO (∼1250 °C) and plagioclase saturation at ∼8 wt% MgO (∼1210 °C; Fig. 18), particularly pronounced at lower pressures of crystallization. This would allow production of a constant composition melt at ∼8–10 wt% MgO, which may have undergone additional fractionation upon ascent to the surface to generate RP8 basalts with ∼7·5 wt% MgO. This clustering of basalt compositions close to the point of plagioclase saturation has been also been noted for the Fortescue large igneous province of the Pilbara Craton, and similarly interpreted as resulting from an efficiently mixed open-system magma chamber (Mole et al., 2018). We stress that this may not be an eruption trigger in itself, but a plausible means of generating a melt with very consistent composition despite extensive low-pressure fractionation. Fig. 18. Open in new tabDownload slide MELTS predicted densities during fractionation of a komatiite parent melt to produce basalt, for best-fit MELTS models. Saturation of olivine (ol), clinopyroxene (cpx), and plagioclase (plag) are indicated. MgO range of RP8 basalts is indicated by the grey band. Average density of local Archaean tonalite through which the WKB volcanic rocks erupted (east of the WKB) is also shown, calculated from gravity profiles (Hosain & Bamburak, 2002). Fig. 18. Open in new tabDownload slide MELTS predicted densities during fractionation of a komatiite parent melt to produce basalt, for best-fit MELTS models. Saturation of olivine (ol), clinopyroxene (cpx), and plagioclase (plag) are indicated. MgO range of RP8 basalts is indicated by the grey band. Average density of local Archaean tonalite through which the WKB volcanic rocks erupted (east of the WKB) is also shown, calculated from gravity profiles (Hosain & Bamburak, 2002). Second, Dufek & Bachmann (2010) suggested that crystal–liquid separation is most efficient in a mixture with between 50 and 70 % crystals, as this crystal fraction shuts down chamber-wide convection, promoting compaction. Around 60 % crystallization of a komatiitic melt is required to generate RP8 basalt compositions, consistent with the window of predicted maximum melt segregation efficiency. The constant composition of RP8 basalts can therefore be attributed to magma homogenization due to a density minimum, efficient segregation of basaltic magma from cumulates at ∼60 % fractional crystallization, or a combination of both of these processes. Finally, we note that it is possible that volatiles played some part in driving the basalt eruptions, as even small quantities of volatiles (<0·1 wt% H2O estimated for Winnipegosis parental melt; Waterton et al., 2017) might be concentrated in the remaining melt by extensive crystallization of anhydrous phases. However, the presence and role of volatiles is difficult to assess in altered ancient rocks, and is beyond the scope of this paper. Implications for other komatiite–basalt associations A fractional crystallization relationship between basalts and komatiites is a long-standing idea originating in the 1970s (Arndt et al., 1977; Francis & Hynes, 1979; Arndt & Nesbitt, 1982; Revillon et al., 2000; Shimizu et al., 2005; Mole et al., 2018; Ghosh et al., 2019). Evidence, presented here and in previous studies, that basaltic compositions can form through extensive crystallization of komatiitic magmas in upper crustal magma chambers contradicts the suggestion that komatiites represent a different geochemical ‘lineage’ to tholeiites (Arndt & Brooks, 1980). In the case of the WKB, as demonstrated here, the komatiites are the high-MgO extreme endmember of the tholeiitic lineage. Although this model of basalt formation may not necessarily apply to all komatiite-associated basalts, it may be relevant to other komatiite–tholeiite sequences, such as the d-basalts and G2 komatiites on the island of Gorgona (Revillon et al., 2000; Kerr, 2005). Gorgona d-basalts have similar REE patterns to the G2 komatiites, with the exception of minor LREE/MREE enrichment, and like the Winnipegosis Komatiite Belt, there is compelling evidence of crustal differentiation of ultramafic magmas in the form of abundant dunite and wehrlite bodies interpreted as cumulates (Revillon et al., 2000). Furthermore, the G2 komatiites and d-basalts have similar radiogenic lithophile and 187Os isotope characteristics (Kerr, 2005), which under this model can be understood as deriving from a shared parental melt rather than a similar depleted mantle source. Finally, similar to the Winnipegosis basalts, Gorgona d-basalts are also relatively Fe enriched, which is consistent with their derivation from a deep-sourced, high-degree parental melt. Similarly, this model may be applicable to other komatiite–basalt associations in which the basalts and komatiites have similar trace element characteristics (Rollinson, 1999; Manikyamba et al., 2008; Schneider et al., 2019) or isotopic compositions (Hamilton et al., 1979; Blichert-Toft & Arndt, 1999; Hollings & Wyman, 1999; Debaille et al., 2013). Furthermore, we suggest that fractional crystallization models for the formation of komatiite-associated basalts may in part explain the bimodality of Archaean mafic–ultramafic volcanism (Kamber & Tomlinson, 2019). If the variation in volcanic rock compositions observed at the surface is merely driven by different degrees of mantle melting at different temperatures, then a smooth continuum of compositions might be expected, due to the continuum of mantle temperatures between the hot plume axis and ambient mantle (e.g. Herzberg & Gazel, 