| Journal of Petrology | Pages |
© 1999 Oxford University Press |
|
|
The origin of the distinctive geochemical signature that characterizes arc magmas is a contentious issue. Geochemical and isotopic data imply that the depleted peridotites of the supra-subduction mantle wedge are refertilized by the influx of a slab-derived component (SZ component), which is either an aqueous fluid or a hydrous silicate melt enriched in incompatible trace elements such as Rb, Ba, Cs, Th, U, Pb, Sr and the light rare earth elements (LREE) (Hawkesworth et al., 1993; Pearce & Parkinson, 1993; Arculus, 1994; Iwamori, 1998). However, considerable uncertainty remains concerning the extent of prior wedge depletion, and with regard to the composition and source of this slab-derived SZ component. Boninites are thought to originate as low-pressure partial melts of extremely depleted mantle wedges (Cameron et al., 1979; Coish et al., 1982; Hickey & Frey, 1982; Kostopoulos & Murton, 1992; Sobolev & Danyushevsky, 1994; Taylor et al., 1994). As such, the trace element signatures of the SZ component should stand out more clearly in boninites than in normal arc magmas. This paper presents field, petrographic and whole-rock geochemical data for a suite of boninites from the Ordovician Betts Cove ophiolite of Newfoundland, Canada (Fig. 1). The composition of the mantle sources of these boninites is calculated, allowing the degree of prior wedge depletion and the composition of the added SZ component to be modelled, so addressing some of the controversies related to arc magma genesis. Figure
The Betts Cove ophiolite (Fig. 1) is inferred to have formed by seafloor spreading (Fig. 2) near the margin of North America (Harris, 1992; Pinet & Tremblay, 1995; Bédard et al., 1998) in the Ordovician (488·6 + 3·1/-1·8 Ma, Dunning & Krogh, 1985). The ophiolitic ocean crust and the sub-conformably overlying lavas and sediments of the Snooks Arm Group were accreted on to the continental margin of eastern North America during the Taconian Orogeny (~470 Ma, Williams, 1979; Hibbard, 1983; Dallmeyer & Hibbard, 1984; Tremblay et al., 1997). The ophiolitic and cover rocks at Betts Cove were then tilted and eroded before deposition of the Cape St John Group in the Silurian (425 Ma, Coyle, 1990). Coeval Silurian granitoids intrude the Betts Cove ophiolite along its northwestern margin. The ophiolite and its cover rocks were then folded into a large-scale syncline (Fig. 1), presumably during Acadian compression in the Devonian (Tremblay et al., 1997). Figure
A new geological map of the Betts Cove ophiolite complex is now available (Bédard et al., 1999a), and descriptions of field relations and petrographic characteristics of the different units have been given by Bédard et al., (2000). All sample locations, the complete structural and geochemical database, and a digital version of the map have been given by Bédard et al., (1999b). The Betts Cove ophiolite complex (Fig. 1) is separated from its country rocks by a marginal band of serpentinite and talc schists. Massive talc-magnesite-ankerite-magnetite schists may represent mantle rocks. The inner part of the ophiolite complex comprises a basal sequence of layered cumulate peridotites, pyroxenites and gabbro-norites (Upadhyay, 1973; Bédard et al., 1999a, 2000). The sequence of crystallization recorded in the cumulates is chromite, olivine, orthopyroxene, clinopyroxene, plagioclase, Fe-Ti oxide, hornblende and quartz. Quartz- and Fe-Ti oxide-bearing gabbro-norites commonly intrude the cumulate-sheeted dyke interface (see
Church & Riccio, 1974), and envelope slivers of boninitic pillow lavas and sheeted dyke septa, yet are themselves cut by dykes of the sheeted dyke complex. The sheeted dyke complex is composed almost entirely of rocks belonging to the boninitic suite, and grades up into a sequence of boninitic lavas locally >1 km in thickness, the Betts Head Formation (Bédard et al., 1999a, 2000). This paper focuses on the boninitic sheeted dykes and lavas. Rocks of the Snooks Arm Group sub-conformably overlie the Betts Head boninites. The lowermost unit of the Snooks Arm Group is the Mount Misery Formation, consisting of ~1 km of submarine basalts with arc tholeiitic composition. These are followed by (Fig. 1) interstratified evolved tholeiitic basalts (Upper Snooks Arm Group tholeiites; Bédard et al., 2000), calc-alkaline tuffs and lavas (Cousineau & Bédard, 2000), and volcanogenic sedimentary rocks (Kessler & Bédard, 2000). The petrogenesis and tectonic implications of Snooks Arm Group magmas will be considered in a subsequent paper. Several deformation episodes overprint this area, but most of the strain was accommodated in the marginal serpentinite and talc schists, and so the intrusive, extrusive and sedimentary rocks constituting the core of the ophiolite rarely exhibit pervasive deformation fabrics. The ophiolite stratigraphy is largely intact, although tilted to the vertical and folded, and most of the faults and shear zones are thought to be syn-oceanic structures. The sheeted dyke unit thins and eventually disappears completely north of Betts Cove (Fig. 1), possibly as the result of excision by spreading-related normal faults and décollements. The basal and upper contacts of the sheeted dyke unit are gradational, faulted, or have been intruded by gabbro-norites. In many places the sheeted dykes grade up into a 50-200 m wide zone characterized by alternating dyke swarms and septa of pillow lavas. More commonly, however, the sheeted dykes are separated from the lavas by faults that are interpreted (Tremblay et al., 1997; Bédard et al., 1998, 1999a, 2000) to have originally been steeply dipping normal faults. Fault-breccias may be impregnated with boninitic matrices, implying that extensional faulting was contemporaneous with seafloor spreading. Rare trondhjemitic dykes (~2 m wide) have shallow palaeo-dips. Spherulitic, sparsely amygdaloidal, sparsely porphyritic (2-6%) boninitic pillow lavas and hyaloclastite breccias constitute the Betts Head Formation in the Betts Cove area (Coish & Church, 1979; Coish et al., 1982; Bédard et al., 1998, 1999a, 2000). The thickest accumulation (up to 1·3 km) is in a ridge-related graben structure south of Betts Cove. Restricted exposures of boninitic volcanic rocks in the Long Pond and Tilt Cove areas (Fig. 1) are pervasively brecciated. Intra-breccia dykes and impregnations are principally boninitic, with compositions similar to those of Betts Head Formation lavas. A few intra-breccia dykes have tholeiitic compositions, similar to those of lavas from the overlying Mount Misery Formation (Snooks Arm Group). Locally heterolithic breccias contain detrital chromite and clasts of basalt, peridotite and gabbro, and probably represent fault talus or slump deposits developed during an amagmatic phase of extension associated with seafloor spreading. The presence of plutonic clasts implies unroofing of oceanic basement rocks during seafloor extensional faulting. Most of the rocks of the Betts Cove complex are affected by greenschist facies hydrothermal metamorphism(Coish, 1977b), although textural pseudomorphism allows reliable identification of most original phases. Electron microprobe compositions of chromite and clinopyroxene were obtained at McGill University with a JEOL superprobe. Methods and accuracy have been described by
Bédard & Hébert, (1996). The full dataset has been given by Bédard et al., (1999b, 2000). Additional data were compiled from Coish, (1977a). Along the shores of Red Cliff Pond and Long Pond are domains of massive talc-magnesite-ankerite-magnetite schists. Within a felted talc-magnesite matrix one can locally recognize olivine and orthopyroxene pseudomorphs. Euhedral magnetite octahedra locally contain ragged chromite relics. Holly-leaf chromites typical of mantle harzburgites are recognizable in places. Several types of dyke have been recognized at Betts Cove (see
