Enriched crustal and mantle components and the role of the lithosphere in generating...

Lithos 114 (2010) 95–108

Contents lists available at ScienceDirect

Lithos

j ourna l homepage: www.e lsev ie r.com/ locate / l i thos

Enriched crustal and mantle components and the role of the lithosphere ingenerating Paleoproterozoic dyke swarms of the Ungava Peninsula, Canada

Charles Maurice a,⁎, Don Francis b

a Bureau de l'exploration géologique du Québec, Ministère des Ressources naturelles et de la Faune, 400 Boul. Lamaque, bureau 1.02, Val d'Or, Canada, J9P 3L4b Earth and Planetary Sciences, McGill University, 3450 University St., Montréal, Canada H3A 2A7

⁎ Corresponding author. Tel.: +1 819 354 4514x253;E-mail address: [email protected] (C

0024-4937/$ – see front matter © 2009 Elsevier B.V. Aldoi:10.1016/j.lithos.2009.08.002

a b s t r a c t
a r t i c l e i n f o
Article history:Received 16 April 2009Accepted 7 August 2009Available online 21 August 2009

Keywords:PaleoproterozoicMafic dyke swarmsContinental lithosphereAlkaline componentCrustal componentNorth American Craton

Paleoproterozoicmafic dyke swarms (2.5–2.0 Ga) of theUngava Peninsula can be divided in three chemicalgroups. Themain group has a wide range of Fe (10–18 wt.% Fe2O3) and Ti (0.8–2.0 wt.% TiO2) contents, and themost magnesian samples have compositions consistent withmelting of a fertile lherzolitic mantle at ~1.5 GPa.Dykes of a low-LREE (light rare earth element) subgroup (La/Yb ≤4) display decreasing Zr/Nb with increasingLa/Yb ratios and positive εNd2.0 Ga values (+3.9 to +0.2) that trend from primitive mantle towards thecomposition of Paleoproterozoic alkaline rocks. In contrast, dykes of a high-LREE subgroup (La/Yb ≥4) displayincreasing Zr/Nb ratios and negative εNd2.0 Ga values (−2.3 to −6.4) that trend towards the composition ofArchean crust. A low Fe–Ti group has low Fe (<11wt.% Fe2O3), Ti (<0.8wt.% TiO2), high field strength elements(HFSE; <6 ppmNb) and heavy rare earth elements (HREE; <2 ppm Yb) contents, but are enriched in large ionlithophile elements (LILE; K/Ti=0.7–3) and LREE (La/Yb >4). These dykes are interpreted as melts of adepleted harzburgitic mantle that has experienced metasomatic enrichment. A positive correlation of Zr/Nbratio and La/Yb ratio, negative εNd2.0 Ga values (−14 to −6), and the presence of inherited Archean zirconsfurther suggest the incorporation of a crustal component. A high Fe–Ti group has high Fe (>14wt.% Fe2O3) andTi (>1.4wt.% TiO2) contents, alongwith higherNa contents relative to themain group dykes. Dykes of a high-Alsubgroup (>12 wt.% Al2O3) share Fe contents, εNd2.0 Ga values (−2.3 to −3.4), La/Yb and Th/Nb ratios withArchean ferropicrites, and may represent evolved ferropicrite melts. A low-Al subgroup (<12 wt.% Al2O3) hasrelatively lower Yb contents (<2 ppm) and fractionated HREE patterns that indicate the presence of garnet intheir melting residue. A comparison with ~5 GPa experimentally-derivedmelts suggests that these dykesmaybe derived from garnet-bearing pyroxenite or peridotite. The εNd2.0 Ga values (−0.3 to−2.0) of these dykes liebetween the compositions of Archean granitoids and Paleoproterozoic alkaline rocks, signifying theirpetrogenesis involved both crustal and mantle components.

Paleoproterozoic dykes containing a crustal component occur within, or close to, an isotopically enrichedArchean terrane (TDM 4.3–3.1 Ga), whereas dykes without this component occur in an isotopically juvenileterrane (TDM<3.1 Ga). The lack of a crustal component and the positive εNd2.0 Ga values of dykes intruding thelatter suggest that the crust they intruded was either too cold to be assimilated, or that its lower crust and/orlithosphere were Paleoproterozoic in age. In contrast, the ubiquitous presence of a crustal component and thediversity of mantle sources for dykes intruding the enriched terrane (lherzolite, harzburgite, pyroxenite)suggest a warmer crust with underlying heterogeneous lithospheric mantle.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

Proterozoic mafic (dolerite) dyke swarms are common to allArchean cratons and provide probes of the chemical composition andevolution of the Earth's mantle. Their message is, however, compli-cated by uncertainties about the relative roles of asthenosphere andlithosphere in their origin, as well as the effects of crystal fractionationand the assimilation of enriched crustal and/or mantle components(Tarney, 1992; Patchett et al., 1994; Condie, 1997). The magmas ofmany Proterozoic dykes have trace element characteristics that are

fax: +1 819 354 4508.. Maurice).

l rights reserved.

similar to those of continental flood basalts (CFB), with light rare earthelement (LREE) and large ion lithophile element (LILE) enrichments,and Nb–Ta depletions relative to primitive mantle values. Petroge-netic models for the generation of such trace element signaturetypically range between two end-members, i.e. contamination ofasthenospheric magmas by crust, or enriched lithospheric mantlesources (Tarney, 1992 and references therein). It is difficult todistinguish between these end-members on the basis of chemistryalone because the trace element signature of the continental crust(enriched in LILE and REE, but depleted in Nb–Ta) is similar to that ofmetasomatized mantle (Tarney and Weaver, 1987; Boily and Ludden,1991; Seymour and Kumarapeli, 1995; Condie, 1997). Furthermore,radiogenic isotope analyses have not clearly resolved the problem, as

mailto:[email protected]

http://dx.doi.org/10.1016/j.lithos.2009.08.002

http://www.sciencedirect.com/science/journal/00244937

96 C. Maurice, D. Francis / Lithos 114 (2010) 95–108

some studies have concluded that enriched and heterogeneouslithospheric mantle is the source for CFB (Hart, 1985), whereas otherspropose an important role for crustal contamination (Carlson et al.,1981; Peng et al., 1994; Baker et al., 2000).

This paper presents a synthesis of the chemical data acquired onPaleoproterozoic dyke swarms emplaced between 2.5 and 2.0 Ga in

Fig. 1. Schematic distribution of mafic dyke swarms (modified from Buchan et al., 2007; Maur(age for the Mistassini dykes from an internal report by M.A. Hamilton, Jack Satterly Geochrswarms) and c) ca. 2.15–1.88 Ga. Acronyms for alkaline rocks as follows: KL, Kenty Lake; LA,SR, Spanish River. The inset in a) shows the location of Fig. 2 and corresponds to the Ungav

the Ungava Peninsula of Canada (Figs. 1 and 2). We show that the Ndsignature of enrichedmafic dykes is inherited from an Archean crustalcomponent, but that trace element systematics require the existenceof two enriched components, one representing the Archean crust andthe other a metasomatized lithospheric component. We argue thatthese dyke swarms reflect the composition of the continental

ice, 2008) and alkaline rocks of the Superior Province emplaced over a) ca. 2.51–2.44 Gaonology Laboratory); b) ca. 2.23–2.17 Ga (the different dyke colors distinguish distinctLac Aigneau; LC, Lac Castignon; LE, Lac Lemoyne; LL, Lac Leclair; CG, Cargill; BO, Borden;a Peninsula.

Fig. 2. Distribution of the mafic dykes (adapted from Maurice, 2008) and mafic dyke samples of the Ungava Peninsula. The initial dyke compilation map included only dykes withwidths >5m, such that samples that appear not to be linkedwith dykes on this figure were taken from small dykes. The dashed line separates the isotopically juvenile Rivière ArnaudTerrane (RAT) to the East from the Hudson Bay Terrane (HBT) to the West and South (Boily et al., 2009). The white stars indicate the location of Paleoproterozoic alkaline rocks: LA,Lac Aigneau lamprophyres; LE, Lac Lemoyne carbonatite; LL, Lac Leclair lamprophyres; KT, Kenty Lake alkaline basalts. Names of the dated swarms are shown with a white shadeclose to each dated sample. The background figure is constructed with the extrapolation of Nd isotopic data acquired on over 300 Archean rocks of the Northeastern SuperiorProvince (see compilation by Maurice, 2007). Each sample was given a value between 1 and 6 that reflects the extent to which an older evolved crust was recycled. Samples with147Sm/144Nd ratios >0.14 (mainly mafic and ultramafic rocks) were classified with respect of the εNd(t) notation, similarly to that reported in Tomlinson et al. (2004). Samples with147Sm/144Nd ratios <0.14 (mainly felsic rocks) were classified with respect of their model ages (TDM). As a result of the interpolation, the darker shades outline areas with lowerεNd(t) or higher TDM values. Symbols are as defined in Table 1.

97C. Maurice, D. Francis / Lithos 114 (2010) 95–108


lithospheric mantle in the Paleoproterozoic and investigate their usein probing its evolution.

2. Geological framework

Our study focuses on the Ungava Peninsula, that comprises rocksof Precambrian age. Its center includes rocks of the ArcheanNortheastern Superior Province (NESP) that are intruded by numer-ous Paleoproterozoic mafic dykes. The NESP is surrounded byPaleoproterozoic supracrustal belts emplaced during the Trans-Hudson Orogeny (Lewry and Collerson, 1990), ca. 1.8 Ga.