2009). If, instead, the variation in composition is driven by fractional crystallization processes, then bimodal behaviour can be established depending on whether parental magmas are able to rapidly ascend to the surface or become stalled in the crust. Those that ascend without stalling erupt as komatiites, whereas those that meet a crustal density trap or an existing magma chamber crystallize until their eruption is triggered to form basalts. In this case, the observed abundance of basaltic compositions could in part be explained by the homogenization of magma chambers melts following plagioclase saturation, and efficient segregation of basaltic magma from cumulates at ∼60 % fractional crystallization. A potential objection to this model is that a fractional crystallization origin for komatiite-associated basalts requires the generation of large volumes of ultramafic cumulates in the Archaean crust. We note that in the case of the Proterozoic Winnipegosis Komatiite Belt, the largest ultramafic cumulate body is actually located stratigraphically below the supracrustal belt, in the underlying Archaean tonalite–trondhjemite–granodiorite (TTG). Similarly, in Archaean cratons where deep crustal sections are exposed, such as the North Atlantic Craton, ultramafic enclaves are a common feature in TTG gneisses (e.g. Chadwick & Crewe, 1986; Szilas et al., 2018). As Archaean crust is widely thought to have been internally mobile, it is also possible that ultramafic cumulates generated from komatiite crystallization delaminated into the underlying mantle (Bédard, 2006; Abbott et al., 2013; Sizova et al., 2015). CONCLUSIONS The Winnipegosis Komatiite Belt (WKB) contains abundant contemporaneous komatiite and basalt volcanism, along with evidence of a large upper crustal magma chamber containing abundant dunite cumulates. Dunites are highly magnesian with whole-rock compositions similar to komatiitic olivine. They crystallized from komatiitic liquids and represent olivine cumulates with minor chromite and <15 wt% intercumulus melt. Basalts show an extremely restricted compositional range and REE patterns that are similar in shape to Winnipegosis komatiites, albeit with higher absolute concentrations and slightly elevated LREE/MREE. MELTS modelling is used to investigate whether the Winnipegosis basalts could have formed from extensive crystallization of a komatiitic parental melt similar to the Winnipegosis komatiites. MELTS is able to successfully reproduce komatiite liquid lines of descent with the exception of FeO, which is overestimated due to errors in predicted olivine KD, and CaO, which is probably underestimated due to MELTS overestimating CaO/MgO in clinopyroxene. The proportion of chromite is overestimated as MELTS does not take into account Cr2O3 in olivine, but this should not affect predicted LLDs. Crystallization of komatiitic magma at higher pressures (up to 4 kbar) leads to a greater degree of Al2O3 enrichment in evolved melts, due to an increasing temperature and compositional gap between clinopyroxene and plagioclase saturation. Differences in fO2 can have a similar effect, with a 1 log unit decrease in fO2 producing roughly the same Al2O3 enrichment as a 0·5 kbar increase in pressure of crystallization. This trade-off means that there is no unique pressure–fO2 solution to form a basalt through extensive fractionation of a komatiitic parental melt. However, almost all aspects of the major element chemistry of Winnipegosis basalts can be matched by ∼60 % fractional crystallization of a komatiite parent, in a magma chamber at pressures of 1·5–2·5 kbar and an oxygen fugacity of QFM − 1 to QFM + 1. A REE partitioning model attached to the MELTS outputs demonstrates that basalt REE patterns can be explained by extensive crystallization of a komatiite parental magma alone and do not require additional crustal contamination. The geochemical evidence that Winnipegosis basalts formed from crystallization of a parental melt similar to the Winnipegosis komatiites is well supported by geological evidence, which suggests that there was extensive crystallization of komatiitic magma in the upper crust below the supracrustal rocks of the WKB. It is inferred that basalts in the WKB are not the product of melting in cooler portions of a plume (Campbell et al., 1989), but instead are the product of upper crustal crystallization of komatiitic magmas (Arndt et al., 1977). Where komatiitic parental melts were able to rapidly ascend to the surface, komatiite flows were formed. However, if the komatiitic melts stalled, they fractionated extensively to form basalts. Eruption of compositionally homogeneous basalts may have been promoted by a density minimum prior to the point of plagioclase saturation (Sparks & Huppert, 1984), and/or efficient segregation of basaltic melt from cumulates after 60 % crystallization (Dufek & Bachmann, 2010). This model of tholeiitic basalt formation may be more broadly applicable, and has the potential to resolve outstanding questions on the bimodal nature of Archaean mafic–ultramafic magmatism (Kamber & Tomlinson, 2019). SUPPLEMENTARY DATA Supplementary data are available at Journal of Petrology online. ACKNOWLEDGEMENTS Part of this work was conducted under the tenure of an NSERC Vanier Scholarship, awarded to Pedro Waterton at the University of Alberta. Nicholas Arndt, Jake Ciborowski, and Paulo Sossi are thanked for in-depth and constructive reviews. Larry Heaman, Bob Luth, Matt Steele-McInnes, and Richard Walker are thanked