This paper presents major, trace-elements, and Sr-Nd isotopes for two prominent sills formed during the opening of the North Atlantic, sampled by the Utgard borehole (6607/5-2) in the Vøring Plateau. The Utgard sills are compared to opening-related lavas recovered from ODP Leg 104 Hole 642E farther west on the Vøring Plateau and on the NE Greenland conjugate margin. The Utgard sills (3.6–5.9 wt % MgO) are enriched in strongly relative to moderately incompatible trace elements and have 87Sr/86Sr and 143Nd/144Nd ratios of 0.70380–70387 and 0.51292–0.51293, respectively, in the Upper Utgard Sill, and 0.70303–0.70306 and 0.51297–0.51299 in the Lower Sill. Alteration is minor. The Utgard melts originated by partial melting of an asthenospheric, depleted mantle source (DMM or Iceland Rift Zone, IRZ, type) with chemical characteristics similar to the source that gave rise to NE Greenland lavas. The Utgard magmas underwent extensive fractional crystallization in the lower crust (Upper Sill: >70%; Lower Sill: >55%) with removal mainly of olivine and pyroxenes, accompanied by ≤1% assimilation of crustal melts. This crystallization formed significant masses of dense cumulates (∼3.25 g/cm3) (underplating). Assuming an areal extent similar to that of the two sills, we estimate a composite layer of ultramafic cumulates mixed with less dense country rocks to be >320 m thick beneath the two Utgard sills and >8.8 km beneath the thickest part of the Vøring Plateau lavas. Opening-related cumulates may thus account for a significant part of the lower crustal high-velocity, high-density bodies (average density 3.1 g/cm3) along the Norwegian margin.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The opening of the NE Atlantic was accompanied by anomalously high breakup related magmatic activity forming the North Atlantic Large Igneous Province (Figure 1) [e.g., Talwani and Eldholm, 1977; Eldholm et al., 1989]. Along the Norwegian margin, the volcanism includes seaward dipping reflectors, sills, and dykes, mainly in the Møre and Vøring basins [e.g., Planke and Eldholm, 1994; Eldholm et al., 1989; Eldholm, 1991; Berndt et al., 2000; Planke et al., 2000, 2005]. In contrast to Greenland, there are no volcanic exposures on the Norwegian margin; petrological-geochemical information is therefore sparse. The conjugate east Greenland margin shows various types of shallow intrusions in addition to large volumes of flood basalts at the Blosseville Kyst, Jameson Land Basin, Hold with Hope, and Wollaston Forland [e.g., Brooks et al., 1976; Upton et al., 1984; Hald and Tegner, 2000; Brooks, 2011] (Figure 1). Seismic surveys have also identified extensive lower crustal high-velocity, high-density bodies (LCBs) beneath both the Norwegian (Figure 2) and east Greenland margins and the adjacent oceanic lithosphere [e.g., Planke et al., 1991; Skogseid et al., 1992; Mjelde et al., 2002, 2009; Voss and Jokat, 2007; White et al., 2008; Reynisson et al., 2010].
 On the East Greenland margin, the timing and chemical characteristics of the breakup magmatism have been studied both on land and in drill cores of the seaward-dipping reflectors (ODP legs 152 and 163). It has been shown that flood basalts and intrusions were emplaced over tens of millions of years (∼61–14 Ma) with a short peak less than 300,000 years long, coinciding with continental rupture ∼56 million years ago [Larsen and Tegner, 2006; Storey et al., 2007b]. The regional, voluminous magmatism before and during continental breakup has been linked to melting of the Iceland mantle plume and records secular changes in melting conditions, mantle sources, and crustal assimilation controlled by lithosphere tectonics [Thirlwall et al., 1994; Fram and Lesher, 1997; Tegner et al., 1998a, 2008; Fitton et al., 2000; Peate et al., 2008; Voss and Jokat, 2007]. So far chemical data on the breakup magmatism on the Norwegian margin is restricted to the lava series recovered in a single drill core, ODP Leg 104 Hole 642E in the outer part of the Vøring Plateau and Deep Sea Drilling Programme (DSDP) Sites 338, 342, and 343 on the northern Vøring Margin (Figure 1) [e.g., Viereck et al., 1988, 1989; Parson et al., 1989; Meyer et al., 2009].