2.1. Northeastern Superior Province (NESP)

The NESP comprises dominantly Neoarchean plutonic suites inwhich small amphibolite to granulite-grade greenstone belts occur(Maurice et al., 2009 and references therein). It is separated into twoisotopically distinct regional terranes (Fig. 2) bordered bymigmatizedpelites to semi-pelites of the Archean Lake Minto metasedimentarybasin (Simard, 2008). To the Northeast, the Rivière Arnaud Terranegroups Archean rocks having juvenile isotopic signatures (T-DM<3.1 Ga), while to the West and South, the Hudson Bay Terranepreserves the remnants of a reworked Meso- to Paleoarchean craton,with Ndmodel ages up to 4.3 Ga (O'Neil et al., 2008; Boily et al., 2009).

2.2. Volcano-sedimentary Circum-Ungava belts

The Paleoproterozoic supracrustal belts surrounding the Archeanrocks of the Ungava Peninsula range in age from 2.17 to 1.88 Ga (seethe review inMaurice et al., 2009 for a complete reference list of U–Pbages). These supracrustal belts include the Labrador Trough to theEast, the Cape Smith Foldbelt to the North, and the Belcher and OttawaIslands and associated coastal continental basalts (the Richmond Gulfand Nastapoka groups) to the West (Fig. 2).

The supracrustal belts host a variety of alkaline rocks that includealkaline basalts near Kenty Lake (Gaonac'h et al., 1992; Modelandet al., 2003), lamprophyres near Lac Leclair (Baragar et al., 2001) andLac Castignon, and carbonatites near Lac Lemoyne (Wright et al.,1999). While the Lac Leclair lamprophyres must have ages >2.04 Ga,all the other alkaline occurrences were emplaced ca. 1.96–1.88 Ga, inthe same time frame as the 1.94 Ga Lac Aigneau mafic to ultramaficlamprophyre dykes that intrude the Archean basement of the easternUngava Peninsula (Maurice et al., 2009).

2.3. Mafic dyke swarms

Numerous mafic dyke swarms ranging in age between 2.51 and2.00 Ga (Buchan et al., 1998; Maurice et al., 2009) intrude the Archeanbasement (Fig. 1). The oldest ages were obtained on the small northtrending Irsuaq (2508±6Ma) and the NE trending Ptarmigan (2505+2/−1 Ma; Fig. 1a) swarms.

These two early swarms are separated by 280 Ma from the manydykes emplaced 2.23–2.17 Ga in the northern Ungava Peninsula. TheWNW trending Klotz dykes (2209±1 Ma; Fig. 1b) represent avoluminousmagmatic event that emplaced large individual dykes (upto 100 m in thickness) that are traceable over strike lengths greaterthan 250 km. The slightly older NW trending Anuc dykes (2220±3 Ma) appear to define a coherent swarm having a distinct trend thatcontrasts with the Klotz dykes. The few WSW trending Kogaluk Bay(2212±3 Ma; Fig. 1b) and Couture (2199±5 Ma) dykes also occurwith the Klotz dykes, and are closer to them in age. Although notprecisely dated, theWNW trendingMaguire dykes in the center of theUngava Peninsula (2229+35/−20 Ma) may be coeval with the Anuc,Kogaluk Bay and Klotz dykes. On the basis of U–Pb ages, Buchan et al.(1998) suggested that theMaguire, Klotz, and Senneterre dykes (Fig. 1b)defined a radiating dyke swarm whose magma spread laterally from a

focus above a mantle plume south of Ungava Bay. This model hashowever been challenged on the basis of a compilation of more recentdata on these dykes and geochemical differences between the swarms(Maurice et al., 2009).

The Payne River dykes (ca. 2.17 Ga) in the northern portion of theUngava Peninsula comprise a dense population of NNW trendingdykes (Fig. 2). The southern dykes of the Ungava Peninsula haverelatively younger ages, with the Rivière du Gué dykes (2149±3 Ma;Fig. 1c) being the oldest. Although no age data is available in thesouthwestern Ungava Peninsula, Roy et al. (2004) suggested that theN to NNW trending dykes therein could belong to the Lac Espritswarm recognized further to the South (2069±1 Ma). Finally, theWNW trending Minto dykes (1998±1 Ma) have the youngest U–Pbage. The Inukjuak swarm is undated, but is believed to be 2.0–1.9 Ga(Buchan and Ernst, 2004). Overall, the time and distribution of 2.2–1.9 Ga mafic dykes of the Ungava Peninsula record an evolution fromolder dykes which occur mainly in the Rivière Arnaud Terrane (2.23–2.17 Ga) to younger mafic and alkaline dykes that occur in the HudsonBay Terrane (2.15–1.94 Ga).

Most Paleoproterozoic mafic dykes of the Ungava Peninsula arefine to coarse grained gabbros dominated by an assemblage ofclinopyroxene and plagioclase. Gabbronoritic compositions are onlydocumented in the oldest 2.5 Ga dykes (Irsuaq and Ptarmiganswarms, Buchan et al., 1998; Maurice et al., 2009). Igneous texturesand mineral assemblages are commonly well preserved, withpyroxenes being variously replaced by amphiboles, and plagioclaseby sericite. Rare olivine crystals are commonly altered to talc, chlorite,iddingsite and serpentine. Small proportions of Fe–Ti oxides, titanite,zircon and baddeleyite complete the mineral assemblage.

3. Data

3.1. Source

Most of the data presented in this paper were obtained during ageological mapping program conducted between 1998 and 2003 bythe Ministère des Ressources naturelles et de la Faune of Québec(Simard, 2008 and references therein). Of the 209 bulk rock analysesdiscussed in this paper (see supplementary data set), 176 wereobtained from the Système d'Information Géominière (SIGEOM)database, 25 were extracted from the geochemical database of theGeological Survey of Canada (project #TGI010004) and 8 wereprovided by J. Bédard (GSC-Québec). Although these data wereanalyzed in a number of different laboratories, all of SIGEOM analyses(85%) were performed at ACME Analytical Laboratories (Vancouver).Nd isotopic analyses were carried out on 31 of the mafic dykes and8 Lac Aigneau lamprophyric dykes and reported in Maurice et al.(2009).

3.2. Sampling rationale

Comparing the chemical composition of samples acquired bymany geologists has a number of problems. The most obvious is theuneven distribution of dyke samples (Fig. 2). While the lack ofgeochemical data in the center of the Peninsula is due to a lack ofdykes (Maurice, 2008), the small number of analyses from the Southand Southeast reflects the small number of samples collected, ratherthan low dyke abundance.

Another problem involves the nature of the samples collectedwithin each dyke. While judicious sample selection by visualinspection (e.g. avoiding secondary alteration and heterogeneity) isalways possible, our database contains no information on the locationof samples with respect to their dyke margins. The position of asample within a dyke can be important, as the relative compositionsof dyke margins and interiors may in part reflect the emplacementprocess(es). In the Mackenzie dyke swarm of northern Canada for


example, thin dykes commonly appear homogeneous, but dykeshavingwidths greater than 30m aremarkedly zoned, with the centerstypically being more evolved in composition than the margins(Gibson et al., 1987). In contrast, the Harp dykes of Labrador arecharacterized by centers that are substantially more magnesian thantheir margins (Cadman et al., 1994). Because both flow banding andhydrous alteration can be more prevalent in dyke interiors (South-wick and Day, 1983), some workers prefer to sample dyke margins(e.g. Southwick and Halls, 1987), a choice that can, however, lead tosamples that have been contaminated by the host rock (Dostal andFratta, 1977; Gill and Bridgwater, 1979). More than 80% of thesamples in our database have dyke widths that were visuallyestimated in the field. Despite a three orders magnitude range indyke widths, there appears to be no correlation between width andchemical indicators of crustal contamination such as K2O/TiO2 and La/Yb (Fig. 3). This suggests that larger dykes have not assimilated morecrust than smaller ones during their emplacement, and thatcomparisons can be made in spite of a lack of knowledge of thelocation of samples within individual dykes.

The interpretation of the chemistry of Precambrian rocks is neverstraightforward because of the post-magmatic processes they arelikely to have experienced. Fifteen dyke samples with anomalouslylow CaO (<6 wt.%) and high LOI (>4 wt.%) were interpreted to beheavily altered and omitted from our dataset.

4. Compositional groups

Dykeswith awide spectrumof ages canbe foundwithin apparentlysingle dyke swarms (French et al., 2004; Jourdan et al., 2004;Mège and

Fig. 3. Crustal contamination indicators a) K/Ti, and b) La/Yb vs. dyke thicknesses fordykes of the Ungava Peninsula. Symbols as defined in Table 1.

Korme, 2004; Jourdan et al., 2006), while in other cases dykes withdistinct trends can belong to the same swarm (Buchan et al., 2007).The attribution of magmatic ages to undated samples on the basis ofswarm affiliation is thus hazardous. Furthermore, although recentefforts have been made to acquire more U–Pb ages on dykes of theNESP (Buchan et al., 1998; Maurice et al., 2009), only 15% of thesamples contained in our database can be confidently attributedmagmatic ages. Our approach consists of grouping dyke samples onthe basis of their geochemical characteristics, in a manner unpreju-diced by apparent dyke swarm affiliation. The dyke analyses aredivided hereafter into three chemically distinct groups: a maingroup, a low Fe–Ti group, and a high Fe–Ti group (Table 1) on thebasis of the relative concentrations of the major elements: Fe, Ti, Mg,Al, and Na.