for their time and helpful feedback on a preliminary version of this paper that appeared in Pedro Waterton’s PhD thesis. Finally, Mark Ghiorso, Paul Asimow, Paula Antoshechkina, Guilherme Gualda, and Roger Nielsen are thanked for running an excellent MELTS workshop at Goldschmidt 2016 that provided the spark for modelling in this paper. FUNDING Analytical work was funded by the Canada Excellence Research Chairs Program to D. Graham Pearson. REFERENCES Abbey S. ( 1983 ). Studies in “Standard Samples” of Silicate Rocks and Minerals 1969–1982 . Geological Survey of Canada Paper 83 , 1 – 114 . Google Scholar OpenURL Placeholder Text WorldCat Abbott D. H. , Mooney W. D. , VanTongeren J. A. ( 2013 ). The character of the Moho and lower crust within Archean cratons and the tectonic implications . Tectonophysics 609 , 690 – 705 . Google Scholar Crossref Search ADS WorldCat Aigner-Torres M. , Blundy J. , Ulmer P. , Pettke T. ( 2007 ). Laser ablation ICPMS study of trace element partitioning between plagioclase and basaltic melts: an experimental approach . Contributions to Mineralogy and Petrology 153 , 647 – 667 . Google Scholar Crossref Search ADS WorldCat Aitken B. G. , Echeverría L. M. ( 1984 ). Petrology and Geochemistry of Komatiites and Tholeiites from Gorgona Island, Colombia . Contributions to Mineralogy and Petrology 86 , 94 – 105 . Google Scholar Crossref Search ADS WorldCat Albarede F. ( 1992 ). How deep do common basaltic magmas form and differentiate? Journal of Geophysical Research 97 , 10997 . Google Scholar Crossref Search ADS WorldCat Ansdell K. M. , Lucas S. B. , Connors K. , Stern R. A. ( 1995 ). Kisseynew metasedimentary gneiss belt, Trans-Hudson orogen (Canada): back-arc origin and collisional inversion . Geology 23 , 1039 – 1043 . Google Scholar Crossref Search ADS WorldCat Arndt N. , Brooks C. ( 1980 ). Komatiites . Geology 8 , 155 – 156 . Google Scholar Crossref Search ADS WorldCat Arndt N. , Lesher C. M. , Barnes S. J. ( 2008 ). Komatiite . Cambridge : Cambridge University Press . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Arndt N. T. , Nesbitt R. W. ( 1982 ). Geochemistry of Munro Township basalts. In: Arndt N. T. , Nisbet E. G. (eds) Komatiites . London : George Allen & Unwin , pp. 309 – 330 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Arndt N. T. , Nisbet E. G. ( 1982 ). What is a komatiite? In: Arndt N. T. , Nisbet E. G. (eds) Komatiites . London : George Allen & Unwin , pp. 19 – 27 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Arndt N. T. , Naldrett A. J. , Pyke D. R. ( 1977 ). Komatiitic and iron-rich tholeiitic lavas of Munro Township, Northeast Ontario . Journal of Petrology 18 , 319 – 369 . Google Scholar Crossref Search ADS WorldCat Asimow P. D. , Longhi J. ( 2004 ). The significance of multiple saturation points in the context of polybaric near-fractional melting . Journal of Petrology 45 , 2349 – 2367 . Google Scholar Crossref Search ADS WorldCat Asimow P. D. , Dixon J. E. , Langmuir C. H. ( 2004 ). A hydrous melting and fractionation model for mid-ocean ridge basalts: Application to the Mid-Atlantic Ridge near the Azores . Geochemistry, Geophysics, Geosystems 5 , Q01E16 . Google Scholar OpenURL Placeholder Text WorldCat Ballhaus C. , Berry R. F. , Green D. H. ( 1991 ). High pressure experimental calibration of the olivine–orthopyroxene–spinel oxygen geobarometer: implications for the oxidation state of the upper mantle . Contributions to Mineralogy and Petrology 107 , 27 – 40 . Google Scholar Crossref Search ADS WorldCat Balta J. B. , McSween H. Y. ( 2013 ). Application of the MELTS algorithm to martian compositions and implications for magma crystallization . Journal of Geophysical Research E: Planets 118 , 2502 – 2519 . Google Scholar Crossref Search ADS WorldCat Baragar W. R. A. , Scoates R. F. J. ( 1981 ). The Circum-Superior Belt: A Proterozoic plate margin? In: Kroner A. (ed.) Developments in Precambrian Geology , volume 4, pp. 297 – 330 . Amsterdam: Elsevier. Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Barnes S. J. ( 1998 ). Chromite in Komatiites, 1. Magmatic Controls on Crystallization and Composition . Journal of Petrology 39 , 1689 – 1720 . Google Scholar Crossref Search ADS WorldCat Beattie P. ( 1993 ). Olivine–melt and orthopyroxene–melt equilibria . Contributions to Mineralogy and Petrology 115 , 103 – 111 . Google Scholar Crossref Search ADS WorldCat Bédard J. H. ( 2006 ). A catalytic delamination-driven model for coupled genesis of Archaean crust and sub-continental lithospheric mantle . Geochimica et Cosmochimica Acta 70 , 1188 – 1214 . Google Scholar Crossref Search ADS WorldCat Bickle M. J. , Nisbet E. G. , Martin A. ( 1994 ). Archean greenstone belts are not oceanic crust . Journal of Geology 102 , 121 – 137 . Google Scholar Crossref Search ADS WorldCat Blichert-Toft J. , Arndt N. T. ( 1999 ). Hf isotope compositions of komatiites . Earth and Planetary Science Letters 171 , 439 – 451 . Google Scholar Crossref Search ADS WorldCat Bolikhovskaya S. V. , Vasil’eva M. O. , Koptev-Dvornikov E. V. ( 1995 ). Modelling of low-Ca pyroxene crystallisation in basic systems (New versions of geothermometers) . Geokhimia 12 , 1710–1727. Google Scholar OpenURL Placeholder Text WorldCat Burnham O. M. ( 2009 ). CAMIRO Project 97E-02, Thompson Nickel Belt: Final Report March 2002, revised and updated 2003. Winnipeg: Manitoba Geological Survey. Campbell I. H. ( 1985 ). The difference between oceanic and continental tholeiites: a fluid dynamic explanation . Contributions to Mineralogy