Upadhyay, 1973; Coish, 1977a, 1977b; Coish & Church, 1979). Boninite dykes contain euhedral microphenocrysts of chromite (0·1-1 mm, 1-2%), and pseudomorphs after olivine and orthopyroxene phenocrysts (<5 mm, <10%). Clinopyroxene phenocrysts have also been reported (Coish, 1977a). Porphyritic boninite dykes (described as perknites in the older literature) contain larger (1-10 mm), more abundant (10-50%) phenocrysts of the same type. The matrix of the boninite dykes is composed of prismatic pyroxene pseudomorphs and interstitial feldspar + quartz. Some boninitic dykes contain plagioclase and clinopyroxene phenocrysts (1 mm, <10%), and abundant groundmass plagioclase. Pink-weathering microgabbroic and ferro-gabbro-noritic dykes are marginal facies and apophyses of the gabbro-norite intrusions. Blue-green-weathering, sparsely phyric diabase dykes are dominated by clinopyroxene and feldspar. Diabase dykes may contain clinopyroxene phenocrysts (~5%) and rare prismatic bastite pseudomorphs after orthopyroxene(?). Trondhjemite dykes contain pseudomorphs after phenocrysts of feldspar, quartz, mafic silicates and Fe-Ti oxides. The Betts Head Formation lavas belong to two subtypes. The most common subtype contains olivine phenocryst pseudomorphs (<10%), sparse, euhedral, chromite microphenocrysts and prismatic bastite pseudomorphs after orthopyroxene phenocrysts. Rare clinopyroxene phenocrysts are present locally (Coish, 1977b). The groundmass is similar to that of the boninite dykes. Interbedded with the olivine + orthopyroxene + chromite-phyric lavas are sequences (up to 200 m thick) of spherulitic lavas that may also contain plagioclase and clinopyroxene microphenocrysts and that have feldspar microlites in their groundmasses. Clinopyroxene phenocrysts in the dykes and lavas have low Ti contents (Coish & Church, 1979; Bédard et al., 1999b, 2000), which is typical of boninites (see
Crawford et al., 1989). Chromite phenocrysts in dykes and lavas have high Cr/(Cr + Al) = 0·70-0·85 (Coish & Church, 1979; Bédard et al., 1999b, 2000), also typical of those found in boninites (see
Crawford et al., 1989), but with a wider than usual range of Fe2+/(Mg + Fe 2+). Chromites from the talc-magnesite-ankerite-magnetite schists are compositionally distinct from those of the layered cumulates (Bédard et al., 2000), with lower Cr/(Cr + Al). New analyses of boninites from Betts Cove include 22 dykes or breccia matrices, 28 lavas and four clasts in breccias (Table 1). Sample locations are given as Universal Transverse Mercator coordinates in Table 1, and have been presented graphically by Bédard et al., (1999b). Altered margins and prominent veins were removed from the samples with a saw. Sawmarks were removed with sandpaper. Samples were then crushed in a steel jaw crusher, and subsequently reduced to powder in an agate shatterbox. Whole-rock powders were analysed for major and trace elements at the Centre géoscientifique de Québec (CGQ) laboratories. Major elements, Cr, Ni, Cu, Zn, Sr, Rb and Ba were analysed by conventional X-ray fluorescence (XRF) methods for samples collected in 1994, and by inductively coupled plasma atomic emission spectrometry (ICP-AES) for samples collected in 1995, 1996 and 1997. Other trace elements were studied by instrumental neutron activation analysis (INAA) for 1994 samples; and by ICP-MS on a VG Turbo Plasma Quad 2+ instrument, using a method similar to that described by
Varfalvy et al., (1997) for other samples. Analytical precision for ICP-MS analysis of tholeiitic lavas is better than 1% for La, Ce, Pr, Nd, Eu, Tb, Lu and Ba; 2% for Sm, Gd, Dy, Ho, Er, Tm, Yb, Th, Rb and Sr; 4% for Zr, Cs, Y and U; and ~10% for Nb and Pb. Precision is lower for the extremely depleted boninites, in which elemental abundances approach detection limits. Laflèche et al., (1998, appendix 1) have published analyses of international rock standards obtained in the CGQ laboratories. The new data reported here are supplemented by previously published data from
Upadhyay, (1973), DeGrace et al., (1976), Coish, (1977a), Jenner, (1977), Coish et al., (1982), Hurley, (1982), Saunders, (1985), Al, (1990) and
Swinden et al., (1997), which were compiled to yield the `average' analyses of Table 2. Most of the pre-1990 trace element data were not plotted, or included in the averages, however. Table 1. Low-Ti boninite lavas and dykes, Betts Cove
Table 2. Average end-member magmas
Samples affected by obvious hydrothermal alteration related to ore deposition (i.e. marked S and Fe enrichment), and extreme enrichment or depletion in Na2O, K2O or CaO, were excluded from the dataset used for petrogenetic interpretation and from the computed averages. Coish, (1977b), Coish & Church, (1979) and
Coish et al., (1982) concluded that Ti, P, Ni, Cr, Zr, Y, the rare earth elements (REE), Al2O3 and FeO*/MgO in the Betts Cove rocks largely reflect magmatic abundances. The concentrations of elements such as CaO and Na2O are commonly perturbed, however. Consequently, it is not possible to classify the Betts Cove boninites in a reliable way using the
Crawford et al., (1989) scheme, which requires accurate values for CaO, SiO2 and alkali elements. Instead, the Betts Cove lavas and dykes have been subdivided using relatively immobile trace elements, upon which the petrogenetic interpretations outlined below also principally depend. The elements Rb, Cs and K display random positive or negative anomalies on mid-ocean ridge basalt (MORB)-normalized trace element variation diagrams (Fig. 3) that are attributable to hydrothermal remobilization. Nevertheless, averaged values of these elements (excluding data from obvious metasomatic zones) do not show large anomalies in such diagrams (Fig. 3) in comparison with nearby, more immobile elements (Th, La), suggesting that Rb, Cs and K were redistributed on the scale of map units, but were not systematically leached or enriched from the ophiolitic complex. Comparison of U vs Th distributions in Betts Cove rocks (Fig. 4a) suggests that only a few strongly perturbed rocks have been enriched in U. In normalized trace element diagrams, Betts Cove boninites always show positive anomalies for Ba, U, Sr and Pb, similar to those of recent fresh boninites (Fig. 3a). One could speculate that these Betts Cove lavas originally had immobile element abundances identical to those of recent boninites from the Bonin Islands, but different concentrations of Ba, U, Sr, Pb and SiO2; and that it is the action of pervasive hydrothermal metasomatism that has somehow replicated boninitic signatures for these elements. However, considering all the other evidence pointing towards a boninitic affinity for the Betts Cove lavas, a hydrothermal origin for the resemblance to boninites in terms of Ba, U, Sr, Pb and SiO2 seems providential. Consequently, it is concluded that the abundances of these elements are similar to those in the original magmas. This is not to say that hydrothermal alteration has had no impact at all. For example, Ba (Fig. 4b), Sr and Pb (not shown) show rather poor correlations when plotted against Th. Hydrothermal alteration is almost certainly responsible for the large degree of dispersion in these elements in geochemical variation diagrams, but the high abundances of Ba, U, Sr, Pb and SiO2 characteristic of most of the rocks are probably not in themselves of hydrothermal origin. Figure
Figure
Betts Cove dyke complex rocks are compositionally indistinguishable from Betts Head Formation lavas (Fig. 5), suggesting that the sheeted dykes and lavas belong to a single comagmatic suite. The porphyritic dykes and lavas (picrites and perknites) have very high Cr, Ni and MgO contents (Fig. 5a) indicative of phenocryst accumulation (Coish & Church, 1979). The lavas of the Betts Head Formation have been subdivided by
Coish & Church, (1979) and