 This paper presents new data on two extensive sills penetrated by borehole 6607/5-2 located at the Utgard High in the Vøring Basin (Figure 1). The aim is to (1) establish the geochemical character and evolutionary history of the Utgard sills; (2) compare the Utgard sills to lavas in ODP Leg 104 Hole 642E on the Vøring Plateau and to outcrops of flood basalt at Hold with Hope and Wollaston Forland on the conjugate NE Greenland margin in order to examine regional variations with respect to crustal contamination, melting processes, and mantle sources; and (3) discuss possible relationships to the LBC in the deep crust beneath the Vøring Basin (Figure 2).
2. Geological Setting
 The thick sequence of lavas forming seaward dipping reflectors covering the outer part of the Møre and Vøring margins and adjacent oceanic crust exceeds 6 km in composite thickness and extruded during the last stages of rifting and earliest stages of seafloor spreading [e.g., Skogseid and Eldholm, 1987, 1989; Planke and Eldholm, 1994]. ODP Leg 104 Hole 642E (hereafter referred to as 642E) was drilled on the Vøring Plateau on continental lithosphere in the innermost part of the lavas that form the seaward dipping reflectors (Figures 1 and 2) [Eldholm et al., 1989].
 The Vøring and Møre basins east of the seaward dipping reflectors contain voluminous magmatic complexes of dominantly subhorizontal sheets (sills) that intruded Cretaceous sedimentary rocks during opening of the northeast Atlantic [e.g., Berndt et al., 2000; Planke et al., 2005]. Sill intrusions and the Mesozoic basins cover >85,000 km2 offshore mid-Norway (Figures 1 and 2). Seismic reflection data and amplitude modeling imply 6–7 sills with thicknesses of up to 100 m for individual sills [Berndt et al., 2000]. The sills have been penetrated by a few industrial boreholes on structural highs, but detailed geochemical studies of the sills have not yet been reported. Hydrothermal vent complexes are also abundant, more than 700 craters up to 12 km in diameter are mapped on the Paleocene-Eocene paleo-seafloor by seismic 2-D and 3-D imaging. Also the large number of craters and their size reflect the great volumes of magma emplaced as sills in the area during a short-time interval. Furthermore, these craters are interpreted to have formed during violent release of gas generated within the contact aureoles around the sills, possibly contributing to the global warming event at the Paleocene-Eocene boundary [Svensen et al., 2004, 2010; Planke et al., 2005; Storey et al., 2007a].
 The Utgard dolerite sills are 91 m (Utgard Upper Sill; located at 3792–3883 m depth) and >50 m (Utgard Lower Sill; upper contact at 4650 m depth), emplaced in Upper Cretaceous mudstones and sandstones (Figure 2) [Berndt et al., 2000]. Drilling terminated 50 m into the lower sill, thus its total thickness remains unknown. The sills are well imaged on seismic profiles and can be followed for more than 100 km westward into the deeper part of the basin (Figure 2). Recent radiometric dating gave U-Pb zircon ages of 55.6 ± 0.3 Ma for the Upper Utgard Sill and 56.3 ± 0.4 Ma for the Lower Utgard Sill [Svensen et al., 2010], showing that the sills were emplaced during the early stages of the breakup-related volcanism [Storey et al., 2007b]. The dolerite fragments forming each sample represent a visually homogenous population regarding color and grain size. Margin samples contained fragments of country rock which were carefully removed from the analyzed fraction.
3.1. Petrography and Compositional Variations
 The Utgard sills consist mainly of plagioclase (∼50%), clinopyroxene (∼42%), olivine (∼3%), and magnetite (∼5%), with small amounts of apatite. In the Upper Sill, olivine grains are <0.5 mm in diameter, rounded, and commonly enclosed by clinopyroxene. Clinopyroxene partly forms large (≤3 mm in diameter) subhedral grains that may enclose plagioclase grains along their boundaries, partly it forms smaller, subhedral to anhedral grains (<0.5–3 mm in diameter) interlocked with, or enclosed by, plagioclase. Also plagioclase shows a wide range in size (<0.5–3 mm long) and is mainly subhedral. Magnetite is up to 1 mm in diameter, the larger grains are rounded. The Lower Sill shows even grain size (<1 mm in diameter). Clinopyroxene is mostly anhedral and may locally enclose several plagioclase grains, plagioclase is mainly subhedral. Both sills show minor alteration expressed by alteration rims on olivine and minor amounts of biotite and chlorite.