and Petrology 91 , 37 – 43 . Google Scholar Crossref Search ADS WorldCat Campbell I. H. , Davies D. R. ( 2017 ). Raising the continental crust . Earth and Planetary Science Letters 460 , 112 – 122 . Google Scholar Crossref Search ADS WorldCat Campbell I. H. , Griffiths R. W. , Hill R. I. ( 1989 ). Melting in an Archaean mantle plume: heads it’s basalts, tails it’s komatiites . Nature 339 , 697 – 692 . Google Scholar Crossref Search ADS WorldCat Chadwick B. , Crewe M. A. ( 1986 ). Chromite in the early Archean Akilia association (ca. 3,800 m.y.), Ivisartoq region, inner Godthabsfjord, southern west Greenland . Economic Geology 81 , 184 – 191 . Google Scholar Crossref Search ADS WorldCat Ciborowski T. J. R. , Minifie M. J. , Kerr A. C. , Ernst R. E. , Baragar B. , Millar I. L. ( 2017 ). A mantle plume origin for the Palaeoproterozoic Circum-Superior Large Igneous Province . Precambrian Research 294 , 189 – 213 . Google Scholar Crossref Search ADS WorldCat Corrigan D. ( 2012 ). Paleoproterozoic crustal evolution and tectonic process: insights from the Lithoprobe program in the Trans Hudson Orogen, Canada. In: Percival, J. A., Cook, F. A. & Clowes, R. M. (eds) Tectonic Styles in Canada: The Lithoprobe Perspective. Geological Association of Canada, Special Paper 49, 237 – 284 . Corrigan D. , Pehrsson S. , Wodicka N. , de Kemp E. ( 2009 ). The Palaeoproterozoic Trans-Hudson Orogen: a prototype of modern accretionary processes. In: Murphy J. B. , Keppie J. D. , Hynes A. J. (eds) Ancient Orogens and Modern Analogues . Geological Society, London , Special Publications 327 , 457 – 479 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Danyushevsky L. V. ( 2001 ). The effect of small amounts of H2O on crystallisation of mid-ocean ridge and backarc basin magmas . Journal of Volcanology and Geothermal Research 110 , 265 – 280 . Google Scholar Crossref Search ADS WorldCat Danyushevsky L. V. , Plechov P. ( 2011 ). Petrolog3: integrated software for modeling crystallization processes . Geochemistry, Geophysics, Geosystems 12 , Q07021 . Google Scholar OpenURL Placeholder Text WorldCat Debaille V. , O’Neill C. , Brandon A. D. , Haenecour P. , Yin Q.-Z. , Mattielli N. , Treiman A. H. ( 2013 ). Stagnant-lid tectonics in early Earth revealed by 142Nd variations in late Archean rocks . Earth and Planetary Science Letters 373 , 83 – 92 . Google Scholar Crossref Search ADS WorldCat DePaolo D. J. ( 1981 ). Trace element and isotope effects of combined wallrock assimilation and fractional crystallisation . Earth and Planetary Science Letters 53 , 189 – 202 . Google Scholar Crossref Search ADS WorldCat Dixon J. E. , Leist L. , Langmuir C. , Schilling J.-G. ( 2002 ). Recycled dehydrated lithosphere observed in plume-influenced mid-ocean-ridge basalt . Nature 420 , 385 – 389 . Google Scholar Crossref Search ADS PubMed WorldCat Droop G. T. R. ( 1987 ). A general equation for estimating Fe3+ concentrations in ferromagnesian silicates and oxides from microprobe analyses, using stoichiometric criteria . Mineralogical Magazine 51 , 431 – 435 . Google Scholar Crossref Search ADS WorldCat Dufek J. , Bachmann O. ( 2010 ). Quantum magmatism: Magmatic compositional gaps generated by melt-crystal dynamics . Geology 38 , 687 – 690 . Google Scholar Crossref Search ADS WorldCat Echeverria L. M. ( 1980 ). Tertiary or Mesozoic komatiites from Gorgona Island, Colombia: field relations and geochemistry . Contributions to Mineralogy and Petrology 73 , 253 – 266 . Google Scholar Crossref Search ADS WorldCat Farmer G. L. ( 2007 ). Continental Basaltic Rocks. In: Holland, H. D. & Turekian, K. K. (eds) Treatise on Geochemistry . Amsterdam : The Mantle and Core , pp. 1 – 39 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Faure F. , Tissandier L. ( 2014 ). Contrasted liquid lines of descent revealed by olivine-hosted melt inclusions and the external magma . Journal of Petrology 55 , 1779 – 1798 . Google Scholar Crossref Search ADS WorldCat Ford C. E. , Russell D. G. , Craven J. A. , Fisk M. R. ( 1983 ). Olivine–liquid equilibria: temperature, pressure and composition dependence of the crystal/liquid cation partition coefficients for Mg, Fe2+, Ca and Mn . Journal of Petrology 24 , 256 – 266 . Google Scholar Crossref Search ADS WorldCat Francis D. ( 1985 ). The Baffin Bay lavas and the value of picrites as analogues of primary magmas . Contributions to Mineralogy and Petrology 89 , 144 – 154 . Google Scholar Crossref Search ADS WorldCat Francis D. M. , Hynes A. J. ( 1979 ). Komatiite-derived tholeiites in the Proterozoic of New Quebec . Earth and Planetary Science Letters 44 , 473 – 481 . Google Scholar Crossref Search ADS WorldCat Gale A. , Dalton C. A. , Langmuir C. H. , Su Y. , Schilling J. G. ( 2013 ). The mean composition of ocean ridge basalts . Geochemistry, Geophysics, Geosystems 14 , 489 – 518 . Google Scholar Crossref Search ADS WorldCat Ghiorso M. S. ( 1997 ). Thermodynamic models of igneous processes . Annual Review of Earth and Planetary Sciences 25 , 221 – 241 . Google Scholar Crossref Search ADS WorldCat Ghiorso M. S. , Gualda G. A. R. ( 2015 ). An H2O–CO2 mixed fluid saturation model compatible with rhyolite-MELTS . Contributions to Mineralogy and Petrology 169 , 1 – 30 . Google Scholar Crossref Search ADS WorldCat Ghosh R. , Vermeesch P. , Gain D. , Mondal R. ( 2019 ). Genetic relationship among komatiites and associated basalts in the Badampahar greenstone belt (3·25–3·10 Ga), Singhbhum Craton, Eastern India . Precambrian Research 327 , 196 – 211 . Google Scholar Crossref Search ADS WorldCat Govindaraju K. ( 1994 ). 