Coish, (1989) into `Low-Ti' and `Intermediate-Ti' suites (Fig. 5c). This first-order subdivision is adopted here. Rocks of the Low-Ti suite correspond to the olivine + orthopyroxene + chromite ± clinopyroxene phyric dykes and lavas, whereas rocks of the Intermediate-Ti suite correspond to the clinopyroxene ± plagioclase ± olivine phyric dykes and lavas. In practice, many rocks have been assigned to these suites on the basis of their TiO2 vs FeO*/MgO (Fig. 5c) and La/Nd vs La (Fig. 5d) distributions. Figure
Overall, the Betts Cove Low-Ti lavas and dykes define trends of decreasing Cr (Fig. 5a) and Ni (not shown) with increasing FeO*/MgO, and a steep, diffuse trend of SiO2 enrichment (Fig. 5b). Rocks of the Low-Ti suite have low contents of most incompatible elements (Figs 6a and 7). Their normalized trace element profiles show relative enrichment in large ion lithophile elements (LILE) and LREE, in comparison with the middle REE (MREE), which gives them `U' shapes (Figs 6a and 7a). Enrichment in LREE is variable, as reflected in the wide range of La/Nd ratios (Fig. 5d). Negative Nb anomalies, and positive Pb, Sr and Zr anomalies are typical (Fig. 7a). Figure
Figure
Most Low-Ti suite rocks are geochemically almost indistinguishable from recent Bonin Island Low-Ca boninites (Figs 5-
7); and have similar mineralogical characteristics, including low-Ti clinopyroxene, high-Cr/(Cr + Al) chromites, and abundant orthopyroxene. The mineralogical and geochemical data support the inference that rocks of the Low-Ti suite are boninites, in accord with
Upadhyay, (1980), Coish et al., (1982), Coish, (1989), and
Swinden et al., (1989). The strong resemblance to Bonin Island lavas suggests that they should be classed as Type 3 Low-Ca boninites in the scheme of
Crawford et al., (1989). Intermediate-Ti suite lavas and dykes (clinopyroxene ± plagioclase ± olivine phyric) have flatter normalized trace element profiles than do the Low-Ti boninites (Figs 6b and 7b), with overall higher contents of moderately incompatible elements, but have similar LILE contents, negative Nb anomalies, and positive Pb and Sr anomalies. Whether the Intermediate-Ti suite rocks are best classified as arc tholeiites or boninites is uncertain (see
Coish, 1989). Swinden et al., (1997) referred to them as arc tholeiites. The presence of feldspar phenocrysts in some rocks does suggest an affinity to tholeiites, and they are transitional towards Mount Misery tholeiites in many respects (Figs 5-
7). The Mount Misery tholeiites have normalized incompatible-element profiles sub-parallel to those of the Intermediate-Ti boninites (Figs 6b and 7b), but have consistently higher contents of moderately incompatible elements in comparison with the boninitic rocks (Figs 5-
7). Because the Intermediate-Ti suite rocks are interbedded with Low-Ti boninites, and fall within the compositional range of boninites in most variation diagrams (e.g. Fig. 5), in this paper they will be referred to as boninites. Can fractional crystallization alone explain the compositional spectrum observed within the Betts Cove boninitic lavas and dykes? Fractionation models involving incremental extraction of olivine, orthopyroxene and chromite were developed to test this possibility (Appendix 1). These models are inappropriate for fractionation of plagioclase-bearing assemblages, so the thermodynamically based
Weaver & Langmuir, (1990) crystal fractionation program was used to model differentiation of the locally plagioclase-phyric Intermediate-Ti magmas. Parental magma compositions for the Low-Ti and Intermediate-Ti boninite series in these calculations are simply the averages of all analyses with FeO*/MgO <0·61 and <0·74, respectively (Table 2). The model results (Fig. 8) show that closed-system fractional crystallization cannot explain the observed ranges of SiO2, FeO*, TiO2 or La/Nd ratios within the Low-Ti or Intermediate-Ti boninite suites. Nor can it generate residual magmas similar to Intermediate-Ti lavas from Low-Ti boninite parental melts, or Mount Misery tholeiites from Intermediate-Ti parents; in agreement with the conclusions of
Coish et al., (1982). Fractionation coupled with assimilation of trace element depleted lower-crustal or mantle lithologies does not yield significantly different results and so cannot change these conclusions. Therefore, the wide range in SiO2 and incompatible trace element ratios in each suite must reflect either melting processes or source heterogeneity. Table 3. Calculated model parental end-member magmas
Figure
Partial melting models are developed to determine whether or not mantle melting processes can explain compositional variations among the Betts Cove lavas and dykes. Rather than use completely ad hoc compositions, the modelling uses mantle compositions calculated from the Betts Cove data. The solutions are non-unique, but they provide a realistic starting point for modelling intra-suite geochemical variation. Betts Cove end-member magmas The first step in the modelling process is to define the different end-member magmas. Stratigraphic relations allow a first-order division into the Mount Misery tholeiites and the Betts Head-sheeted dyke boninites. The boninites are subdivided into Low-Ti and Intermediate-Ti suites, as proposed by
Coish & Church, (1979). The Low-Ti boninites are further subdivided into end-members to allow quantitative modelling of source heterogeneity. This is done by defining a number of end-member boninite `poles' or subtypes that bracket the range of observed compositions. These end-members do not correspond to precise stratigraphic intervals, and should be viewed only as the most extreme poles of a continuous compositional spectrum. The La/Nd ratio should reflect the amount of the slab-derived SZ component added to the depleted mantle wedge (Pearce & Parkinson, 1993), and this ratio is used to define one set of end-members. Thus, `High-La/Nd' and `Low-La/Nd' end-members are defined as the Low-Ti boninites with the highest and lowest La/Nd ratios, respectively (Table 2). During the course of the modelling it became apparent that some Low-Ti boninites had higher Th abundances than most of the others, whrereas others were richer in Nb. This suggests that additional components or processes might be recorded by these Th- or Nb-rich boninites, and, for this reason, a pair of `High-Th/La' and `High-Nb' end-members were also defined. Recalculation to bring end-member magma compositions into equilibrium with the mantle All of these end-member magma compositions have FeO*/MgO ratios that are too high to coexist with normal or depleted peridotite mantle, and so a parental magma composition was calculated for each end-member magma (Table 3) assuming fractional crystallization (Appendix 1) of a mineral assemblage comprising 99% olivine and 1% chromite. The Intermediate-Ti boninite parent and most Low-Ti boninite parents were brought into equilibrium with Fo~92 olivine, which would be appropriate for a refractory wedge peridotite. The Low-La/Nd end-member is extremely depleted in incompatible trace elements, which might reflect a greater extent of prior mantle depletion. For this reason, the Low-La/Nd end-member was recalculated to bring it into equilibrium with a slightly more refractory olivine of Fo93·6 composition. In all cases, the degree of calculated back-fractionation is small (8-13%, Table 3), and the assumptions made about the extent of fractionation and the phases involved (olivine or orthopyroxene) have only a small effect on the absolute abundances of incompatible trace elements in the calculated parental magmas (e.g. see Fig. 8b). Furthermore, fractionation of small amounts of olivine, spinel or orthopyroxene does not significantly change the characteristic shape of incompatible trace element profiles, and so all calculated parental magma profiles (Fig. 9) retain the small positive Zr anomalies, large positive Pb + Sr anomalies, negative Nb anomalies, and the enrichment in Th + LILE of the uncorrected data. Figure