4.1. Main group

The main group comprises 65% of the dyke analyses contained inthe database and have a large range of Fe (10–18 wt.% Fe2O3) and Ti(0.8–2.0 wt.% TiO2) contents (Fig. 4). Samples of the main group withlowMg overlap those of the high Fe–Ti group in terms of Fe and Ti, butcan be distinguished on the basis of their lower Na (Fig. 5a), and/orhigher V/Ti ratios (30–50; Fig. 6). The most magnesian samples of themain group have Mg, Si and Al contents similar to those of ~12 wt.%MgO experimental melts of fertile lherzolite (HK-66 and KLB-1) atpressures ~1.5 GPa (Figs. 4c and 5d). The increase in Fe and Ticontents with decreasing Mg in the quartz-normative dykes of themain group is similar to the liquid lines of descent of experimentaltholeiitic liquids undergoing crystal fractionation between 1 atm and1 GPa (Figs. 4 and 7). The decrease in Al content with increasing Si inthe main group dykes is, however, less than that produced byfractionation at 1 GPa, and best matches trends produced at pressuresbetween 1 atm and 0.7 GPa (Figs. 4c and 5c).

Dykes of the main group have a wide range of high field strength(HFSE; 25–300 ppm Zr, 2–30 ppm Nb) and heavy rare earth elements(HREE; 1–6 ppm Yb), and can be further divided into two subgroupson the basis of their La/Yb and Zr/Nb ratios (Figs. 8 and 9). The low-LREE main group dykes have less fractionated REE profiles, and thuslower La/Yb ratios (≤4; Fig. 9d). They have Zr/Nb, Th/Nb, and K/Tiratios that overlap those of the high-LREE dykes, but do not extend tothe higher values of the latter (Fig. 10). The least enriched samples ofthis subgroup have trace element ratios that plot near those ofprimitivemantle (PM; Figs. 10 and 11), but their Zr/Nb ratios decreasewith increasing La/Yb towards the compositions of Paleoproterozoicalkalinemagmas of the Ungava Peninsula. Eight of the low-LREE dykesyielded positive εNd2.0 Ga values ranging between +3.9 and +0.2(Fig. 10d). The low-LREE dykes occur largely in the northeasternportion of the Ungava Peninsula, in the Rivière Arnaud Terrane, asdykes of the Klotz (2.21 Ga), Couture (2.20 Ga) and Payne River(2.17 Ga) swarms, but are also found in few scattered localitieselsewhere (Fig. 2).

The high-LREE main group dykes have similar HREE, but elevatedLa/Yb ratios (≥4) compared to the low-LREE main group dykes(Figs. 8 and 9). These enriched dykes have La/Yb ratios that are similarto those of the low Fe–Ti dyke group, but most have distinctly lowerZr/Nb, Th/Nb and K/Ti ratios (Fig. 10). In a La–Yb–Nb diagram, thehigh-LREE main group dykes scatter from a position close to that ofprimitive mantle towards the composition of the Archean granitoidsof the Ungava Peninsula (Fig. 11). The five high-LREE dykes analyzedfor Nd isotopes are isotopically enriched, with negative εNd2.0 Ga

values ranging between −2.3 and −6.4 (Fig. 10d). These dykesinclude a dated dyke from the Rivière du Gué swarm (2.15 Ga; Fig. 1),along withmany other dykes of unknown age that occurmainly in thewestern and southern parts of the Ungava Peninsula, in the HudsonBay Terrane (Fig. 2).

Table 1Synoptic table of symbols, characteristics, possible components and sources for geochem-ical groups of mafic dyke swarms of the Ungava Peninsula.


4.2. Low Fe–Ti group

Dykes in our database having distinctly lower Fe (<11 wt.% Fe2O3)and Ti contents (<0.8 wt.% TiO2; Fig. 4) relative to the main groupdykes are grouped into a low Fe–Ti group. They display a slightincrease in Fe, Ti and Al with decreasing Mg, are characterized bysystematically higher Mg at any Si content (Fig. 5b), low Fe similar toharzburgite experimental melts (Figs. 4a and 5c), and low V and Ticontents (Fig. 6). The low Fe–Ti dykes are further characterized by low

Fig. 4. a) Fe, b) Ti, c) Al and d) Ca/Al vs. Mg in cation units for themafic dykes of the Ungava Pe9 and 4 wt.% MgO. Crystal fractionation experiments conducted at 0.7 and 1.0 GPa (Villiger ethe 1.5 GPa KLB-1 lherzolitic residuum of sample #19 from Hirose and Kushiro (1993). Expmaterial. The positions of the 0.7 and 1.0 GPa vectors are offset+1.5 cat. Fe and+0.5 cat. Ti ostarting materials used in the experiments. White numbers are the pressures to which experproduced (1+, 1.0 and 1.5 GPa; 2+, 2.0 and 2.5 GPa; 3, 3 GPa; 4+, 4.0 to 4.6 GPa; 5+, 5.0 to2.5 GPa; 5, 5 GPa) from melting garnet pyroxenite MIX1G (Hirschmann et al., 2003; Kogiharzburgite (Falloon and Danyushevsky, 2000). Trend 1 is produced by the low pressure fractiocpx. The data of Archean Western Superior ferropicrites are from Goldstein and Francis (2008

HFSE (25–125 ppm Zr; <5 ppm Nb) and HREE contents (<2 ppm Yb),but are relatively enriched large ion lithophile element (LILE; 1–5 ppmTh) contents at any Zr content (Fig. 9). They have pronounced negativeNb–Ta anomalies in primitivemantle-normalized diagrams (Fig. 8), andhigh ratios of incompatible trace elements (Th/Nb, K/Ti, Zr/Nb) thatscatter with those of the high-LREE main group dykes towards those ofArchean granitoids (Fig. 10). Four dykes within this chemical groupyielded strongly negativewhole rock εNd2.0 Ga values (−6 to−14) thatapproach those of Archean granitoids (Fig. 10d). The dated dykes of theIrsuaq (2.51 Ga) and Maguire (2.23 Ga) swarms (Fig. 1), that occuralong the boundary of the two isotopic Archean terranes, belong tothe low Fe–Ti group (Fig. 2), but this group also includes other dykesof unknown age found mainly in the Hudson Bay Terrane.

4.3. High Fe–Ti group

Samples of the high Fe–Ti group are characterized by high Fe(>14 wt.% Fe2O3) and Ti (>1.4 wt.% TiO2) contents (Fig. 4) and havesystematically higher Na (Fig. 5a), and/or lower V/Ti ratios (12–30;Fig. 6) relative to the main group dykes. Most of these dykes are ol-normative basalts (Fig. 7), but are not alkaline sensu stricto (i.e.nepheline-normative). They have, however, high Na (Fig. 5a) andincompatible trace element contents (Figs. 8 and 9; Zr>100 ppm),and low V/Ti ratios which approach those of alkaline rocks of the

ninsula. The dotted arrows show the vectors produced by experimental liquids betweent al., 2004, 2007) used a starting material in equilibrium with a composition identical toeriments conducted at 1 atm (Thy et al., 1998) used natural basalt #11R4 as a startingwing to the higher Fe and Ti contents of the Ungava Peninsula mafic dykes relative to theimental melts from fertile lherzolites KLB-1 and HK-66 (Hirose and Kushiro, 1993) were7.0 GPa). Black numbers are the pressures to which experimental liquids (2+, 2.0 andso et al., 2003) were produced. HZ are the 1.5 GPa experimental melts of a syntheticnation of 35% ol+65% cpx, while trend 2 is produced by the fractionation of 70% ol+30%).

Fig. 5. a) Na, b) Mg, c) Fe and c) Al vs. Si in cation units for the mafic dykes of the Ungava Peninsula. Samples of the low Fe–Ti group are not plotted for clarity in panel a). High Fe–Tidykes overlapping with the Na contents of main dykes have lower V/Ti ratios and higher Fe contents. Conversely, dykes of the main group plotting at high Na have higher V/Ti ratiosand lower Fe contents. The dotted arrows in panel c) show the vectors produced by experimental liquids between 9 and 4 wt.% MgO, as explained in Fig. 4. The significance of blackand white numbers and HZ are as explained in Fig. 4. Other symbols as defined in Table 1, LH, lherzolite; PX, pyroxenite.

Fig. 6. V vs. Ti in ppm for the mafic dykes and alkaline rocks of the NESP. Source of datafor the alkaline rocks as follows: Kenty Lake; Gaonac'h et al. (1992) and Modeland et al.(2003), Lac Leclair; Baragar et al. (2001), Lac Aigneau (SIGEOM database), and LacCastignon (unpublished data). Mafic dyke symbols as defined in Table 1.


Ungava Peninsula (Fig. 6). The high Fe–Ti group dykes can be dividedinto two subgroups on the basis of their Al contents.

The high-Al subgroup includes dykes with >12 wt.% Al2O3 whoseAl contents decrease with Si (Fig. 5d). They display the lowest Si, buthighest Fe contents, of all the Paleoproterozoic dykes of the UngavaPeninsula, and are similar to Archean ferropicrites of the WesternSuperior Province (FP; Figs. 4a and 5c). The most primitive samples ofthis subgroup have Al contents similar to experimental melts of fertilelherzolite and pyroxenite produced at ~2.0 GPa (Fig. 5d), but theirrelatively low Si contents at any Mg (Fig. 5b) more closely resemblepyroxenite experimental melts, whereas their high Fe contentsresemble the ferropicrites (Fig. 4a). Most of the high-Al subgroupdykes are olivine-normative with low normative diopside contents,and scatter from the diopside–olivine join towards the diopside–hypersthene join (Fig. 7). The high-Al subgroup dykes have high HFSE(>100 ppm Zr; >7 ppm Nb) and HREE (≥2 ppm Yb) contents and La/Yb ratios that are similar to the most enriched high-LREE main groupdykes (Fig. 9). They differ from the high-LREE main group dykes inhaving lower Th/Nb and Zr/Nb ratios similar to those of Archeanferropicrites (Fig. 10). Four of these samples yielded negative εNd2.0-

Ga values that range between−2.3 and−3.4, while a fifth sample hasa positive value of 3.1 (Fig. 10d). The dated dyke of the Minto swarm(2.00 Ga; Fig. 1) belongs to the high-Al subgroup, along with manydykes of unknown age from the Hudson Bay Terrane (Fig. 2).