 Major and trace element data are given in Table 1 and shown in Figures 3-6. In the total alkalis versus silica (TAS) system [Le Bas et al., 1986], the sills classify as basalts and fall in the subalkaline/tholeiitic field of MacDonald  and Kuno  (Figure 1a in supporting information).1 CIPW norm calculations show small amounts of normative quartz (0.04–7.4 mole percent). The rocks have low contents of MgO (3.6–5.9 wt%), Ni (35–71 ppm), and Cr (22–99 ppm), and high concentrations of total iron (12.3–15.4 wt % Fe2O3) and incompatible trace elements (e.g., 0.6–1.9 ppm Th, 6–18 ppm La, and 28–43 ppm Y; Table 1) relative to primitive tholeiitic basalts. Samples from the Lower Sill are, on average, slightly more MgO-rich than those from the Upper Sill (Figure 3).
Table 1. Major Element (Weight %) and Trace Element (ppm) Data on Samples From the Utgard Sillsa
Upper Sill (US)
Upper Sill (US)
Lower Sill (LS)
Sample numbers show depths in m.
 The sills show considerable scatter with respect to major elements, but Al2O3, Na2O, and Sr increase with decreasing MgO, whereas CaO and particularly Ni, Sc, and Cr, decrease (Figures 3 and 4). The Lower Sill also shows increasing mg# and decreasing concentrations of incompatible elements with stratigraphic height, whereas no such relationships are seen in the Upper Sill (Figure 7 and Table 1). Samples from each sill show essentially parallel REE-patterns strongly enriched in light rare earth elements (LREE), relative to heavy REE (HREE; Figure 6). The Upper Sill is enriched in strongly incompatible elements relative to the Lower Sill (e.g., Upper Sill: ThN = 17–22, [Ce/Yb]N = 3.0–3.3, Lower Sill: ThN = 7.6–12.9, (Ce/Yb)N = 1.7–1.9), but both sills have negative K-, Sr-, and P-anomalies. There is little variation in ratios of incompatible elements in each sill, but incompatible element ratios are distinct in the two sills (e.g., Th/Y, Nb/Y, and La/Y; Figure 6), indicating distinct origins.
 Sr-Nd isotope data on selected Utgard samples are listed in Table 2 and shown in Figure 8. The Upper Sill samples gave 143Nd/144Nd ratios almost within analytical error (0.512917 ± 0.000004 to 0.512932 ± 0.000005), whereas the range in 87Sr/86Sr ratios is significant (0.703801 ± 0.000012 to 0.703872 ± 0.000011). The highest 87Sr/86Sr ratios were obtained from samples close to the contacts against sedimentary wall rocks (Figure 8), suggesting mild local contamination. The two Lower Sill samples show significantly different Sr and Nd isotopic ratios and have slightly higher Nd and lower Sr isotopic ratios than the Upper Sill. Both sills have higher 87Sr/86Sr and lower 143Nd/144Nd than mid-ocean ridge basalts (MORB) and most Icelandic compositions [e.g., Thirlwall et al., 2004; Kokfeldt et al., 2006] (Figure 9a).
Table 2. Sr-Nd Isotope Data on Selected Samples From the Utgard Sillsa
Sample numbers show depths (m).
4.1. Correlation of Conjugate Margins
 The Utgard sills and lavas on the Vøring Plateau (642E) and on the conjugate NE Greenland margin (Hold with Hope and Wollaston Foreland) represent opening-related magmatism at different distances from, and different sides of, the continent-ocean boundary (Figures 1 and 2). Chemical similarities and differences between the magmatic rocks at the three locations provide important information about their evolutionary histories. Below we therefore summarize the petrology and geochemistry of 642E and NE Greenland lavas, followed by a discussion of the evolutionary history of breakup volcanism at the conjugate margins of the Greenland-Norwegian Sea.