Source composition and inversion models To reconstruct the composition(s) of the mantle source from a lava, it is necessary to make assumptions about the nature of the melting process (equilibrium, fractional, or critical melting), to constrain the modal composition of the residue in equilibrium with the parental magmas, and to estimate the degree of fusion involved. The inversion calculations presented here assume equilibrium, or batch melting, as there are no unique inversion solutions for fractional or critical melting. This necessary simplification is probably not a fatal flaw, as equilibrium and pooled fractional melts have similar compositions for moderately incompatible elements (Fig. 8c, and see below, Fig. 13). Furthermore, comparison of profiles representing the residues of fractional, critical and batch melting suggests that the form of the moderately incompatible trace element profile in the residues is not strongly dependent on the melting process. The abundances of the very highly incompatible elements (Cs, Th, U, Nb, La, Ce), in contrast, are strongly dependent on the nature of the melting process. However, they are even more sensitive to source composition (Pearce & Parkinson, 1993), and so the source mantle profiles calculated from different end-member magmas (assuming similar residual modes and degrees of fusion) should provide a useful baseline for comparison. The trace element composition of the mantle residue in equilibrium with a given magma can be calculated by summation:
The modal proportion of the different residual minerals is [phi]. The concentration of a trace element in the melt is CL, and CRM is the composition of a trace element in the residual mantle. The mineral-liquid trace element partition coefficient for a given element is D (see Appendix 2, Table A1). The composition of the pre-melting source mantle (CSM) can then be reconstructed by adding appropriate proportions of the melt (Table 3) to its assumed residual mantle assemblage (Table 4):
Table 4. Model mantle source compositions
The composition of the melt (CL) dominates the trace element budget of the model sources [equation (2)] at degrees of melting >1%. Varying the assumed value of F (degree of melting) from 5 to 10% has a negligible effect on the shape of the model source's trace element profile (Fig. 10), although it does have a small effect on the absolute abundances of trace elements. Although small errors in the assumed value of F cannot change the conclusions obtained, the models will be more realistic if accurate estimates of F are used. Van der Laan et al., (1989) compared boninite compositional data with experimentally determined liquidus phase boundaries and proposed that Low-Ca boninites formed through 7-13% mantle melting. Consequently, most Low-Ti Betts Cove parental boninite end-members are assumed to represent 10% melting. A lower degree of melting is assumed for the depleted Low-La/Nd parental magma (6%), so as to constrain its model source mantle to have a moderately incompatible trace element content less than that of the model High-La/Nd source mantle, and thus permit mass balance calculation of SZ component compositions (see below). The Intermediate-Ti boninite parental magma, which is less depleted in incompatible elements, is assumed to have formed through a slightly greater degree of fusion (12%). Figure
The modal composition of the residue ([Sigma][phi]) must also be assumed in equations (1) and (2). Although small variations in the assumed residual mode (e.g. clinopyroxene 0-3·5%, orthopyroxene 8·7-13%) have only a small effect on the calculated source profiles (Fig. 10), the results will be more accurate if a realistic residual assemblage is used. The residual mode used in the models is determined as follows. First, an initial modal estimate is obtained for each end-member magma type on the basis of phase equilibria studies, which imply that boninite magmas separate from extremely depleted, clinopyroxene-free dunitic to harzburgitic residues, in contrast to most tholeiites and arc basalts, which separate from less depleted harzburgite to lherzolite residues (Howard & Stolper, 1981; Dick et al., 1984; Crawford et al., 1989; Falloon et al., 1989; Parkinson & Pearce, 1998). Consequently, a residual harzburgite assemblage (DM-10Bon residual mode in Table 4) is assumed for an initial calculation of all Low-Ti boninite residual mantle subtypes using equation (1), and pre-melting model source mantles using equation (2). Because the Intermediate-Ti boninites are less depleted in incompatible trace elements, a more fertile clinopyroxene harzburgite residue (clinopyroxene:orthopyroxene:olivine:spinel = CPX:OPX:OL:SP = 1:10:74:3) is assumed for them. The trace element profiles of these initial models are then compared with theoretical residual mantle profiles formed by equilibrium melting of FMM (fertile MORB mantle, Table A2, see Appendix 3). The modal assemblage of the most closely matching theoretical FMM residue is dunitic for most Low-Ti boninite source mantles and is lherzolitic for the Intermediate-Ti boninite source mantle (Table 4). These residual modes were adopted for a second (and final) inversion calculation using equations (1) and (2). The first- and second-pass results of the calculations are essentially identical, and no further iterations are necessary. Comparison of the model source mantles calculated from the Betts Cove model parental magmas with the theoretical residues of FMM allows the degree of prior melt extraction from their source mantles to be determined (Fig. 11; the full profiles are shown in Fig. 12, and the values are given in Table 4). The comparison implies that the Betts Cove Low-Ti boninite source had already lost 19-22% equilibrium melt; whereas the source of the Intermediate-Ti boninites was less depleted, having lost only ~12% equilibrium melt. These estimates of the extent of prior source depletion are similar to published estimates for wedge mantle depletion (Ewart & Hawkesworth, 1987; McCulloch & Gamble, 1991; Pearce & Parkinson, 1993; Woodhead et al., 1993, 1998; Pearce et al., 1995; Ewart et al., 1998; Parkinson & Pearce, 1998). This depletion may be related to seafloor spreading at an oceanic ridge, to back-arc spreading, or to earlier arc magmatism. The lack of relative HREE/MREE fractionation among the different model sources (Figs 11 and 12) suggests that this prior depletion took place in the spinel lherzolite field. Figure
Figure
Most model source mantle profiles calculated using equation (2) for the different Low-Ti boninite end-members are very similar. Like the melt compositions from which they were computed, the model source mantle profiles (Fig. 12) typically show enrichment in Ba, Th, U, LREE, Sr and Pb (also Rb and Cs, not shown), variable Zr enrichment, and Nb depletion. The Low-La/Nd source mantle profile differs from the others in having a smaller positive Zr peak and in lacking Th and LREE enrichment. Like the Low-Ti boninite model mantles, the Intermediate-Ti model mantle is enriched in LILE, Pb and Sr, and depleted in Nb, but it is notably more enriched in MREE to HREE. The massive talc-magnesite schists along the shore of Red Cliff and Long Ponds are mineralogically and compositionally distinct from the layered cumulates and may represent slivers of mantle rock. Model mantles calculated from the lavas are compared with analyses of these talc-magnesite schists in Fig. 12d. Pervasive hydrothermal alteration and addition of carbonate to the talc-magnesite schists has remobilized some elements, leading to considerable scatter in Ba, U, Sr and Eu. When the less mobile elements are considered, however, the profile of the average talc-magnesite schists is rather similar to the model mantle compositions computed here (Fig. 12d). The overall similarity of composition between the analyses of