The low-Al subgroup dykes have <12 wt.% Al2O3 (Fig. 4c) anddisplay increasing Al with increasing Si, but decreasing Mg content, incontrast to the high-Al subgroup (Figs. 4c and 5d). Two distinct liquid

lines of descent appear to be defined by the variations in their Ca/Alratios and Mg contents (Fig. 4d). The increase in Al content with Si inthese trends diverge from olivine control lines and are better matchedby fractionation trends involving clinopyroxene varying in propor-tions between 65% (trend 1) and 30% (trend 2). The low Al and high Fecontents of this subgroup distinguishes them from all other dykes, and

Fig. 7. Ternary projection of CIPW (cation) normative minerals in Ol-Di-Hy, Hy-Di-Qz and Ol-Ne-Di planes. The signification of arrows and squares are as explained in Figs. 4 and 5,but MgO contents decrease from 13 to 4 wt.% MgO in the crystallization experiments. Symbols as defined in Table 1.


their most magnesian samples have low Al (Fig. 4c) and high Ca/Alratios (Fig. 4d) that resemble experimental melts of garnet pyroxeniteand/or fertile lherzolite produced at 5 GPa. Most of the low-Al dykeshave olivine-normative compositions that are high in normativediopside, and appear to scatter from the high pressure experimentalmelts of garnet pyroxenite towards the diopside–hypersthene join(Fig. 7). The low-Al subgroup dykes typically display higher Thcontents than the high-Al subgroup, which gives them higher Th/Nbratios (>0.2; Figs. 9 and 10). Although their Zr/Nb ratios overlap thoseof the high-Al dykes, many have higher La/Yb ratios (Fig. 10) becauseof their distinctly more fractionated HREE patterns (Fig. 8) and lowerYb contents (Yb ≤2 ppm; Fig. 9). Six of the low-Al subgroup dykesyielded negative εNd2.0 Ga values ranging between −0.3 and −2.0and are displaced from the array of other dykes towards the KentyLake alkaline basalts and Lac Aigneau lamprophyres (Fig. 10d). Thedated dykes of the Anuc (2.22 Ga) and Kogaluk Bay (2.21 Ga) swarmsbelong to the low-Al subgroup (Fig. 1), together with many dykes of

Fig. 8.Multi-element diagram normalized over primitive mantle (Sun andMcDonough,1989) for representative mafic dyke samples with ~8 wt.% MgO of the five chemicalgroups and subgroups of the Ungava Peninsula. Symbols as defined in Table 1. Samplenumbers as follows: low-LREE main group, 1999022816; high-LREE main group,2000025651; low Fe–Ti group, 2001035385; low-Al high Fe–Ti group, 1999020157;high-Al high Fe–Ti group 2001035397.

unknown age mainly from the western and southern portions of theUngava Peninsula, in the Hudson Bay Terrane (Fig. 2).

4.4. Geographical occurrence of mafic dykes and the nature of theUngava Peninsula

The mafic dykes intruding the Rivière Arnaud Terrane belongdominantly to the low-LREE main group. In distinct contrast, virtuallyall the other dyke groups occur within the older Hudson Bay Terrane,or along the contact between the two terranes. In addition to thisterrane dependence, all dykes emplaced in the 2.23–2.17 Ga periodare spatially associated with a decrease in lithospheric thicknesstowards the North of the Ungava Peninsula (Fig. 2). Younger dykesemplaced in the 2.15–2.00 Ga period are underlain by a thickerlithosphere. Although few data exist for dykes in the crust overlyingthe thickest lithosphere to the Southeast (>200 km), most samples inthis area are low-Al high Fe–Ti group dykes spatially associated withthe Lac Aigneau lamprophyre dykes.

5. Discussion

Mafic dykes are documented in a variety of geological settings,with parallel linear swarms commonly being associated with failedrifts or the breakup of continental margins (Fahrig, 1987; Ernst andBuchan, 2001). The tectonic significance of giant radiating swarms is,however, model dependent. The fanning that is characteristic of theseswarms can be related to lithospheric stress regimes associated withplate tectonics (McHone et al., 2005) or to the rise of deep mantleplumes (Ernst et al., 1995; Ernst and Buchan, 2001). The tendency ofthese swarms to radiate from a focal point may imply the lateraltransport of magmas from a central plume source, but direct evidencefor flow direction is disputed. Studies of the anisotropy of magneticsusceptibility (AMS) hold lateral flow in some dyke swarms (Ernstand Baragar, 1992; Ernst, 1994), but the original AMS fabrics of otherswarms have been shown to be vertical, before being replaced by latehorizontal fabrics (Cadman et al., 1992). Such constraints may implythat crustal fractures are initially filled vertically, but then magmapropagates laterally (Tarney, 1992). The many rift zones surroundingthe Ungava Peninsula and the older ages of most mafic dykes relativeto the supracrustal rocks (Maurice et al., 2009) suggest that a largenumber of dykes may be associated with the early breakup of the

Fig. 9. a) Nb, b) Th, c) Yb and d) La/Yb vs. Zr for mafic dykes of the Ungava Peninsula. Symbols as defined in Table 1. The flat and the sloped arrows in panel d) are the products ofcrystal fraction and assimilation fractional crystallization, respectively. The bulk partition coefficients for Zr, La, and Yb are quantified with a gabbroic crystal fractionation from amodel involving olivine (ol), clinopyroxene (cpx), and plagioclase (plag) fractionating in proportions close to the low pressure gabbroic cotectic (20% ol, 30% cpx and 50% plag). Thismodel reproduces the Mg, Ti, and Fe variations of much of the main group dykes. The ol and plag partition coefficients, along with cpxKdZr, are from the experiments from a tholeiiticmafic magma reported in Fujimaki et al. (1984). cpxKdLa (0.042) and cpxKdYb (0.56) were calculated from a predictive model for REE partitioning between clinopyroxene andanhydrous silicate melts (Wood and Blundy, 1997) with sample 1999022816 (8 wt.% MgO, and 1 wt.% TiO2). Bulk partition coefficients are as follows: DLa=0.03, Dzr=0.045 andDYb=0.18. The assimilation factor (r) in the assimilation fractional crystallization model (AFC) was set to 0.05 (r=dMa/dMfc, where dMa is the assimilation rate of wallrock anddMfc is the crystallization rate at which fractionating phases are removed). The enriched contaminant is an average granite with La/Yb=55 and Zr=165 ppm. The position of thedividing line between the low- and high-LREE main group dykes follows a slope similar to that of the AFC model.


lithosphere. Although some dykes may belong to radiating swarmswith magma sources located off the cratonic margins, we assume inthe following discussion that dykes record the regional nature of theimmediately underlying mantle.

Another debate concerning the origin of continental dyke swarms(and continental flood basalts) centers on whether their traceelement enrichment reflects crustal contamination or an enrichedmantle source (Collerson and Sheraton, 1986; Patchett et al., 1994;Hergt and Brauns, 2001). Two broad models have emerged to explainthe incompatible element and isotopic features. The first involves twodistinct source regions, with asthenosphere-derived magmas inter-acting with either the continental crust (Arndt et al., 1993), or smalldegree partial melts of the subcontinental lithosphere (McKenzie,1989; Ellam and Cox, 1991; Cadman et al., 1995). The second class ofmodel holds that these chemical characteristics are inherited from alithospheric mantle that had previously been enriched while beneathcontinents (Hawkesworth et al., 1984; Hergt and Brauns, 2001).

The differences between chemical groups of mafic dykes of theUngava Peninsula may reflect a number of different processes,including crystal fractionation, different degrees of partial melting,mixing between mantle reservoirs, melting of heterogeneous mantlesources, crustal contamination, metasomatic source enrichment, or acombination of these processes (Collerson and Sheraton, 1986;Tarney, 1992; Patchett et al., 1994; Seymour and Kumarapeli, 1995;Condie, 1997). Because the compositional changes produced by low

pressure crystallization obscure the primarymagmatic composition ofmagmatic suites (Cox, 1980; Grove and Baker, 1984; Klein andLangmuir, 1987), and it is necessary to consider the effects ofdifferentiation on the compositional arrays of dykes.

5.1. Main group

The low Mg contents (Fig. 4) and quartz-normative compositions(Fig. 7) of most main group dykes imply that their magmas haveevolved since leaving the mantle. The increasing Fe and Ti withdecreasing Mg contents of the low-LREE and high-LREE main groupdykes (Fig. 4) is typical of tholeiitic magmatic suites. Most dykes areconsistent with experimental liquid lines of descent of tholeiiticbasalts fractionating between 1.0 GPa and 1 atm (Figs. 4, 5 and 7), butthe lower Al with decreasing Mg (Fig. 4a) and increasing Si (Fig. 5d)contents relative to 1.0 Gpa experimental liquids suggests that mostfractionated at intermediate to shallow crustal levels between 0.7 GPaand 1 atm. A gabbroic crystal fractionation model involving olivine(20%), clinopyroxene (30%), and plagioclase (50%) reproduces thetrend defined by the 1 atm experimental liquids. It indicates that therange of observed main group dyke compositions corresponds to~65% fractionation of a parental magma with 9 wt.% MgO, whosecumulate assemblages may lie hidden in the crust. A fractionationmodel cannot, however, reproduce the increase in La/Yb ratios and Zrcontents within each of themain group dyke subgroups (Fig. 9), much

Fig. 10. a) Th/Nb, b) K/Ti, c) Zr/Nb and d) εNd2.0 Ga vs. La/Yb for Paleoproterozoic mafic dykes and alkaline rocks and Archean granitoids of the Ungava Peninsula. Mafic dyke symbolsas defined in Table 1, alkaline rocks as in Fig. 6. White diamonds are for Archean tonalites and trondhjemites (TT is the average of 53 samples), while grey diamonds are for granitesand granodiorites (GG is the average of 177 samples). Data for Archean granitoids are from Percival andMortensen (2002), the SIGEOM database, and unpublished data by J. Bédard.Panel c) shows melting trajectories obtained using a nonmodal batch melting model by Aldanmaz et al. (2006). Melting curves are for spinel–lherzolite and garnet–lherzolite havingbulk compositions similar to that of primitive mantle. The Nd isotopic data of panel d) for mafic dykes and Lac Aigneau alkaline lamprophyres are fromMaurice et al. (2009) and datafor Kenty Lake from Gaonac'h et al. (1992). The source of data for Archean granitoids is compiled in Maurice (2007).


less the La/Yb differences between the two subgroups. Although acombined assimilation-fractional crystallization (AFC) process with asmall proportion (r=0.05) of a La/Yb enriched component canproduce the increase in La/Yb ratios within each subgroup, no

Fig. 11. La-Yb-Nb ternary diagram formafic dykes, alkaline rocks andArchean granitoidsof the Ungava Peninsula. Symbols as defined in Table 1, Figs. 6 and 10 respectively.

reasonable r factor can produce the high-LREE dykes from the low-LREE dykes, and these subgroups must therefore have distinct origins.