1994 compilation of working values and sample description for 383 geostandards . Geostandards Newsletter 18 , 1 – 158 . Google Scholar Crossref Search ADS WorldCat Grove T. L. , Baker M. B. ( 1983 ). Effects of melt density on magma mixing in calc-alkaline series lavas . Nature 305 , 416 – 418 . Google Scholar Crossref Search ADS WorldCat Grove T. L. , Gerlach D. C. , Sando T. W. ( 1982 ). Origin of calc-alkaline series lavas at Medicine Lake Volcano by fractionation, assimilation and mixing . Contributions to Mineralogy and Petrology 80 , 160 – 182 . Google Scholar Crossref Search ADS WorldCat Gualda G. A. R. , Ghiorso M. S. , Lemons R. V. , Carley T. L. ( 2012 ). Rhyolite-MELTS: A modified calibration of MELTS optimized for silica-rich, fluid-bearing magmatic systems . Journal of Petrology 53 , 875 – 890 . Google Scholar Crossref Search ADS WorldCat Hamilton P. J. , Evensen N. M. , O’Nions R. K. , Smith H. S. , Erlank A. J. ( 1979 ). Sm–Nd dating of Onverwacht Group Volcanics, southern Africa . Nature 279 , 298 – 300 . Google Scholar Crossref Search ADS WorldCat Heaman L. M. , Peck D. , Toope K. ( 2009 ). Timing and geochemistry of 1·88 Ga Molson Igneous Events, Manitoba: insights into the formation of a craton-scale magmatic and metallogenic province . Precambrian Research 172 , 143 – 162 . Google Scholar Crossref Search ADS WorldCat Herzberg C. T. , Asimow P. D. ( 2015 ). PRIMELT3 MEGA.XLSM software for primary magma calculation: Peridotite primary magma MgO contents from the liquidus to the solidus . Geochemistry, Geophysics, Geosystems 16 , 563 – 578 . Google Scholar Crossref Search ADS WorldCat Herzberg C. , Gazel E. ( 2009 ). Petrological evidence for secular cooling in mantle plumes . Nature 458 , 619 – 622 . Google Scholar Crossref Search ADS PubMed WorldCat Herzberg C. , O’Hara M. J. ( 2002 ). Plume-associated ultramafic magmas of Phanerozoic age . Journal of Petrology 43 , 1857 – 1883 . Google Scholar Crossref Search ADS WorldCat Herzberg C. , Condie K. , Korenaga J. ( 2010 ). Thermal history of the Earth and its petrological expression . Earth and Planetary Science Letters 292 , 79 – 88 . Google Scholar Crossref Search ADS WorldCat Hirschmann M. M. , Ghiorso M. S. , Wasylenki L. E. , Asimow P. D. , Stolper E. M. ( 1998 ). Calculation of peridotite partial melting from thermodynamic models of minerals and melts. I. Review of methods and comparison with experiments . Journal of Petrology 39 , 1091 – 1115 . Google Scholar Crossref Search ADS WorldCat Hofmann A. W. ( 1988 ). Chemical differentiation of the Earth: the relationship between mantle, continental crust, and oceanic crust . Earth and Planetary Science Letters 90 , 297 – 314 . Google Scholar Crossref Search ADS WorldCat Hollings P. , Wyman D. A. ( 1999 ). Trace element and Sm–Nd systematics of volcanic and intrusive rocks from the 3 Ga Lumby Lake Greenstone belt, Superior Province: evidence for Archean plume–arc interaction . Lithos 46 , 189 – 213 . Google Scholar Crossref Search ADS WorldCat Hosain I. T. , Bamburak J. D. ( 2002 ). Interpretation of Thompson-type geophysical signatures (NTS 63B and 63F) and the possible cause of the Camperville gravity low (NTS 62N and 63C), West–central Manitoba. Report of Activities 2002. Manitoba Industry, Trade and Mines. Winnipeg: Manitoba Geological Survey, pp. 117 – 130 . Hulbert L. , Stern R. , Kyser T. K. , Pearson J. , Lesher M. , Grinenko L. ( 1994 ). The Winnipegosis komatiite belt, central Manitoba. Manitoba Mining and Minerals Convention 1994, Program and Abstracts, 21. IAG ( 2015 a). OKUM certificate of analysis. http://iageo.com/certified-reference-material/. Nottingham, UK: International Association of Geoanalysts. IAG ( 2015 b). MUH-1 certificate of analysis. http://iageo.com/certified-reference-material/. Nottingham, UK: International Association of Geoanalysts. JCGM ( 2008 ). International vocabulary of metrology—Basic and general concepts and associated terms (VIM) Vocabulaire international de métrologie—Concepts fondamentaux et généraux et termes associés (VIM) . Geneva : International Organization for Standardization , 3 , 104 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jochum K. P. , Nohl U. , Herwig K. , Lammel E. , Stoll B. , Hofmann A. W. ( 2005 ). GeoReM: a new geochemical database for reference materials and isotopic standards . Geostandards and Geoanalytical Research 29 , 333 – 338 . Google Scholar Crossref Search ADS WorldCat Jochum K. P. , Weis U. , Schwager B. , Stoll B. , Wilson S. A. , Haug G. H. , Andreae M. O. , Enzweiler J. ( 2016 ). Reference Values Following ISO Guidelines for Frequently Requested Rock Reference Materials . Geostandards and Geoanalytical Research 40 , 333 – 350 . Google Scholar Crossref Search ADS WorldCat Jolly W. T. ( 1982 ). Progressive metamorphism of komatiites and related Archaean lavas of the abitibi area, Canada. In: Arndt N. T. , Nisbet E. G. (eds) Komatiites . London : George Allen & Unwin , pp. 247 – 266 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Kamber B. S. , Tomlinson E. L. ( 2019 ). Petrological, mineralogical and geochemical peculiarities of Archaean cratons . Chemical Geology 511 , 123 – 151 . Google Scholar Crossref Search ADS WorldCat Kamenetsky V. ( 1996 ). Methodology for the study of melt inclusions in Cr-spinel, and implications for parental melts of MORB from FAMOUS area . Earth and Planetary Science Letters 142 , 479 – 486 . Google Scholar Crossref Search ADS WorldCat Keller C. B. , Schoene B. ( 