Al, (1990) and the models presented here suggests that the model results shown in Fig. 12 are at least approximately correct. The largest discrepancy is for Th, which is systematically high in the rocks analysed by
Al, (1990). This probably reflects the relative imprecision of the Th data (all are 0·1 ppm) from
Al, (1990). Closed-system melting models Melting models were developed (Appendix 3, Figs 8c and 13) to test whether variable extents of melting of a single source mantle could have produced the range of compositions (Figs 5d, 6 and 7) observed in the boninites at Betts Cove. Given that the model sources defined above approximate the composition of the actual sources, then the results of the melting models imply that fractional, equilibrium or critical melting of a single source mantle cannot generate the range of trace element ratios seen in the Betts Cove boninites. Figure
Metasomatic models It has been proposed that percolation of partial melts through their own mantle residues generates arc-like geochemical signatures (depletion in Nb and Ti, and relative LREE/MREE enrichment) in the magmas (Kelemen et al., 1990, , 1993). A powerful argument against an autometasomatic model is the heterogeneity of isotopic signatures from arc-related volcanic suites, including boninites, which implies the involvement of an external SZ component (Brown et al., 1982; White & Dupré, 1986; Woodhead, 1989; Olive et al., 1997). Limited isotopic data from Betts Cove (Coish et al., 1982; Swinden et al., 1997) also imply an external SZ component. There may be complex metasomatic reactions associated with the influx of such a fluid-rich SZ component into the depleted mantle wedge (e.g. Bodinier et al., 1990; Van der Wal & Bodinier, 1996), but these processes cannot easily be reconstructed from the lava chemistry. The simplifying assumption that will be made here is that the Betts Cove lava sources can be modelled by simple addition of an SZ component to a depleted mantle wedge. How many subduction zone components? To create the characteristic trace element and isotopic patterns of arc magmas, one or more SZ components derived from the subducting slab need to be added to the depleted mantle wedge (Rogers et al., 1985; Ellam & Hawkesworth, 1988; Elliott et al., 1997; Kepezhinskas et al., 1997; Turner et al., 1997; Ewart et al., 1998; Gribble et al., 1998). There are considerable data implying that the elements B, Pb, Ba, Rb, U and Sr are largely carried by a hydrous fluid (SZ-Hydrous) produced when the subducting, hydrothermally altered oceanic crust is heated and devolatilizes (Hickey & Frey, 1982; White & Dupré, 1986; Davidson, 1987; Tatsumi, 1989; Lin et al., 1990; McDermott et al., 1993; Miller et al., 1994; Brenan et al., 1995; Elliott et al., 1997; Turner et al., 1997; Iwamori, 1998; Schmidt & Poli, 1998; Stalder et al., 1998; Woodhead et al., 1998). Experimental data and numerical modelling (Peacock et al., 1994; Iwamori, 1998; Schmidt & Poli, 1998) imply that when young, hot, oceanic crust is subducted, the hydrated basaltic-gabbroic part of the subducting slab may begin to melt at shallow depths, yielding wet melts of adakitic, trondhjemitic or tonalitic composition (e.g. Cameron, 1985; Defant & Drummond, 1990; Schiano et al., 1995). These melts will be referred to as SZ-ATT henceforth. Pearce et al., (1992) and
Taylor et al., (1994) have suggested that SZ-ATT was involved in boninite genesis. The isotopic signatures of SZ-Hydrous and SZ-ATT are probably indistinguishable, as they have the same source, but trace element chemistry should allow discrimination, because SZ-ATT is a silicate melt and so should contain significant concentrations of Zr, Th and LREE (Fig. 14a). In contrast, because of the relative insolubility of these elements in water (Brenan et al., 1995; Keppler, 1996; Ayers et al., 1997; Stalder et al., 1998), SZ-Hydrous should contain little or no Zr, Th or LREE. Figure
Analysis of in situ partial melts of oceanic crust from ophiolites (plagiogranites or trondhjemites) and of modern adakites can also help to constrain the composition of the SZ-ATT end-member (Figs 14a and 15). There are basically two types of ophiolitic trondhjemites (e.g. Pederson & Malpas, 1984; Elthon, 1991; Flagler & Spray, 1991; Jenner et al., 1991): those that formed by anatexis of amphibolitized gabbros, and those derived through fractionation of basalts. Typical anatectic ophiolitic trondhjemites have rather flat REE profiles (Fig. 14a), with low La and La/Nd (
Fig. 15a), and high Zr and Zr/Sm (Fig. 15b). The negative Nb and positive U anomalies of ophiolitic trondhjemites probably reflect the arc affinity of their protoliths (e.g. Elthon, 1991). Adakites differ from trondhjemites in showing LILE-Pb-Sr enrichment (Fig. 14a). Although the overall similarity to trondhjemites (Fig. 14a) supports the concept that adakites are indeed melts of the oceanic crust, many of their trace element characteristics (Fig. 15) more closely resemble those of sediments, which implies that a sedimentary component must also be involved in their genesis (e.g. Defant & Drummond, 1990; Maury et al., 1996; Stern & Kilian, 1996). Figure
In many arcs, distinctive trace element and isotopic data appear to require the involvement of a component derived from wet melting of subducted sediments (White & Dupré, 1986; Davidson, 1987; Ellam & Hawkesworth, 1988; Woodhead, 1989; McDermott et al., 1993; Cousens et al., 1994; Brenan et al., 1995; Pearce et al., 1995; Elliott et al., 1997; Turner et al., 1997; Plank & Langmuir, 1998). I will refer to this component as SZ-Sediment henceforth. Oceanic sediments typically show prominent positive LILE and Pb anomalies, and negative Ti and Nb anomalies (Fig. 14a). They rarely show Sr or Zr enrichment relative to the REE. SZ-Sediment has been identified as the principal source of enrichment in 10Be, LREE and Th in arc magmas, and may also contribute to the Zr, U and Pb budget. Many of the trace element signatures interpreted by some as SZ-Sediment have been attributed to involvement of an SZ component similar to ocean island basalts (SZ-OIB) by others (Fig. 14a). It was originally suggested that SZ-OIB resides in the supra-subduction zone wedge as dispersed melt or veins (e.g. Morris & Hart, 1983; Stern & Ito, 1983; Falloon & Crawford, 1991; Stern et al., 1991; Kostopoulos & Murton, 1992; Lin, 1992). Stern et al., (1991) proposed that SZ-Hydrous derived from the slab leaches the dispersed SZ-OIB component from the mantle wedge and carries it up to the melting zone. More recently, others have proposed that SZ-OIB enters the arc source through subduction of hotspot-related seamounts (Turner et al., 1997; Ewart et al., 1998), or by movement of plume-related material around the edge of a subducting slab (Wendt et al., 1997; Ewart et al., 1998), or by refusion of OIB-related residues (Danyushevsky et al., 1995). Ascent of undepleted mantle during back-arc extension is another possible mechanism to explain the presence of OIB-like signatures in arc magmas (Lin et al., 1990; Gribble et al., 1998). Determining which of these numerous potential SZ components is responsible for the distinctive geochemical and isotopic signatures of a given arc suite is a difficult problem. Typically, the identity and proportion of these components are identified with isotopic data. However, there is always the suspicion that some isotopic systematics may be decoupled from the trace element concentrations, so that it would be reassuring if an independent argument could be constructed using the trace element data alone. In the following sections I use the model mantle trace element compositions calculated previously to determine the compositions of the SZ component involved in the petrogenesis of the Betts Cove boninites, in an attempt to discriminate between competing hypotheses. The first step is to calculate the composition of the depleted wedge component from the least enriched model