A comparison of the trends of olivine fractionation for fertile-mantle experimental melts to the few high magnesian samples of themain group dykes suggests parental magmas have either been derivedfrom a mantle similar to KLB-1 (Mg#=90) at ~3 GPa, or a more Fe-rich mantle similar to HK-66 (Mg#=85) at ~1.5 GPa (Fig. 4a).However, these high magnesian samples exhibit the high Si contentsexpected for melting under pressures of ~1.5 GPa (Fig. 5c and d), andtheir relatively high Fe contents would rather be consistent with themain group dykes being sourced from a Fe-rich fertile lherzolitesimilar to HK-66. Because the minimum pressure for garnet stabilityin peridotites is >2.2 GPa (Kinzler, 1997;Walter, 1998; Longhi, 2002),the lack of HREE depletion of the main group dykes (Figs. 8 and 9) isfurther consistent with melting a Fe-rich lherzolite at ~1.5 GPa.

Partial melting of a peridotitic mantle source produces negativeslopes in the Zr/Nb vs. La/Yb space, with a garnet-bearing assemblagedefining a flatter slope than a spinel-bearing one (Fig. 10c). Althoughdykes of the low-LREE subgroup, and to a lesser extent the high-LREEsubgroup, respectively fall along the melting arrays predicted forspinel- and garnet-lherzolite having primitive mantle compositions,their similar range in Al (Figs. 4 and 5) and HREE (Yb; Fig. 9) contentsmakes it unlikely that the distinct La/Yb ratios reflect the presence orabsence of garnet in their source. The rise in the La/Yb ratio with Zr ofthe low-LREE dykes (Fig. 9) along with their trend towards thePaleoproterozoic alkaline magmas of the Ungava Peninsula (Figs. 10cand 11), rather suggest the possible involvement of an alkalinemantlecomponent. In contrast, the high-LREE dykes appear to have


incorporated Archean granitoids of the Ungava Peninsula, a featureconsistent with their low εNd2.0 Ga values (Fig. 10d).

5.2. Low Fe–Ti group

The low Fe, but high Mg and Si contents of the low Fe–Ti groupdykes with respect to the other dyke groups of the Ungava Peninsulaare most similar to experimental melts of depleted harzburgite (HZ;Figs. 4a, 5b and c). A depleted refractory source for the low Fe–Tidykes is also consistent with their low HREE and HFSE contents(Fig. 9). Most of these dykes have high Mg contents and the limitedincrease in Fe and Ti is inconsistent with low pressure liquid lines ofdescent for magmas derived from a fertile source (Fig. 4). This featureis similar to trends defined by calc-alkaline volcanic suites that havefractionated under hydrous conditions, but contamination by evolvedfelsic material yields an essentially identical signature, and thedistinction between these two petrogenetic processes is difficult.

The higher relativemagmatic temperatures of many of theMg-richdykes in the low Fe–Ti groupmean that they had a greater potential toassimilate contaminants. The displacement of the low Fe–Ti dykestowards the composition of the ArcheanUngava crust (Figs. 10 and 11)is coupledwith strongly enriched Nd isotopic compositions approach-ing those of the Archean granitoids (Fig. 10d). This, along with thepresence of inherited Archean-age zircons in an Irsuaq dyke belongingto this group (Maurice et al., 2009), suggests that the trace elementenrichments of the low Fe–Ti dykes are best explained by theincorporation of an Archean crustal component. The LILE enrichmentsof the low Fe–Ti dykes compared to the high-LREE main dykes (highTh/Nb and K/Ti ratios, Fig. 10) may also imply that the depletedharzburgite source for the low Fe–Ti dykes was metasomatically re-fertilized by fluids prior or during melting. The geographicalassociation of many low Fe–Ti dykes with the boundary of the twoArchean isotopic terranes (e.g. the Irsuaq and Maguire dykes; Fig. 2)may indicate that such a harzburgitic mantle wedge, depleted by meltextraction in the Archean, was sandwiched between the two terranes.The existence of low Fe–Ti dykes that are not associated with knowntectonic boundaries, however, requires that depleted mantle domainsalso exist peppered beneath the Hudson Bay Terrane (Fig. 2).

5.3. High Fe–Ti group

The parental magmas of the high Fe–Ti group dykes lie close to theolivine–diopside join and give insights on the conditions under whichthey fractionated. Crystallization experiments on mildly alkalinebasaltic liquids (3–5% normative nepheline) indicate that liquidlines of descent are a function of pressure (Mahood and Baker, 1986;Thy, 1991). During the initial stages of crystallization at 1 atm,experimental liquids remain near the thermal crest defined by theolivine–plagioclase–clinopyroxene plane, but with higher degrees ofcrystallization move away from the divide towards quartz-normativecompositions. At pressures above 0.8 GPa, however, alkaline liquidsare driven towards increasingly ne-normative compositions. Thesesystematics indicate that mildly alkaline basaltic liquids lying close tothe thermal divide would fractionate at low pressure to produceresidual compositions similar to those of the most evolved high Fe–Tigroup dykes (Fig. 7).

The alkaline affinities of the high Fe–Ti group dykes may furtherimply a range of possible mantle sources. Small-degree partial melts ofperidotite at high pressure can be alkaline (Takahashi and Kushiro,1983), especially if generated in the presence of carbonate (Hirose,1997). However, partial melting of lower crustal eclogite or pyroxeniteplus peridotite mixtures (Kogiso et al., 1998; Yaxley and Green, 1998),or pyroxenite itself (Hirschmann et al., 2003; Kogiso et al., 2004), alsoyield alkalinemelts. Pyroxenite is aminor, but ubiquitous, component inthe mantle (Hirschmann and Stolper, 1996). Because the range ofpyroxenite compositions is rather large, a wide spectrum of composi-

tions may modify the predominantly peridotitic mantle melts. Increas-ing the pressure of melting of peridotite increases the Fe content(Langmuir and Hanson, 1980), but decreases the Si content of initialmelts (Green, 1973; Takahashi and Kushiro, 1983; Hirose and Kushiro,1993), such that Si and Fe are negatively correlated with increasingpressure (Fig. 5c). In contrast, melting experiments of silica-deficientpyroxenite produces more Si-rich liquids with increasing pressure.Furthermore, the expansion of the phase volume of garnet relative toclinopyroxenewith increasing pressure yields liquidswith higher Ca/Al,but lower MgO, compared with garnet peridotite-derived partial melts(Fig. 4d, Kogiso et al., 2004).

Despite their similar alkaline affinities, the low-Al dykes of thehigh Fe–Ti group cannot simply be related to the high-Al dykes throughcrystal fractionation because of their distinctly lower Fe and higher Sicontents (Figs. 4 and 5). Melting a fertile lherzolite with a Fe contentintermediate between HK-66 and KLB-1 at ~5 GPa, followed by olivine-dominated crystal fractionation (Fig. 4a), would explain the higher Fe,and lower Al contents of the low-Al dykes relative to the main groupdykes (Figs. 4c, 5b and d). These features are, however, also consistentwithmelting a garnet pyroxenite source at similarpressures ~5GPa. Thetrends of increasing Al with decreasing Mg (Fig. 4c) and increasing Si(Fig. 5d) contents defined by most low-Al dykes significantly divergesfrom olivine fractionation alone, and require the involvement ofclinopyroxene. The existence of these distinct trends defined by thelow-Al dykes in the Ca/Al vs. Mg spacemay signify that both pyroxeniteand peridotite existed, with samples having the highest Ca/Al ratios atanyMg contents beingderived fromapyroxenite source, and thosewiththe lowestCa/Al ratios beingderived fromaperidotite source. These twotrends appear to reflect the fractionation of two different proportions ofolivine and clinopyroxene (35%olivine+65%clinopyroxene for trend1;70% olivine+30% clinopyroxene for trend 2; Fig. 4d). Regardless of theexact nature of the mantle source(s) for the low-Al subgroup dykes,their low Al and Yb contents (<2 ppm; Fig. 9) and fractionated HREEprofiles (Gd/Yb >2; Fig. 8), indicate the presence of significant residualgarnet in their source, suggesting higher pressure melting than that forthemaingroupdykes. Furthermore, the spatial associationofmany low-Al dykes with the Lac Aigneau lamprophyre dykes that may have a verydeepmantle source (ca. 9 GPa, Francis and Patterson, 2009), along withtheir location over the thick lithospheric root of the southeasternUngava Peninsula, are both consistent with a deep-seated source.