2012 ). Statistical geochemistry reveals disruption in secular lithospheric evolution about 2·5 Gyr ago . Nature 485 , 490 – 493 . Google Scholar Crossref Search ADS PubMed WorldCat Kerr A. C. ( 2005 ). La Isla de Gorgona, Colombia: a petrological enigma? Lithos 84 , 77 – 101 . Google Scholar Crossref Search ADS WorldCat Kinzler R. J. , Grove T. L. ( 1985 ). Crystallization and differentiation of Archean komatiite lavas from northeast Ontario: phase equilibrium and kinetic studies . American Mineralogist 70 , 40 – 51 . Google Scholar OpenURL Placeholder Text WorldCat Kinzler R. J. , Grove T. L. ( 1992 ). Primary magmas of mid-ocean ridge basalts 2 . Journal of Geophysical Research 97 , 6885 – 6906 . Google Scholar Crossref Search ADS WorldCat Klein E. M. , Langmuir C. H. ( 1987 ). Global correlation of ocean ridge basalt chemistry with axial depth and crustal thickness . Journal of Geophysical Research 92 , 8089 – 8115 . Google Scholar Crossref Search ADS WorldCat Knaack C. , Cornelius S. , Hooper P. ( 1994 ). Trace Element Analyses of Rocks and Minerals by ICP-MS. Department of Geology, Washington State University, Pullman, Open-File Report. Lange R. A. ( 2002 ). Constraints on the preeruptive volatile concentrations in the Columbia River flood basalts . Geology 30 , 179 – 182 . Google Scholar Crossref Search ADS WorldCat Larsen L. M. , Pedersen A. K. ( 2000 ). Processes in high-Mg, High-T magmas: evidence from olivine, chromite and glass in Palaeogene picrites from West Greenland . Journal of Petrology 41 , 1071 – 1098 . Google Scholar Crossref Search ADS WorldCat Laubier M. , Schiano P. , Doucelance R. , Ottolini L. , Laporte D. ( 2007 ). Olivine-hosted melt inclusions and melting processes beneath the FAMOUS zone (Mid-Atlantic Ridge) . Chemical Geology 240 , 129 – 150 . Google Scholar Crossref Search ADS WorldCat Lee C. T. A. , Luffi P. , Plank T. , Dalton H. , Leeman W. P. ( 2009 ). Constraints on the depths and temperatures of basaltic magma generation on Earth and other terrestrial planets using new thermobarometers for mafic magmas . Earth and Planetary Science Letters 279 , 20 – 33 . Google Scholar Crossref Search ADS WorldCat Lightfoot P. C. , Hawkesworth C. J. , Devey C. W. , Rogers N. W. , Calsteren P. W. C. V. ( 1990 ). Source and differentiation of Deccan trap lavas: Implications of geochemical and mineral chemical variations . Journal of Petrology 31 , 1165 – 1200 . Google Scholar Crossref Search ADS WorldCat Longhi J. ( 1981 ). Multicomponent phase diagrams and the phase equilibria of basalts. Workshop on Magmatic Processes of Early Planetary Crusts. Houston: Lunar and Planetary Institute, volume 82-01, pp. 90 – 94 . Lucas S. B. , White D. J. , Bleeker W. , Hajnal Z. , Lewry J. F. , Weber W. ( 1996 ). Crustal structure of the Superior Boundary Zone (Thompson Belt) from new Lithoprobe seismic reflection data. University of British Columbia, Trans-Hudson Orogen transect . Lithoprobe Report 55 , 82 – 94 . Google Scholar OpenURL Placeholder Text WorldCat Manikyamba C. , Kerrich R. , Khanna T. C. , Keshav Krishna A. , Satyanarayanan M. ( 2008 ). Geochemical systematics of komatiite–tholeiite and adakitic-arc basalt associations: The role of a mantle plume and convergent margin in formation of the Sandur Superterrane, Dharwar craton, India . Lithos 106 , 155 – 172 . Google Scholar Crossref Search ADS WorldCat Matzen A. K. , Baker M. B. , Beckett J. R. , Stolper E. M. ( 2011 ). Fe–Mg partitioning between olivine and high-magnesian melts and the nature of Hawaiian parental liquids . Journal of Petrology 52 , 1243 – 1263 . Google Scholar Crossref Search ADS WorldCat McGregor C. R. ( 2011 ). Open File OF2011-1: GIS Compilation of Relogged Sub-Phanerozoic Precambrian Exploration Drillcore from the Thompson Nickel Belt, eastern Flin Flon Belt and Winnipegosis Komatiite Belt (parts of NTS 63B, C, F, G, J, K). Winnipeg: Manitoba Geological Survey. McKay G. , Wagstaff J. , Yang S. R. ( 1986 ). Clinopyroxene REE distribution coefficients for shergottites: The REE content of the Shergotty melt . Geochimica et Cosmochimica Acta 50 , 927 – 937 . Google Scholar Crossref Search ADS WorldCat Meisel T. , Schoner N. , Paliulionyte V. , Kahr E. ( 2002 ). Determination of rare earth elements, Y, Th, Zr, Hf, Nb and Ta in geological reference materials G-2, G-3, SCo-1 and WGB-1 by sodium peroxide sintering and inductively coupled plasma-mass spectrometry . Geostandards and Geoanalytical Research 26 , 53 – 61 . Google Scholar Crossref Search ADS WorldCat Mertzman S. ( 2000 ). K–Ar results from the southern Oregon–northern California Cascade Range . Oregon Geology 62 , 99 – 122 . Google Scholar OpenURL Placeholder Text WorldCat Minifie M. J. , Kerr A. C. , Ernst R. E. , Hastie A. R. , Ciborowski T. J. R. , Desharnais G. , Millar I. L. ( 2013 ). The northern and southern sections of the western ca. 1880 Ma Circum-Superior Large Igneous Province, North America: The Pickle Crow dyke connection? Lithos 174 , 217 – 235 . Google Scholar Crossref Search ADS WorldCat Mole D. R. , Barnes S. J. , Yao Z. , White A. J. R. , Maas R. , Kirkland C. L. ( 2018 ). The Archean Fortescue large igneous province: A result of komatiite contamination by a distinct Eo-Paleoarchean crust . Precambrian Research 310 , 365 – 390 . Google Scholar Crossref Search ADS WorldCat Morimoto N. , Fabries J. , Ferguson A. K. , Ginzburg I. V. , Ross M. , Seifert F. A. , Zussman J. , Aoki K. , Gottardi G. ( 1988 ). Nomenclature of pyroxenes . Mineralogical Magazine 52 , 535 – 550 . Google Scholar