 Lavas collected in the 642E core on the Vøring Plateau (Figure 1a) consist of an Upper Series (US) of tholeiitic E-MORB-type basalts, and a very heterogeneous Lower Series (LS) that comprises rhyolitic ignimbrites, tholeiitic basalts, basaltic andesites, and dacites [Viereck et al., 1988, 1989; Parson et al., 1989]. Based on trace elements and Sr-Nd isotope data, Viereck et al. [1988, 1989], Parson et al. , and Meyer et al.  concluded that the 642E US lavas are mostly uncontaminated and have essentially preserved the trace element and isotopic signature imprinted by a mantle source somewhat more radiogenic than the depleted MORB mantle (DMM). The LS assimilated significant proportions of anatectic melts formed from upper crustal rocks. The pattern of contamination seen in the Vøring Plateau succession is very similar to that seen on the SE Greenland margin and onshore East Greenland [e.g., Fitton et al., 1998, 2000].
 On the conjugate margin in NE Greenland (Figure 1a), Thirlwall et al.  divided the lavas into three main groups based on Sr-Nd-Pb isotope and trace element compositions. Most Hold with Hope (HwH) and Wollaston Forland (WolF) Lower Series (LS) lavas are largely uncontaminated quartz tholeiites. One HwH group, termed “normal,” have (La/Yb)N ratios of about 5.5, whereas a group with (La/Yb)N ratios of 7.6–26 was termed “anomalous” (this group is disregarded below). The HwH Upper Series (US) lavas are heterogeneous and have isotope and trace element compositions that indicate significant assimilation of (upper) crustal material combined with fractional crystallization (AFC processes), and may also contain a small component from the subcontinental lithospheric mantle [Thirlwall et al., 1994]. Thirlwall et al.  found the NE Greenland basalt sequence to be produced from the Iceland plume source, with no component from the DMM. They also concluded that the lavas are generated by mixing between relatively small degree melts from garnet facies with relatively high degree melts from spinel facies mantle.
 We have adopted the terms “contaminated” and “uncontaminated” used in the discussion of the Vøring Plateau (642E) and in NE Greenland (HwH + WolF) lavas. “Uncontaminated” lavas are those that have 87Sr/86Sr below 0.7037 and 143Nd/144Nd above 0.5129 (Figure 9). With respect to Sr-Nd isotopes, both Utgard sills fall in the “uncontaminated” group and share similarities in particular with the NE Greenland LS lavas. The NE Greenland lavas are, on average, richer in MgO than the Utgard sills, and the highest MgO contents are found in the 642E lavas. The Lower Utgard Sill has many compositional features common with the NE Greenland lavas. In major and trace element diagrams, the NE Greenland lavas define trends that coincide with, or pass close to, the Lower Utgard Sill, the two rock groups overlap with respect to trace element patterns and have similar ratios between strongly and moderately incompatible elements (Figures 3-6). The Upper Utgard Sill has lower concentrations of compatible elements, higher concentrations of strongly incompatible elements, and higher ratios between strongly and moderately incompatible elements. The 642E lavas, in contrast, show considerable scatter with respect to major elements and incompatible trace elements and most samples are less evolved (higher concentrations of MgO, Ni, Sc, and Cr than the other groups; Figures 3 and 4). The 642E lavas also differ from the Utgard sills and the NE Greenland lavas by having relatively flat HREE patterns, whereas the Utgard and the NE Greenland samples exhibit regularly decreasing enrichment factors from La to Yb (Figure 5). In the Sr-Nd isotope diagram, the NE Greenland LS (NEGLS) and Utgard Upper Sill define a common trend with the Eastern Mohns Ridge (EMR) falling at the low-87Sr/86Sr, and the Upper Utgard Sill at the high-87Sr/86Sr end (Figure 9b). This trend lies within the field of Iceland basalts. The Lower Utgard Sill falls slightly above this trend. The 642E US likewise plots above the EMR-NEGLS-Upper Utgard Sill trend with uniform 143Nd/144Nd ratios and 87Sr/86Sr ratios extending almost to 0.7060. Finally, the 642E US lavas have much wider ranges in Nb/Th and Nb/La ratios than the Utgard sills and NE Greenland LS (Figures 9b and 10), and they show no correlation between 87Sr/86Sr and Nb/Th or Nb/La ratios.