The relatively higher Yb contents of the high-Al dykes areinconsistent with a garnet-bearing source and indicate melting atrelatively lower pressures. Although the high Al and low Si contents ofthe high-Al dykes are similar to experimental liquids produced bymelting fertile lherzolite at ~2 GPa (Fig. 5d), olivine fractionation of afertile source cannot yield both low Mg and Si contents, a feature thatmay imply a more Si-poor source (Fig. 5b). The low Si contents of thehigh-Al subgroup dykes are similar to those of 2–2.5 GPa experimen-tal melts of Si-undersaturated pyroxenite and to those of Archeanferropicrites of the Western Superior Province (Figs. 4a and 5c). Theirexceptional Fe-rich nature and alkaline affinities may suggest theyrepresent evolved ferropicrite melts. Although no experimental meltsexist, Archean ferropicrites may also be considered as a possiblesource for the high-Al dykes because of their similarly enrichedεNd(2.0 Ga) values and La/Yb ratios, but depleted Th/Nb and Zr/Nbratios (Fig. 10). Melting an isotopically enriched Archean ferropicritereservoir would be sustained by the high density estimated for thesemagmas (e.g. 3.33 g/cm3), which suggest they would have difficultiesrising to the Earth's surface and stagnate or sink within the mantle(Goldstein and Francis, 2008).

5.4. Nature of the mantle beneath the Ungava Peninsula

The two distinct arrays in a La–Yb–Nb ternary diagram, onedefined by the high-LREE main group dykes and low Fe–Ti dykes thatscatter from primitive mantle towards Archean granitoids, and a


second by the low-LREEmain group dykes towards the composition ofthe Paleoproterozoic alkaline magmas (Fig. 11), suggest the existenceof two distinctly enriched components. The high Fe–Ti group dykes liebetween the two arrays, which may imply contributions from bothcrustal and enriched lithospheric components. The mafic dykesexhibiting the crustal component are dominantly in the Hudson BayTerrane (Fig. 2), but no chemical difference other than isotopicenrichment appears to exist between granitoid suites of the twoisotopic terranes (Boily and Maurice, 2008), such that the geograph-ical prevalence of the crustal component is difficult to explain bydifferences in the crust dykes intrude.

In some localities of Labrador and Greenland, only dyke marginsamples have been documented to be locally contaminated by thecountry rock, which exchanged the alkalis, Ca, and the LILE with thedyke (Dostal and Fratta, 1977; Gill and Bridgwater, 1979). In contrast,a more recent study has shown that wallrock contamination is un-common (Baragar et al., 1996). Furthermore, country rocks intrudedby mafic dykes rarely show evidence of melting, even when themagmas are high-temperature picrites (Tarney, 1992), and onlyactively convecting (kilometer-scale) magma chambers providesufficient heat capacity to assimilate their host rocks (e.g. the Muskoxintrusion, Francis, 1994). The lack of correlation between dykecomposition and width in our dataset (Fig. 3) indicates that it isunlikely that local crustal contamination is responsible for theevidence of the Archean crustal component seen in the Hudson BayTerrane. Although the two subgroups of the main group dykes exhibitidentical major element systematics, a crustal component is onlyrecorded in the high-LREE subgroup. This may imply a higher flux ofradioactive heat for the crust of the Hudson Bay Terrane, in which casethe assimilation of crustal material bymafic magmaswould have beeneased by smaller temperature contrasts (Thompson et al., 2002).Alternatively, the distinct enrichments for dykes intruding bothterranes may have occurred at greater depth than fractionationestimates, through the assimilation of chemically distinct lower crustsor lithospheres.

In order to explain the voluminous ca. 2.21 Ga low-LREE maingroup dykes occurring above the thinnest lithosphere of the UngavaPeninsula, Maurice et al. (2009) suggested the magmas wereproduced by decompression melting of the asthenosphere, followingthe removal of the northern Archean lithospheric keel and/or lowercrust. In such a scenario, the parental magmas of dykes emplacedcoevally, and after the delamination event, recorded the deep crustalor lithosphere signature with which they coexisted. The lack of acrustal component and the presence of an isotopically-depleted alkalinecomponent in the low-LREE main group dykes (Figs. 10 and 11) wouldbe explained if their parental melts have reacted with the thinnednascent Paleoproterozoic lithosphere that may now underlie much ofthe Rivière Arnaud Terrane.

The nature of the alkaline component recorded in the high Fe–Tidykes across the long-lived Hudson Bay Terrane is more enigmatic.The Archean lower crust and lithosphere of the Hudson Bay Terranelikely contain a heterogeneous assortment of mafic lithologiesintroduced periodically in the Archean (Maurice et al., 2009), whichoffer a compositionally diverse range of possible sources and mantlecomponents (e.g. pyroxenite, eclogite, ferropicrite, hybrid peridotite).In the case of the high-Al high Fe–Ti dykes, their alkaline componentmay be similar to that in Archean ferropicrites with which they shareidentical enriched εNd2.0 Ga values, La/Yb ratios and depleted Th/Nbratios (Fig. 10). The slightly lower εNd2.0 Ga values at any La/Yb ratio ofthe low-Al high Fe-Ti dykes with respect to all other dyke groups(Fig. 10d) may be indicative of the incorporation of two isotopicallydistinct enriched components in their high pressure pyroxeniteand/or peridotite parental melts. A dominant component likelyreflects the assimilation of the Archean crust or melting of anArchean mafic lithology with negative εNd2.0 Ga values, while aminor component may reflect the assimilation of a mantle

component similar to the Paleoproterozoic alkaline magmas withpositive values.

Although Archean lithospheric roots are generally considered to bedepleted residues left after extensive partial melting of the mantle(Boyd, 1989; Herzberg, 1993), the widespread main group dykesrequire a relatively fertile mantle source beneath the Ungava Peninsulain the Paleoproterozoic. The fertile source of the low-LREE main groupdykes of the Rivière Arnaud Terrane may represent Paleoproterozoicmantle upwelling to replace Archean lower crust and/or lithosphere.The occurrence of the high-LREE main group, low Fe–Ti group and highFe–Ti group dykes within or along the border of the older Hudson BayTerrane implies a complex and heterogeneous lithosphere composed ofa variety of mantle lithologies, including lherzolite, harzburgite andpyroxenite.

6. Conclusions

Our study has shown that the Ungava Peninsula hosts diversegroups of Paleoproterozoic dykes swarms requiring a number ofdifferent mantle sources. The most significant correlation is betweendyke groups and the Archean terrane that they intruded. The low-LREE main group dykes intruding the juvenile Rivière Arnaud Terraneare derived from a fertile and isotopically depletedmantle, and appearto contain an alkaline component. They lack evidence of any Archeancrustal component, either signifying the crust they intruded was toocold to be assimilated, or that the lower crust and lithosphere werePaleoproterozoic in age. At least 4 possible mantle sources are neededto explain the diverse groups of Paleoproterozoic dykes intruding theolder Hudson Bay Terrane. The high-LREE main group dykes of thisterrane appear to be derived from fertile lherzolite, but haveassimilated an Archean crustal component, suggesting that fertilemantle domains persisted below this terrane in Paleoproterozoictimes despite a long history of melt extraction. Volumetrically lessimportant, but ubiquitous, high Fe–Ti and high-Al dykes within theHudson Bay Terrane may represent ferropicrite melts. High Fe–Ti butlow-Al dykes appear to represent deep-seated (5+ GPa) melts thatwere sourced in garnet-bearing pyroxenite and fertile peridotite. LowFe–Ti group dykes occur within the Hudson Bay Terrane, and alongthe boundary of the two terranes. These dykes may represent themelts of depleted harzburgite that may occur preferentially as awedge sandwiched during terrane accretion in the Archean. In theabsence of direct mantle samples, the composition of mafic dykeswarms yield an effective means of probing the nature andarchitecture of subcontinental mantle roots.

Acknowledgments

This research has been supported by a National Scientific andEngineering Research Council of Canada (NSERC) discovery grant(RGPIN7977-00) to D. Francis. J.-Y. Labbé (MRNF) supervised the re-analysis of mafic dykes of the Ungava Peninsula, J. Bédard (GSC-Québec) provided 8 unpublished analyses from Payne River and Klotzdyke swarms, and R. Ernst (Ernst Geosciences) provided analysesobtained during the TGI project #010004 (2001–2003). Manygeologists of the MRNF are acknowledged for the beneficial discus-sions. Comprehensive reviews by D. Canil and R. Ernst, and competenthandling by editor in-chief A. Kerr resulted in improvements of anearlier version of the manuscript.Ministère des Ressources naturelles etde la Faune contribution #2009-2010-8439-02.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.lithos.2009.08.002.


References

Aldanmaz, E., Koprubasi, N., Gurer, O.F., Kaymakci, N., Gournaud, A., 2006. Geochemicalconstraints on the Cenozoic, OIB-type alkaline volcanic rocks of NW Turkey:implications for mantle sources and melting processes. Lithos 86 (1–2), 50–76.

Arndt, N.T., Czamanske, G.K., Wooden, J.L., Fedorenko, V.A., 1993. Mantle and crustalcontributions to continental flood volcanism. Tectonophysics 223 (1–2), 39–52.

Baker, J.A., et al., 2000. Resolving crustal and mantle contributions to continental floodvolcanism, Yemen; constraints from mineral oxygen isotope data. Journal ofPetrology 41 (12), 1805–1820.

Baragar, W.R.A., Ernst, R.E., Hulbert, L., Peterson, T., 1996. Longitudinal petrochemicalvariation in the Mackenzie dyke swarm, Northwestern Canadian shield. Journal ofPetrology 37 (2), 317–359.