Crossref Search ADS WorldCat Murck B. W. , Campbell I. H. ( 1986 ). The effects of temperature, oxygen fugacity and melt composition on the behaviour of chromium in basic and ultrabasic melts . Geochimica et Cosmochimica Acta 50 , 1871 – 1887 . Google Scholar Crossref Search ADS WorldCat Nagasawa H. , Schreiber H. D. , Morris R. V. ( 1980 ). Experimental mineral/liquid partition coefficients of the rare earth elements (REE), Sc and Sr for perovskite, spinel and melilite . Earth and Planetary Science Letters 46 , 431 – 437 . Google Scholar Crossref Search ADS WorldCat Nebel O. , Campbell I. H. , Sossi P. A. , Van Kranendonk M. J. ( 2014 ). Hafnium and iron isotopes in early Archean komatiites record a plume-driven convection cycle in the Hadean Earth . Earth and Planetary Science Letters 397 , 111 – 120 . Google Scholar Crossref Search ADS WorldCat Nesbitt R. W. , Sun S. ( 1976 ). Geochemistry of Archaean spinifex-textured peridotites and magnesian and low-magnesian tholeiites . Earth and Planetary Science Letters 31 , 433 – 453 . Google Scholar Crossref Search ADS WorldCat Nicklas R. W. , Puchtel I. S. , Ash R. D. , Piccoli P. M. , Hanski E. , Nisbet E. G. , Waterton P. , Graham D. G. , Anbar A. D. ( 2019 ). Secular mantle oxidation across the Archean–Proterozoic boundary: Evidence from V partitioning in komatiites and picrites . Geochimica et Cosmochimica Acta 250 , 49 – 75 . Google Scholar Crossref Search ADS WorldCat O’Hara M. J. ( 1968 ). The bearing of phase equilibria studies in synthetic and natural systems on the origin and evolution of basic and ultrabasic rocks . Earth-Science Reviews 4 , 69 – 133 . Google Scholar Crossref Search ADS WorldCat Parman S. W. , Dann J. C. , Grove T. L. , de Wit M. J. ( 1997 ). Emplacement conditions of komatiite magmas from the 3·49 Ga Komati Formation, Barberton Greenstone Belt, South Africa . Earth and Planetary Science Letters 150 , 303 – 323 . Google Scholar Crossref Search ADS WorldCat Pearce J. A. ( 2008 ). Geochemical fingerprinting of oceanic basalts with applications to ophiolite classification and the search for Archean oceanic crust . Lithos 100 , 14 – 48 . Google Scholar Crossref Search ADS WorldCat Pearce T. H. ( 1968 ). A contribution to the theory of variation diagrams . Contributions to Mineralogy and Petrology 19 , 142 – 157 . Google Scholar Crossref Search ADS WorldCat Pearson D. G. , Irvine G. J. , Ionov D. A. , Boyd F. R. , Dreibus G. E. ( 2004 ). Re–Os isotope systematics and platinum group element fractionation during mantle melt extraction: a study of massif and xenolith peridotite suites . Chemical Geology 208 , 29 – 59 . Google Scholar Crossref Search ADS WorldCat Prægel N. O. , Holm P. M. ( 2001 ). Replenishment episodes and crustal assimilation in the development of an early Tertiary magma chamber, East Greenland: Evidence from layered cumulates of the Kælvegletscher ultramafic complex, Kangerlussuaq . Mineralogy and Petrology 73 , 279 – 304 . Google Scholar Crossref Search ADS WorldCat Puchtel I. S. , Touboul M. , Blichert-Toft J. , Walker R. J. , Brandon A. D. , Nicklas R. W. , Kulikov V. S. , Samsonov A. V. ( 2016 ). Lithophile and siderophile element systematics of Earth’s mantle at the Archean–Proterozoic boundary: evidence from 2·4 Ga komatiites . Geochimica et Cosmochimica Acta 180 , 227 – 255 . Google Scholar Crossref Search ADS WorldCat Puchtel I. S. , Blichert-Toft J. , Touboul M. , Walker R. J. ( 2018 ). 182W and HSE constraints from 2·7 Ga komatiites on the heterogeneous nature of the Archean mantle . Geochimica et Cosmochimica Acta 228 , 1 – 26 . Google Scholar Crossref Search ADS WorldCat Putirka K. D. ( 2005 ). Mantle potential temperatures at Hawaii, Iceland, and the mid-ocean ridge system, as inferred from olivine phenocrysts: Evidence for thermally driven mantle plumes . Geochemistry, Geophysics, Geosystems 6 , Q05L08 . Google Scholar OpenURL Placeholder Text WorldCat Revillon S. , Arndt N. T. , Chauvel C. , Hallot E. ( 2000 ). Geochemical study of ultramafic volcanic and plutonic rocks from Gorgona Island, Colombia: the plumbing system of an oceanic plateau . Journal of Petrology 41 , 1127 – 1153 . Google Scholar Crossref Search ADS WorldCat Rollinson H. ( 1999 ). Petrology and geochemistry of metamorphosed komatiites and basalts from the Sula Mountains greenstone belt, Sierra Leone . Contributions to Mineralogy and Petrology 134 , 86 – 101 . Google Scholar Crossref Search ADS WorldCat Russell J. K. , Nicholls J. , Stanley C. R. , Pearce T. H. ( 1990 ). Pearce element ratios: A paradigm for testing hypotheses . EOS Transactions American Geophysical Union 71 , 234 – 247 . Google Scholar Crossref Search ADS WorldCat Sarbas B. , Nohl U. ( 2008 ). The GEOROC database as part of a growing geoinformatics network. In: Brady, S., Sinha, A. K. & Gundersen, L. C. (eds) Geoinformatics 2008—Data to Knowledge, Proceedings. VA, USA: USGS, pp. 42 – 43 . Schneider K. P. , Hoffmann J. E. , Münker C. , Patyniak M. , Sprung P. , Roerdink D. , Garbe-Schönberg D. , Kröner A. ( 2019 ). Petrogenetic evolution of metabasalts and metakomatiites of the lower Onverwacht Group, Barberton Greenstone Belt (South Africa ). Chemical Geology 511 , 152 – 177 . Google Scholar Crossref Search ADS WorldCat Sen Gupta J. G. ( 1994 ). Characterisation of major, minor and trace elements in six CCRMP (TDB-1, WGB-1, UMT-1, WPR-1 and WMS-1) and one IWG (ZW-C) geochemical candidate reference materials . Geostandards and Geoanalytical Research 18 , 111 – 122 . Google Scholar Crossref Search ADS WorldCat