4.2. AFC Processes
 The only compositional evidence of in situ contamination is the slightly elevated Sr isotope ratios at the roof and floor of the Utgard Upper Sill, as compared to its central parts (Figure 8). The following discussion is therefore focused on samples from the interior parts of the sills.
 The low concentrations in MgO and other compatible elements (e.g., 3.9–5.9 wt % MgO, 30–75 ppm Ni; Table 1 and Figures 3 and 4) indicate that the magmas that gave rise to the Utgard sills have an extensive history of fractional crystallization. In Figure 4 (Ni-MgO diagram), we propose two stages of fractional crystallization. Stage 2 represented by the observed compositional range within each sill occurred after intrusion of the magmas into the upper crust. Stage 1 represented by the compositional difference between an assumed initial melt and the least evolved sample in each sill, took place before intrusion into the upper crust. Fractional crystallization is discussed in detail in supporting information.1 In summary, Stage 1 involved extensive removal of olivine, clinopyroxene, and possibly minor amounts of plagioclase. In the Lower Sill, Stage 1 may also have involved removal of minor amounts of magnetite. It was concluded that crystallization during Stage1 occurred in the lower crust. During Stage 2, the magmas crystallized the observed mineral assemblage, olivine + clinopyroxene + plagioclase + magnetite + apatite.
 Both Utgard sills have lower Nd and higher Sr isotopic ratios than many of the “uncontaminated” NE Greenland and 642E lavas (Figure 9) and the two sills have slightly different Sr-Nd isotope ratios. These differences suggest that, in addition to fractional crystallization, their parent magmas may have been subjected to crustal contamination. Minor contamination in the Utgard sills is supported by their Rb/Zr and K/Nb ratios (Figure 11). Both ratios are typically low in basalts derived from depleted mantle sources, and high in crustal rocks. As Zr and K have lower mineral/melt partition coefficients than Rb and Nb in basaltic systems [e.g., Sun and McDonough, 1989], fractional crystallization is expected to cause significant increases in the Rb/Zr ratio and mild decreases in the K/Nb ratio. Because of the high ratios in crustal rocks, assimilation of crustal melts will cause both ratios to increase. Both the 642E and NE Greenland lavas show trends from low toward high Rb/Zr and K/Nb ratios with the “uncontaminated” lavas (642E US and NE Greenland LS) at the low-ratio end (Figure 11); the highest K/Nb ratios are found among the “contaminated” 642E LS. The wide field shown by the Viereck et al.  data on 642E US, as compared to the data of Meyer et al. , is probably due to a less rigorous selection of unaltered rocks. The positions of the different series in Figure 11 are in perfect agreement with the conclusions that the 642E LS and NE Greenland US are contaminated [Parson et al., 1989; Thirlwall et al., 1994; Meyer et al., 2009]. The Utgard sills fall at the transition between “uncontaminated” and “contaminated” 642E and NE Greenland series. The relatively high K/Nb and Rb/Zr ratios of the Utgard sills (Figure 11) thus support crustal contamination in addition to fractional crystallization.
 The relatively uniform compositions of each Utgard sill might be taken as evidence against contamination. If so, the Utgard sills, the 642E lavas and the NE Greenland lavas must be derived from a heterogeneous mantle source or different sources. In our opinion, the uniform compositions of each Utgard sill are not evidence against contamination. In a study of sills and dykes in the Golden Valley Complex in the Karoo Basin (South Africa), Neumann et al.  found each sill and dyke to have uniform ratios between incompatible trace elements and uniform Sr and Nd isotope ratios. However, the units showed different ratios, so that such ratios could be used to “fingerprint” the magma batch that gave rise to a given unit. When the various units within a restricted area were compared, it became apparent that they fell on a common trend compatible with different degrees of assimilation and fractional crystallization (AFC processes) in the deep crust. Each batch of magma had clearly homogenized before intrusion into the upper crust. AFC processes during the evolution of the Utgard sills have been tested using the energy-constrained assimilation-fractional crystallization (EC-RAFC) model of Spera and Bohrson .