Baragar, W.R.A., Mader, U., LeCheminant, G.M., 2001. Paleoproterozoic carbonatiticultrabasic volcanic rocks (meimechites?) of Cape Smith Belt, Quebec. CanadianJournal of Earth Sciences 38 (9), 1313–1334.

Boily, M., Ludden, J.N., 1991. Trace-element and Nd isotope variations in earlyProterozoic dyke swarms emplaced in the vicinity of the Kapuskasing structuralzone— enrichedmantle or assimilation and fractional crystallization (AFC) process.Canadian Journal of Earth Sciences 28 (1), 26–36.

Boily, M., Maurice, C., 2008. Géochimie et données isotopiques du néodyme du Nord-Est de la Province du Supérieur. In: Simard, M. (Ed.), Synthèse du Nord-Est de laProvince du Supérieur. Ministère des Ressources Naturelles et de la Faune, Québec,pp. 87–129.

Boily, M., Leclair, A., Maurice, C., Bédard, J.H., David, J., 2009. Paleo- to Mesoarcheanbasement recycling and terrane definition in the Northeastern Superior Province,Québec, Canada. Precambrian Research 168 (1–2), 23–44.

Boyd, F.R., 1989. Compositional distinction between oceanic and cratonic lithosphere.Earth and Planetary Science Letters 96 (1–2), 15–26.

Buchan, K.L., Ernst, R.E., 2004. Mafic dyke swarms and related units in Canada andadjacent regions. Geological Survey of Canada, Ottawa, pp. map 2022A.

Buchan, K.L., Mortensen, J.K., Card, K.D., Percival, J.A., 1998. Paleomagnetism and U–Pbgeochronology of diabase dyke swarms of Minto block, Superior Province, Quebec,Canada. Canadian Journal of Earth Sciences 35 (9), 1054–1069.

Buchan, K.L., Goutier, J., Hamilton, M.A., Ernst, R.E., Matthews, W.A., 2007. Paleomag-netism, U–Pb geochronology, and geochemistry of Lac Esprit and other dykeswarms, James Bay area, Quebec, and implications for Paleoproterozoic deforma-tion of the Superior Province. Canadian Journal of Earth Sciences 44 (5), 643–664.

Cadman, A.C., Park, R.G., Tarney, J., Halls, H.C., 1992. Significance of anisotropy ofmagnetic-susceptibility fabrics in Proterozoic mafic dykes, Hopedale Block,Labrador. Tectonophysics 207 (3–4), 303–314.

Cadman, A.C., Tarney, J., Baragar, W.R.A., Wardle, R.J., 1994. Relationship betweenproterozoic dykes and associated volcanic sequences — evidence from the HarpSwarm and Seal-Lake-Group, Labrador, Canada. Precambrian Research 68 (3–4),357–374.

Cadman, A.C., Tarney, J., Baragar, W.R.A., 1995. Nature of mantle source contributionsand the role of contamination and in situ crystallisation in the petrogenesis ofProterozoic mafic dykes and flood basalts Labrador. Contributions to Mineralogyand Petrology 122 (3), 213–229.

Carlson, R.W., Lugmair, G.W., Macdougall, J.D., 1981. Columbia River volcanism — thequestion of mantle heterogeneity or crustal contamination. Geochimica etCosmochimica Acta 45 (12), 2483–2499.

Collerson, K.D., Sheraton, J.W., 1986. Age and geochemical characteristics of a maficdyke swarm in the Archean Vestfold Block, Antarctica — inferences aboutProterozoic dyke emplacement in Gondwana. Journal of Petrology 27 (4), 853–886.

Condie, K.C., 1997. Sources of Proterozoic mafic dyke swarms: constraints from Th/Taand La/Yb ratios. Precambrian Research 81 (1–2), 3–14.

Cox, K.G., 1980. A model for flood basalt vulcanism. Journal of Petrology 21, 629–650.Dostal, J., Fratta, M., 1977. Trace-element geochemistry of a Precambrian diabase dike

from Western Ontario. Canadian Journal of Earth Sciences 14 (12), 2941–2944.Ellam, R.M., Cox, K.G., 1991. An interpretation of Karoo picrite basalts in terms of

interaction between asthenospheric magmas and the mantle lithosphere. Earthand Planetary Science Letters 105 (1–3), 330–342.

Ernst, R.E., 1994. Mapping the magma flow pattern in the Sudbury dyke swarm usingmagnetic fabric analysis. Current Research 1994-E. Geological Survey of Canada,Ottawa.

Ernst, R.E., Baragar, W.R.A., 1992. Evidence from magnetic fabric for the flow pattern ofmagma in the Mackenzie Giant Radiating Dyke Swarm. Nature 356 (6369),511–513.

Ernst, R.E., Buchan, K.L., 2001. The use of mafic dike swarms in identifying and locatingmantle plumes. In: Ernst, R.E., Buchan, K.L. (Eds.), Mantle Plumes: TheirIdentification Through Time. Geological Society of America, pp. 247–265.

Ernst, R.E., Head, J.W., Parfitt, E., Grosfils, E., Wilson, L., 1995. Giant radiating dykeswarms on Earth and Venus. Earth-Science Reviews 39 (1–2), 1–58.

Fahrig, W.F., 1987. The tectonic settings of continental mafic dyke swarms: failed armsand early passive margin. In: Fahrig, W.F., Halls, H.C. (Eds.), Mafic dyke swarms.Geological Association of Canada, pp. 331–348.

Falloon, T.J., Danyushevsky, L.V., 2000. Melting of refractory mantle at 1.5, 2 and 2.5 GPaunder, anhydrous and H2O-undersaturated conditions: implications for thepetrogenesis of high-Ca boninites and the influence of subduction componentson mantle melting. Journal of Petrology 41 (2), 257–283.

Francis, D., 1994. Chemical interaction between picritic magmas and upper crust alongthe margins of the Muskox intrusion, Northwest Territories. Geological Survey ofCanada, paper 92-12. 26 pp.

Francis, D., Patterson, M., 2009. Kimberlites and aillikites as probes of the continentallithospheric mantle. Lithos 109 (1–2), 72–80.

French, J.E., et al., 2004. Application of Electron Microprobe Chemical BaddeleyiteDating to Reconnaissance Geochronological Investigations of Mafic Dyke Swarmsfrom the Slave Province, NT. InGAC-MAC, St. Catharines, Ontario, p. A586.

Fujimaki, H., Tatsumoto, M., Aoki, K.I., 1984. Partition coefficients of Hf, Zr, and REEbetween phenocrysts and groundmasses. Journal of Geophysical Research 89,B662–B672.

Gaonac'h, H., Ludden, J.N., Picard, C., Francis, D., 1992. Highly alkaline lavas in aProterozoic rift zone: implications for Precambrian mantle metasomatic processes.Geology 20 (3), 247–250.

Gibson, I.L., Sinha, M.N., Fahrig, W.F., 1987. The geochemistry of the Mackenzie dykeswarm, Canada. In: Halls, H.C., Fahrig, W.F. (Eds.), Mafic Dyke Swarms. GeologicalAssociation of Canada, pp. 109–121.

Gill, R.C.O., Bridgwater, D., 1979. Early Archaean basic magmatism in west Greeland —

geochemistry of the Ameralik Dykes. Journal of Petrology 20 (4), 695–726.Goldstein, S.B., Francis, D., 2008. The petrogenesis and mantle source of Archean

Ferropicrites from the Western Superior Province, Ontario, Canada. Journal ofPetrology 49 (10), 1729–1753.

Green, D.H., 1973. Experimental melting studies on a model upper mantle compositionat high-pressure under water-saturated and water-undersaturated conditions.Earth and Planetary Science Letters 19 (1), 37–53.

Grove, T.L., Baker, M.B., 1984. Phase-equilibrium controls on the tholeiitic versus calc-alkaline differentiation trends. Journal of Geophysical Research 89 (NB5), 3253–3274.

Hart, W.K., 1985. Chemical and isotopic evidence for mixing between depleted andenriched mantle, Northwestern USA. Geochimica Et Cosmochimica Acta 49 (1),131–144.

Hawkesworth, C.J., Rogers, N.W., Vancalsteren, P.W.C., Menzies, M.A., 1984. Mantleenrichment processes. Nature 311 (5984), 331–335.

Hergt, J.M., Brauns, C.M., 2001. On the origin of Tasmanian dolerites. Australian Journalof Earth Sciences 48 (4), 543–549.

Herzberg, C.T., 1993. Lithosphere peridotites of the Kaapvaal Craton. Earth andPlanetary Science Letters 120 (1–2), 13–29.

Hirose, K., 1997. Partial melt compositions of carbonated peridotite at 3 GPa and role ofCO2 in alkali-basalt magma generation. Geophysical Research Letters 24 (22),2837–2840.

Hirose, K., Kushiro, I., 1993. Partial melting of dry peridotites at high pressures;determination of compositions of melts segregated from peridotite usingaggregates of diamond. Earth and Planetary Science Letters 114 (4), 477–489.

Hirschmann, M.M., Stolper, E.M., 1996. A possible role for garnet pyroxenite in theorigin of the “garnet signature” in MORB. Contributions to Mineralogy andPetrology 124 (2), 185–208.

Hirschmann, M.M., Kogiso, T., Baker, M.B., Stolper, E.M., 2003. Alkalic magmasgenerated by partial melting of garnet pyroxenite. Geology 31 (6), 481–484.

Jourdan, F., et al., 2004. The Karoo triple junction questioned: evidence from Jurassicand Proterozoic Ar-40/Ar-39 ages and geochemistry of the giant Okavango dykeswarm (Botswana). Earth and Planetary Science Letters 222 (3–4), 989–1006.

Jourdan, F., et al., 2006. Basement control on dyke distribution in Large IgneousProvinces: case study of the Karoo triple junction. Earth and Planetary ScienceLetters 241 (1–2), 307–322.