mantle composition (Low-La/Nd) defined above. Then, mixing calculations between the depleted end-member and the most enriched end-members (High-La/Nd, High-Th/La, High-Nb) allow the composition of the SZ component(s) to be calculated. The procedure should be valid for any suite of primitive magmas where there is a range of trace element profile shape. The mantle wedge: the depleted component The Low-La/Nd model mantle has the most depleted trace element profile of all the Betts Cove model mantles calculated above (Fig. 12a), and so is assumed to best represent the pre-fertilization depleted mantle wedge (Pearce, 1983; Pearce & Parkinson, 1993). However, even this depleted Low-La/Nd mantle has prominent positive Ba + U + Sr + Pb anomalies (Fig. 12a). As these are precisely the elements thought to be associated with SZ-Hydrous, it seems reasonable to attribute this enrichment pattern to addition of SZ-Hydrous to the depleted wedge. The Low-La/Nd mantle lacks enrichment in Th + LREE + Zr, however, and so it is logical to infer that other SZ components did not affect the Low-La/Nd boninite mantle source. Conversely, the ubiquitous presence of prominent positive Ba + U + Sr + Pb anomalies in all Betts Cove boninites and Mount Misery tholeiites (Fig. 7) suggests that SZ-Hydrous may have more widely dispersed in the mantle under Betts Cove. The composition of the pre-metasomatic wedge mantle can be approximated in a manner analogous to that proposed by
Pearce, (1983) and
Pearce & Parkinson, (1993), by removing the prominent spikes (SZ-Hydrous component) from the model Low-La/Nd mantle by interpolation and extrapolation (Fig. 12a). This pre-SZ Low-Ti boninite wedge source is referred to as the Low-La/Nd(I) mantle (I for interpolated), henceforth, and its composition is given in Table 4. A flat profile that extends from Nd and passes just below Pr and Nb is inferred for Low-La/Nd(I), with a shallow, positively sloped profile linking Nd to Eu (Fig. 12a). Use of a differently sloped profile for the Nd-Cs segment of Low-La/Nd(I) would change the absolute abundances and slopes of the model SZ-component profiles calculated below, but would not affect their shapes (peaks and troughs), and would only slightly change their ratios. An Intermediate-Ti(I) profile can be calculated in a similar fashion, by passing a straight line from Ti through Nb (Fig. 12c, Table 4), thus providing an estimate of the pre-SZ Intermediate-Ti boninite wedge source. Trace element abundances for Low-La/Nd(I) and Intermediate-Ti(I) source mantles fall on the depleted mantle curve of
Pearce et al., (1995, see Fig. 16c), suggesting that the estimates are good. Calculation of SZ component compositions The composition of the different SZ components at Betts Cove can be determined by mass balance using the model mantle compositions calculated above. As argued above, the Low-La/Nd model mantle is identified as the depleted wedge + SZ-Hydrous. Therefore, the composition of SZ-Hydrous can be calculated from the Low-La/Nd and Low-La/Nd(I) model mantles if the proportions of the components in the mixture are fixed. If z is the fraction of SZ-Hydrous in the mixture, and Ci is the concentration of a trace element in a given reservoir, then
This calculation probably yields results that are only approximately correct for the Betts Cove suite for two reasons. First, the Low-La/Nd(I) profile is inferred, and small shifts in the real position of this profile will have large effects on the computed value of SZ-Hydrous. Second, the two rocks constituting the Low-La/Nd end-member are rather altered, and most of the elements that define the prominent peaks (Sr, Pb, Ba, U) have probably been perturbed somewhat. The calculation of SZ-Hydrous would be more applicable to fresher, modern rocks. By comparing the model Low-La/Nd(I) mantle with the different Low-Ti boninite end-member model mantles, it is possible to calculate the composition of the total SZ contribution to the depleted wedge for each end-member. This calculation is more robust than the calculation of SZ-Hydrous, as the end-member profiles are computed from larger numbers of relatively fresh rocks. The value of SZ-Total is the sum of all the SZ components involved (Hydrous, Sediment, ATT or OIB). Taking the High-La/Nd boninite model mantle as an example,
The compositions of SZ-Total components involved in the genesis of the High-Nb and High-Th/La end-members are calculated in the same way. The composition of SZ-Total involved in generation of Intermediate-Ti boninites can be calculated in a similar fashion:
Figure 14b shows how the abundances and profiles of the model SZ-Total component associated with the High-La/Nd end-member change with the mixing proportion (z). Changing z does not affect the characteristic profile shapes or incompatible element ratios, although assuming a smaller value of z does cause the absolute abundance of the calculated SZ-Total component to increase. The SZ-Total components calculated from the different boninite end-members (Intermediate-Ti, High-La/Nd, High-Th/La, and High-Nb) are all rather similar (Fig. 14b and Table 5), and for values of z = 0·25% provide approximate matches for immobile trace element abundances with putative SZ components (OIB, Sediment, ATT, Fig. 14b). Such small proportions of the SZ component are consistent with the proportions inferred from isotope data for Tertiary boninites and other arc suites in modern environments, and for boninites in other Appalachian ophiolites (McDermott et al., 1993; Pearce & Parkinson, 1993; Olive et al., 1997; Turner et al., 1997). Table 5. Model SZ components
The model SZ component profiles calculated with equations (4) and (5) (except for the SZ component associated with the High-Nb end-member) have characteristic troughs at Nb and Ti (Fig. 14b); this feature is probably inconsistent with the involvement of a dispersed SZ-OIB component at Betts Cove. Comparison of the Model SZ-Total incompatible trace element profiles with published data on putative SZ components allows the identity of the SZ component at Betts Cove to be further refined (Figs 14c and 15). Typical trondhjemites have flatter REE profiles than typical sediments (Fig. 14a), with lower La and La/Nd, and higher Zr and Zr/Sm (Figs 14a and 15). The model SZ-Total components calculated from the Betts Cove Low-Ti boninites closely resemble typical sediments in terms of the LILE, LREE, Nd and Pb (Figs 14b, c and 15a); but more closely resemble trondhjemites in terms of the MREE and Zr (Fig. 15b). This strongly suggests that both SZ-Sediment and SZ-ATT were involved in refertilizing the depleted mantle wedge source of the Low-Ti boninites. The SZ-Total calculated from Intermediate-Ti boninites, in contrast to the Low-Ti end-members, shows a much greater resemblance to the trondhjemites for most incompatible elements (Figs 14 and 15), and one could posit that SZ-ATT is the dominant influence in their genesis. The pronounced enrichment in Pb and LILE shown by the model SZ-Total components also suggests the involvement of SZ-Sediment, but as the Betts Cove lavas are altered, less confidence can be placed on variations of these elements. Nevertheless, it is interesting that the so-called mobile elements suggest exactly the same conclusions as do the immobile ones. Elements such as Ba, U, Pb and Sr are very enriched in most SZ-Total models, with abundances greater than typical terrigenous SZ-Sediment components (Fig. 14b). This is interpreted to be due in part to addition of a ubiquitous SZ-Hydrous component. Alternatively, some of these enrichments might reflect the presence of carbonate (Sr) in the subducted sediment, or simply the natural heterogeneity to be expected in sediments. Support for the interpreted presence of small amounts of SZ-Sediment also comes from published Nd-isotopic data for these rocks (Coish et al., 1982; Swinden et al., 1997). The computed values for the SZ components calculated from the Betts Cove data are used as end-members for mixing calculations shown in Fig. 16. Values for Ba in SZ-Total were calculated with equation (4). Ba is shown, as it may be less mobile than the other LILE. High-Ba/La in arc lavas is typically interpreted to signify involvement of SZ-Hydrous, whereas high values of La/Sm (or La/Nd) signify higher proportions of SZ-Sediment. It should be noted that involvement of both SZ-Hydrous and SZ-Sediment are needed to explain the compositions of Betts Cove and other boninitic suites (Fig. 15a). In addition, the presence of High-Th/La boninites (Fig. 15b) suggests that some of these lavas involve a High-Th sediment component as well. Figure