Shimizu K. , Nakamura E. , Maruyama S. ( 2005 ). The Geochemistry of Ultramafic to Mafic Volcanics from the Belingwe Greenstone Belt, Zimbabwe: Magmatism in an Archean Continental Large Igneous Province . Journal of Petrology 46 , 2367 – 2394 . Google Scholar Crossref Search ADS WorldCat Sizova E. , Gerya T. , Stüwe K. , Brown M. ( 2015 ). Generation of felsic crust in the Archean: a geodynamic modeling perspective . Precambrian Research 271 , 198 – 224 . Google Scholar Crossref Search ADS WorldCat Sobolev A. V. ( 1996 ). Melt inclusions in minerals as a source of principle petrological information . Petrology 4 , 209 – 220 . Google Scholar OpenURL Placeholder Text WorldCat Sobolev A. V. , Chaussidon M. ( 1996 ). H2O concentrations in primary melts from supra-subduction zones and mid-ocean ridges: Implications for H2O storage and recycling in the mantle . Earth and Planetary Science Letters 137 , 45 – 55 . Google Scholar Crossref Search ADS WorldCat Sparks R. S. J. , Huppert H. E. ( 1984 ). Density changes during the fractional crystallization of basaltic magmas: fluid dynamic implications . Contributions to Mineralogy and Petrology 85 , 300 – 309 . Google Scholar Crossref Search ADS WorldCat Stamatelopoulou-Seymour K. , Francis D. , Ludden J. ( 1983 ). The Petrogenesis of the Lac Guyer Komatiites and Basalts and the Nature of the Komatiite–Komatiitic Basalt Compositional Gap . Contributions to Mineralogy and Petrology 84 , 6 – 14 . Google Scholar Crossref Search ADS WorldCat Storey M. , Mahoney J. J. , Kroenke L. W. , Saunders A. D. ( 1991 ). Are oceanic plateaus sites of komatiites formation? Geology 19 , 376 – 379 . Google Scholar Crossref Search ADS WorldCat Sun C. , Liang Y. ( 2013 ). The importance of crystal chemistry on REE partitioning between mantle minerals (garnet, clinopyroxene, orthopyroxene, and olivine) and basaltic melts . Chemical Geology 358 , 23 – 36 . Google Scholar Crossref Search ADS WorldCat Sun S.-S. , Nesbitt R. W. ( 1978 ). Petrogenesis of Archaean ultrabasic and basic volcanics: Evidence from rare earth elements . Contributions to Mineralogy and Petrology 65 , 301 – 325 . Google Scholar Crossref Search ADS WorldCat Szilas K. , van Hinsberg V. , McDonald I. , Næraa T. , Rollinson H. , Adetunji J. , Bird D. ( 2018 ). Highly refractory Archaean peridotite cumulates: petrology and geochemistry of the Seqi Ultramafic Complex, SW Greenland . Geoscience Frontiers 9 , 689 – 714 . Google Scholar Crossref Search ADS WorldCat Tappe S. , Pearson D. G. , Nowell G. , Nielsen T. , Milstead P. , Muehlenbachs K. ( 2011 ). A fresh isotopic look at Greenland kimberlites: Cratonic mantle lithosphere imprint on deep source signal . Earth and Planetary Science Letters 305 , 235 – 248 . Google Scholar Crossref Search ADS WorldCat Taylor H. P. ( 1980 ). The effects of assimilation of country rocks by magmas on 18O/16O and 87Sr/86Sr systematics in igneous rocks . Earth and Planetary Science Letters 47 , 243 – 254 . Google Scholar Crossref Search ADS WorldCat Thomson A. , Maclennan J. ( 2013 ). The distribution of olivine compositions in Icelandic basalts and picrites . Journal of Petrology 54 , 745 – 768 . Google Scholar Crossref Search ADS WorldCat Viljoen M. J. , Viljoen R. P. ( 1969 a). The geology and geochemistry of the lower ultramafic unit of the Onverwacht group and a proposed new class of igneous rocks . Geological Survey of South Africa, Special Publication 21 , 55 – 85 . Google Scholar OpenURL Placeholder Text WorldCat Viljoen M. J. , Viljoen R. P. ( 1969 b). Evidence for the existence of a mobile extrusive peridotitic magma from the Komati Formation of the Onvernacht Group . Geological Survey of South Africa, Special Publication 21 , 87 – 112 . Google Scholar OpenURL Placeholder Text WorldCat Waterton P. ( 2018 ). The 1.9 Ga Winnipegosis Komatiite: Implications for Earth Accretion, Mantle Dynamics, and Komatiite Formation. PhD thesis, University of Alberta. Waterton P. , Pearson D. G. , Kjarsgaard B. , Hulbert L. , Locock A. , Parman S. W. , Davis B. ( 2017 ). Age, Origin, and Thermal Evolution of the ultra-fresh ∼1.9 Ga Winnipegosis Komatiites, Manitoba, Canada . Lithos 268–271 , 114 – 130 . Google Scholar Crossref Search ADS WorldCat White R. S. , Bown J. W. , Smallwood J. R. ( 1995 ). The temperature of the Iceland plume and origin of outward- propagating V-shaped ridges . Journal of the Geological Society, London 152 , 1039 – 1045 . Google Scholar Crossref Search ADS WorldCat Wilson S. A. ( 1998 ). United States Geological Survey Preliminary Certificate of Analysis. Basalt, Hawaiian Volcanic Observatory, BHVO-2 . Denver, CO : US Geological Survey . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Wooden J. L. , Czamanske G. K. , Fedorenko V. A. , Arndt N. T. , Chauvel C. , Bouse R. M. , King B. S. W. , Knight R. J. , Siems D. F. ( 1993 ). Isotopic and trace-element constraints on mantle and crustal contributions to Siberian continental flood basalts, Noril’sk area, Siberia . Geochimica et Cosmochimica Acta 57 , 3677 – 3704 . Google Scholar Crossref Search ADS WorldCat © The Author(s) 2020. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - A Fractional Crystallization Link between Komatiites, Basalts, and Dunites of the Palaeoproterozoic Winnipegosis Komatiite Belt, Manitoba, Canada
JF - Journal of Petrology
DO - 10.1093/petrology/egaa052
DA - 2020-11-08
UR - https://www.deepdyve.com/lp/oxford-university-press/a-fractional-crystallization-link-between-komatiites-basalts-and-A16eC7PHBy
VL - 61
IS - 5
DP - DeepDyve
ER -