Kinzler, R.J., 1997. Melting of mantle peridotite at pressures approaching the spinel togarnet transition: application to mid-ocean ridge basalt petrogenesis. Journal ofGeophysical Research 102 (B1), 853–874.

Klein, E.M., Langmuir, C.H., 1987. Global correlations of ocean ridge basalt chemistrywith axial depth and crustal thickness. Journal of Geophysical Research–Solid Earthand Planets 92 (B8), 8089–8115.

Kogiso, T., Hirose, K., Takahashi, E., 1998. Melting experiments on homogeneousmixtures of peridotite and basalt: application to the genesis of ocean island basalts.Earth and Planetary Science Letters 162 (1–4), 45–61.

Kogiso, T., Hirschmann, M.M., Frost, D.J., 2003. High-pressure partial melting of garnetpyroxenite: possible mafic lithologies in the source of ocean island basalts. Earthand Planetary Science Letters 216 (4), 603–617.

Kogiso, T., Hirschmann, M.M., Pertermann, M., 2004. High-pressure partial melting ofmafic lithologies in the mantle. Journal of Petrology 45 (12), 2407–2422.

Langmuir, C.H., Hanson, G.N., 1980. An evaluation of major element heterogeneity inthe mantle sources of basalts. In: Bailey, D.K., Tarney, J., Dunham, K. (Eds.), Theevidence for chemical heterogeneity in the Earth's mantle. Philosophical Transac-tions of the Royal Society of London, Series A: Mathematical and Physical Sciences.Royal Society of London, London, United Kingdom, pp. 383–407.

Lewry, J.F., Collerson, K.D., 1990. The Trans-Hudson orogen: extent, subdivision, andproblems. In: Stauffer, J.F.L.a.M.R. (Ed.), The Early Proterozoic Trans-HudsonOrogen of North America. Geological Association of Canada, pp. 1–4.

Longhi, J., 2002. Some phase equilibrium systematics of lherzolite melting: I.Geochemistry, Geophysics, Geosystems 3 (3), 1020. doi:10.1029/2001GC000204.

Mahood, G.A., Baker, D.R., 1986. Experimental constraints on depths of fractionation ofmildly alkalic basalts and associated felsic rocks: Pantelleria, Strait of Sicily.Contributions to Mineralogy and Petrology 93 (2), 251–264.

Maurice, C., 2007. New isotopic neodymium data in the Northeastern Superior Province.RP 2006-06(A). Ministère des Ressources naturelles et de la Faune, Québec. 13 pp.

Maurice, C., 2008. Les essaims de dykes mafiques du nord-est de la Province duSupérieur. In: Simard, M. (Ed.), Synthèse du nord-est de la Province du Supérieur.Ministère des Ressources naturelles et de la Faune, Québec, pp. 137–141.

Maurice, C., David, J., Bédard, J.H., Francis, D., 2009. Evidence for a widespread maficcover sequence and its implications for continental growth in the NortheasternSuperior Province. Precambrian Research 168 (1–2), 45–65.

Maurice, C., David, J., O'Neil, J., Francis, D., 2009. Age and tectonic implications ofPaleoproterozoic mafic dyke swarms for the origin of 2.2 Ga enriched lithospherebeneath the Ungava Peninsula, Canada. Precambrian Research 174 (1–2), 163–180.


McHone, J.G., Anderson, D.L., Beutel, E.K., Fialko, Y.A., 2005. Giant dikes, rifts, floodbasalts, and plate tectonics; a contention of mantle models. Special Paper —Geological Society of America 388, 401–420.

McKenzie, D., 1989. Some remarks on the movement of small melt fractions in themantle. Earth and Planetary Science Letters 95 (1–2), 53–72.

Mège, D., Korme, T., 2004. Dyke swarm emplacement in the Ethiopian large igneousprovince: not only a matter of stress. Journal of Volcanology and GeothermalResearch 132 (4), 283–310.

Modeland, S., Francis, D., Hynes, A., 2003. Enriched mantle components in Proterozoiccontinental-flood basalts of the Cape Smith foldbelt, northern Quebec. Lithos 71(1), 1–17.

O'Neil, J., Carlson, R.W., Francis, D., Stevenson, R.K., 2008. Neodymium-142 Evidence forHadean Mafic Crust. Science 321 (5897), 1828–1831.

Patchett, P.J., Lehnert, K., Rehkamper, M., Sieber, G., 1994. Mantle and crustal effects onthe geochemistry of Proterozoic dikes and sills in Sweden. Journal of Petrology 35(4), 1095–1125.

Peng, Z.X., Mahoney, J., Hooper, P., Harris, C., Beane, J., 1994. A role for lowercontinental-crust in flood-basalt genesis — isotopic and incompatible elementstudy of the lower 6 formations of the western Deccan traps. Geochimica EtCosmochimica Acta 58 (1), 267–288.

Percival, J.A., Mortensen, J.K., 2002. Water-deficient calc-alkaline plutonic rocks ofnortheastern Superior Province, Canada; significance of charnockitic magmatism.Journal of Petrology 43 (9), 1617–1650.

Roy, P., Turcotte, S., Sharma, K.N.M., David, J., 2004. Géologie de la région du lacMontrochand (SNRC 33O). RG 2003-10. Ministère des Ressources naturelles, de laFaune et des Parcs, Québec. 39 pp.

Seymour, K.S., Kumarapeli, P.S., 1995. Geochemistry of the Grenville Dyke Swarm— roleof plume-source mantle in the magma genesis. Contributions to Mineralogy andPetrology 120 (1), 29–41.

Simard, M., 2008. Synthèse du Nord-Est de la Province du Supérieur. Ministère desRessources naturelles et de la Faune, Québec. 196 pp.

Southwick, D.L., Day, W.C., 1983. Geology and petrology of proterozoic mafic dikes,North-Central Minnesota and western Ontario. Canadian Journal of Earth Sciences20 (4), 622–638.

Southwick, D.L., Halls, H.C., 1987. Compositional characteristics of the KenoraKabetogama dyke swarm (early Proterozoic), Minnesota and Ontario. CanadianJournal of Earth Sciences 24 (11), 2197–2205.

Sun, S.S., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanic basalts;implications for mantle composition and processes. In: Saunders, A.D., Norry, M.J.(Eds.), Magmatism in the ocean basins. Geological Society Special Publications.Geological Society of London, London, United Kingdom, pp. 313–345.

Takahashi, E., Kushiro, I., 1983. Melting of a dry peridotite at high-pressures and basaltmagma genesis. American Mineralogist 68 (9–10), 859–879.

Tarney, J., 1992. Geochemistry and significance of mafic dyke swarms in theProterozoic, Proterozoic crustal evolution. Developments in Precambrian Geology.Elsevier, Amsterdam, Netherlands, pp. 151–179.

Tarney, J., Weaver, B.L., 1987. Geochemistry and petrogenesis of early Proterozoic dykeswarms. In: Halls, H.C., Fahrig, W.F. (Eds.), Mafic Dyke Swarms. GeologicalAssociation of Canada, pp. 81–94.

Thompson, A.B., Matile, L., Ulmer, P., 2002. Some thermal constraints on crustalassimilation during fractionation of hydrous, mantle-derived magma withexamples from Central Alpine batholiths. Journal of Petrology 43 (3), 403–422.

Thy, P., 1991. High and low pressure phase equilibria of a mildly alkalic lava from the1965 Surtsey eruption: experimental results. Lithos 26 (3–4), 223–243.

Thy, P., Lesher, C.E., Fram, M.S., 1998. Low pressure experimental constraints on theevolution of basaltic lavas from site 917, southeast Greenland continental margin.In: Saunders, A.D., Larsen, H.C., Wise, S.W. (Eds.), Proceedings of the Ocean DrillingProgram, Scientific Results, College Station, TX, pp. 359–372.

Tomlinson, K.Y., Stott, G.M., Percival, J.A., Stone, D., 2004. Basement terrane correlationsand crustal recycling in the western Superior Province; Nd isotopic character ofgranitoid and felsic volcanic rocks in the Wabigoon Subprovince, N. Ontario,Canada. Precambrian Research 132 (3), 245–274.

Villiger, S., Ulmer, P., O., M., 2007. Equilibrium and fractional crystallizationexperiments at 0.7 GPa; the effect of pressure on phase relations and liquidcompositions of tholeiitic magmas. Journal of Petrology 48 (1), 159–184.

Villiger, S., Ulmer, P., O., M., Thompson, A.B., 2004. The liquid line of descent ofanhydrous, mantle-derived, tholeiitic liquids by fractional and equilibriumcrystallization—an experimental study at 1.0 GPa. Journal of Petrology 45 (12),2369–2388.

Walter, M.J., 1998. Melting of garnet peridotite and the origin of komatiite and depletedlithosphere. Journal of Petrology 39 (1), 29–60.

Wood, B.J., Blundy, J.D., 1997. A predictive model for rare earth element partitioningbetween clinopyroxene and anhydrous silicate melt. Contributions to Mineralogyand Petrology 129 (2–3), 166–181.

Wright, W.R., Mariano, A., Hagni, R.D., 1999. Pyrochlore, mineralization and glimmeriteformation in the Eldor (Lake LeMoyne) carbonatite complex, Labrador Trough,Quebec, Canada. In: CIM (Ed.), 33rd Forum on the Geology of Industrial Minerals,pp. 205–213.

Yaxley, G.M., Green, D.H., 1998. Reactions between eclogite and peridotite: mantlerefertilisation by subduction of oceanic crust. Schweizerische Mineralogische etPetrologische Mitteilung 78, 243–255.

Enriched crustal and mantle components and the role of the lithosphere in generating...

Documents

Transcript of Enriched crustal and mantle components and the role of the lithosphere in generating...