Intermediate-Ti boninites Intermediate-Ti boninites at Betts Cove cannot form through partial melting of the same sources as the Low-Ti boninites (Figs 8c and 12b), because their La/Nd ratios are too low, and their HREE contents are too high for this to be plausible. Model calculations (Figs 11 and 14c) imply that the Intermediate-Ti boninites formed in much the same way as the Low-Ti boninites, i.e. through melting of a depleted source coupled with an influx of a trace element enriched SZ component. However, the Intermediate-Ti source mantle was less depleted initially than was the source of the Low-Ti boninites (Fig. 11). Model results imply that the genesis of both Low-Ti and Intermediate-Ti boninites involved similar SZ components (Figs 14c and 15), but that the Intermediate-Ti boninites were refertilized by a greater relative proportion of SZ-Hydrous (higher Ba/La) and SZ-ATT (Figs 15 and 16). This inference is consistent with the isotopic data of
Swinden et al., (1997), which indicate that less SZ-Sediment was involved in genesis of the Intermediate-Ti boninites (their Island Arc Tholeiites). At Betts Cove, the sheeted dykes, the cumulates and the Betts Head lavas constitute a comagmatic suite of boninitic affinity (Bédard et al., 1998, 2000). As Tertiary boninites appear to be restricted to forearcs (Cameron et al., 1979; Hawkins et al., 1984; Murton, 1989; Johnson & Fryer, 1990; Stern & Bloomer, 1992), this suggests that seafloor spreading at Betts Cove may have been initiated in a forearc environment (Bédard et al., 1998), probably in a peri-continental setting (Harris, 1992; Pinet & Tremblay, 1995). This would account for the common occurrence of ophiolitic boninites along the margin of North America (Church, 1977; Bédard & Hébert, 1996). A detailed discussion of the interrelation of Taconic deformation and the accretion of ophiolites to North America is beyond the scope of this paper. However, the data from Betts Cove provide constraints on the sequence of events attendant on formation of one of these accreted Ordovician peri-continental marginal basins. The narrow active volcanic zone of arcs is commonly interpreted to reflect focused transfer of an LILE- and volatile-rich SZ component from the subduction zone into the mantle wedge above (Fig. 2) (Gill, 1981; Tatsumi, 1989; Arculus, 1994; Keppler, 1996; Iwamori, 1998). Any change in the geometry of subduction (e.g. in response to trench rollback) should disrupt the focused volatile-rich efflux derived from the slab and distribute it over a larger volume of the depleted mantle wedge (Bédard et al., 1998). Considering the small proportion of volatile-rich SZ component (<0·5%) calculated for Betts Cove boninites, pure flux melting alone is probably not capable of accounting for the huge volumes of boninitic melt required, and so decompression melting (Fig. 2) must also be involved (see
Pearce & Peate, 1995). This implies that extension of the overriding plate must have accompanied subduction. If slab rollback and extension of the overriding plate are synchronous, then the combination of a dispersed volatile component, decompression-melting of the mantle at fairly low pressures, and a depleted wedge source, would all favour production of boninitic magmas, in addition to providing the opportunity for magma-dominated extension (seafloor spreading). At Betts Cove, there is a systematic progression from extremely depleted boninites (Low-Ti Betts Head Formation), to less depleted boninites (Intermediate-Ti), to still less depleted arc tholeiites (Mount Misery Formation), and eventually to the fairly enriched Upper Snooks Arm Group tholeiites. These fertile tholeiites are interstratified with calc-alkaline lavas and pyroclastic rocks, and the sedimentary and volcanic facies (Bédard et al., 2000; Cousineau & Bédard, 2000; Kessler & Bédard, 2000) indicate that both tholeiites and calc-alkaline volcanics are extremely proximal to their respective vents. This systematic stratigraphic relationship is inconsistent with a model where the chemical signatures of the different lava types reflect the heterogeneous distribution of a fossil SZ-OIB component in the wedge, as the most fertile mantle domains should melt first, not the most refractory ones. One must therefore posit a spatial compositional zonation in the mantle beneath the Betts Cove ophiolite in the Ordovician. As the magmas derived from the more fertile mantle domains post-date the onset of boninitic seafloor spreading, it is reasonable to infer that the more fertile mantle domains were located at greater depth, and were entrained into the extending supra-subduction zone wedge to fill the space created by slab rollback and seafloor spreading (Fig. 2). The Betts Cove Ophiolite records the initiation of seafloor spreading in a marginal basin characterized by magmas of boninitic affinity. The source mantle of the Low-Ti boninites was a refractory harzburgite or dunite residual after 19-22% melting of a fertile MORB mantle (FMM). Decompression melting was assisted by the addition of small proportions (<0·25%) of a fluxing SZ component, identified as a mixture of volatiles derived from the subducting oceanic crust (SZ-Hydrous), siliceous melts derived from fusion of subducted sediments (SZ-Sediment), and partial melts of the subducting oceanic crust (SZ-ATT). There is no evidence for the involvement of OIB-like components at Betts Cove. Intermediate-Ti boninites were derived from less depleted mantle sources (~12% prior melting of FMM), fluxed with a greater proportion of SZ-Hydrous and SZ-ATT, and less SZ-Sediment (in comparison with the Low-Ti boninites). The gradation from extremely depleted boninites to less depleted boninitic and tholeiitic magmas with time implies a change in source composition, with less depleted sources being entrained into the zone of melting, perhaps in response to slab rollback and extension of the overriding plate. A. Tremblay and K. Lauzière contributed to all phases of this study and their assistance is gratefully acknowledged. Charles Langmuir supplied modelling software. Louise Corriveau, Marc Laflèche, Ray Coish, Julian Pearce, Ian Parkinson, Robert J. Stern, Rosemary Hickey-Vargas and Marjorie Wilson commented on earlier versions of the manuscript. Their contributions were invaluable. K. Lauzière prepared Fig. 1. The people of Tilt Cove, LaScie and Snooks Arm made the fieldwork unforgettable. Fieldwork and geochemical analyses were financed through a Federal-Provincial Mineral-Development Agreement, by an Industrial Partenariat programme between the Geological Survey of Canada and Ressources Noveder, Exploration Sulliden and Ressources Dianor. This is Geological Survey of Canada Contribution 1999044.INTRODUCTION
REGIONAL GEOLOGICAL FRAMEWORK
FIELD RELATIONS
Sheeted dykes
Betts Head Formation lavas
PETROGRAPHY AND MINERALOGY
Massive talc-magnesite-ankerite-magnetite schists
Sheeted dykes
Betts Head Formation lavas
Mineral chemistry
WHOLE-ROCK GEOCHEMISTRY
Sample preparation, analytical methods and data sources
Impact of hydrothermal metamorphism
Betts Cove lavas and dykes
PETROGENESIS
Fractional crystallization
Partial melting models or source heterogeneity?
(1)
(2)
Multiple SZ components added to the depleted wedge
CiLow-La/Nd(I) + zCiSD-Hydrous = CiLow-La/Nd.
(3)
CiLow-La/Nd(I) + zCiSZ-Total High-La/Nd = CiHigh-La/Nd.
(4)
CiIntermediate-Ti(I) + zCiSZ-Total Intermediate-Ti = CiIntermediate-Ti.
(5)
SEAFLOOR SPREADING IN A MARGINAL BASIN
CONCLUSIONS
ACKNOWLEDGEMENTS
REFERENCES
Fractional crystallization was modelled through incremental subtraction of equilibrium olivine, orthopyroxene and spinel in steps of 0·5%. Comparisons with results calculated with the Rayleigh fractionation equation showed negligible deviations after 90% fractionation. Melt Fe3+/Fe2+ ratios were calculated from liquid compositions and temperatures using the SPINMELT program of Ariskin & Nikolaev (1996) with the procedure of Sack et al. (1980), assuming that oxygen fugacity was buffered at quartz-fayalite-magnetite (QFM) conditions. An initial liquidus temperature of 1300°C was assumed, with the temperature-crystallization curve as described by Bédard & Hébert (1998) for the Tongan boninite parent. The average phenocryst spinel composition from Betts Cove lavas and dykes was used in all calculations. Olivine FeO/MgO compositions were calculated with an exchange coefficient of 0·3 (Roeder & Emslie, 1970); Ni in olivine with ol/liqDNi = 10 (Kinzler et al., 1990), and Cr2O3 in olivine with ol/liqDCr = 1. Olivine MnO was fixed at 0·3 wt %. Orthopyroxene Si was fixed at 1·99 f.u. (formula units), Mn at 0·005, Ca at 0·04, and Ti at 0·001. Orthopyroxene NiO was set equal to 0·2 olivine NiO [equation (2) of Barnes (1986b)]. Al in orthopyroxenewas calculated from the liquid Al2O3 content using an equation derived by fitting a line through the experimental data of Barnes (1986a):
| Alopx = [(Al2O3(liquid) * 0·0148) - 0·14178] - 0·00167. | (A1) |
Values of Al calculated to be <0·01 were set at 0·01. Orthopyroxene Cr2O3 was calculated using equation (4) of Barnes (1986b):
| Cr2O3orthopyroxene = Cr2O3(liquid) * [(20400/T) - 11·67)]. | (A2) |
Temperature T is in degrees Kelvin. Orthopyroxene FeO/MgO was calculated using an exchange coefficient of 0·27 ( Barnes, 1986a). Ferric iron was ignored. Assuming stoichiometry, (Fe + Mg)orthopyroxene = 1 - (Al - 0·01) - Ca - Mn - Cr - Ti. Knowing both (Fe/Mg)orthopyroxene and (Fe + Mg)orthopyroxene, Fe and Mg contents of orthopyroxene can be calculated.
To avoid the problems of non-linearity associated with the equations of DePaolo (1981), coupled assimilation and fractionation was also modelled in a stepwise fashion. Model contaminants were added to the melt, the compositions were normalized to 100%, and then a small fractionation step (0·5%) was calculated, and so on.
Different experimental studies rarely yield the same values for mineral-melt partition coefficients D. The problem probably resides in a combination of kinetic factors, and in the dependence of D values on melt composition and temperature [references given by Maaløe (1995)]. Kinetic effects are important because rapid rates of melting or crystallization prevent full equilibration, causing D to tend towards a value of unity [references given by Bédard (1989)]. Published olivine-liquid D values are poorly constrained because trace elements in olivine are hard to analyse, and being so small, the D values are sensitive to disequilibrium effects. Despite these complexities, there is broad agreement on certain trends. The `conventional' order of incompatibility of trace elements (e.g. Pearce, 1983; Pearce & Parkinson, 1993) reflects a systematic trend in D values, allowing interpolation of some D values. Care must be taken, however, because, for crystal chemical reasons, olivine-liquid and orthopyroxene-liquid D values for high field strength cations such as Zr, Ti and Nb are probably about an order of magnitude higher than those of immediately adjacent REE (Kelemen et al., 1990). The partition coefficients (Table A1) used here were compiled from published experimental data, with poorly constrained values adjusted to yield smooth profiles (Fig. A1).
Figure Table A1. Mineral-liquid partition coefficients
The trace element compositions of equilibrium melts were calculated using the batch melting equation of Hanson (1980):
 |
(A3) |
where CL is the concentration of the element in the melt, C0 is the initial concentration in the system, F is the fraction of melt, D is mineral-melt partition coefficient, [Sigma][phi]D is bulk partition coefficient, and [phi] is normalized weight fraction of the residual phases at a given F. The composition of the solid residue is CS, and is equal to CL[Sigma][phi]D.
Fractional melts and residues were calculated by extracting small increments (0·1% steps) of equilibrium melt from the source mantle, recalculating C0, and then repeating the process. Critical melts were modelled as per the fractional melts, except that one-third of the melt was retained in the source at each melt extraction step.
Preferred compositions of the primitive upper mantle (PUM) and fertile MORB mantle (FMM) compiled from the literature are given in Table A2. FMM is normal N-MORB-source mantle (Pearce & Parkinson, 1993). A primitive mantle mode of 15% Cpx, 25·5% Opx, 57% Ol, and 2·5% Sp was assumed to correspond to PUM (Williamson et al., 1995). Values of the melting mode (Table A3) were adjusted to force disappearance of clinopyroxene after 25% melting (Pearce & Parkinson, 1993). Using these values, and the data of Table A1, the residual mantle data trends shown as fig. 5 of Pearce & Parkinson (1993) were bracketed by fractional and critical melting (assuming one-third melt retention) simulations. The trace element profile of FMM closely resembles the residue of 0·2% melting of primitive mantle (PUM), and so the starting FMM mode was set as the residue of 0·2% melting of the PUM mode.
Table A2. Mantle reservoirs
Table A3. Melting modes