 The EC-RAFC model requires assumptions about the compositions of the primitive melts and assimilants. Regarding the primary melt, we tested several candidates. The low 87Sr/86Sr and high 143Nd/144Nd ratios of the Utgard sills (≥0.7039 and ≤0.51292, respectively; Figure 9) indicate that the mantle source(s) of these sills must be depleted. Depleted MORB mantle, DMM, is the most common mantle source associated with sea-floor spreading. We therefore tested a primary melt, Melt1, with the Sr-Nd isotope composition of DMM [Workman and Hart, 2005] and the trace element concentrations of average N-MORB [Sun and McDonough, 1989]. As indicated above, the mantle source that gave rise to the NE Greenland lavas is thought to be the Iceland plume (termed IRZ: Iceland Rift Zone) [Debaille et al., 2009]. Melt2 has the Sr-Nd isotope compositions the mantle source IRZ [Debaille et al., 2009] and the trace element concentrations of a picritic lava from the Reykjanes Penninsula in Iceland with the same Sr and Nd istope ratios as IRZ. The source of the 642E lava series is proposed to be slightly more radiogenic than the DMM [Meyer et al., 2009], but details have not been given. Melt3 represents an attempt to find the chemical characteristics of a primary, uncontaminated source for the 642E lavas. Details on mantle sources and primary melts are given in supporting information.
 Data on the crustal basement beneath the Vøring Margin are not available. Rocks from the adjacent Norwegian mainland, the Western Gneiss Region (WGR), were therefore used as proxy. The choice of assimilants are discussed in detail in supporting information. Data for primary melts, assimilants, other parameters used in the modeling, and references are listed in Tables 3 and 4.
Table 3. Parameters Used in EC-RAFC Modeling
Tlm—Liquidus of magma
Tmo—Initial temperature of magma
Tla—Liquidus of assimilant
Tao—Initial temperature of assimilant
Ts—Solidus (melt and assimilant)
Cpm—Specific heat of magma
1495 J/kg K
Cpa—Specific heat of assimilant
1400 J/kg K
Hcry—Heat of crystallization
Hfus—Heat of fusion
 The results of the EC-RAFC modeling, based on Sr-Nd isotope relationships, and Nb/Th and Nb/La ratios, are shown in Figures 9 and 10. The discussion in supporting information shows that the best fit to the Utgard data is provided by EC-RAFC trends involving assimilants with chemical characteristics typical of the lower crust. The best results are summarized in trend1, which is based on Melt2 as initial melt and a compromise between the Sr-Nd isotope composition of assimilant FAR-31 (monzonitic gneiss), and Nb/Th and Nb/La ratios somewhat higher then those of FAR-31 (ratios between those of FAR-31and eclogite 8815B). The low 87Sr/86Sr ratio (0.71212) of FAR-31 and relatively high Nb/Th and Nb/La of the assumed assimilant are typical of the lower crust. The EC-RAFC modeling thus gives strong support to the conclusion from the discussion based on fractional crystallization that Stage 1 (Figure 4) took place in the lower crust. The results are relatively similar for Melt1 and Melt2 as initial melts, whereas Melt3 does not give acceptable fits with any of the assimilants used in the modeling. Modeling based on the parameters listed in Tables 3 and 4 suggests that the extent of crustal contamination in the Utgard Sills is very minor (Figure 9). Using Melt2 (IFZ as mantle source) as initial melt, the estimates imply assimilation of less than 0.5% molten wall rock relative to the mass of the initial magma. The degree of fractional crystallization indicated by the EC-RAFC modeling is strongly dependent on the temperature difference between the initial temperature and the solidus of the assimilant (Tao and Ts, respectively, in Tables 3 and 4); modeling with Tao = 600 and Ts = 700°C gives 40% crystallization, whereas 600 and 750°C, respectively, gives 80%. An estimate of degree of fractional crystallization based on the niobium content in the least evolved Utgard Upper Sill sample (16.2 ppm Nb; Table 1), and Melt2 as initial melt (Nb = 1.47 ppm; Tables 3 and 4) proposes that ∼90% of the initial melt (F < 0.1) is removed by fractional crystallization, using bulk mineral/melt distribution coefficients for Nb in the range 0.01–0.1 (based on data by Green et al. , Norman et al. , and Severs et al. ). The degree of fractional crystallization obtained for the least evolved Lower Utgard sample (Nb = 8.2 ppm) is somewhat lower, ∼80% (F ∼ 0.2). In situ crystallization (represented by the range 16.2–22.3 ppm Nb in the Upper Sill) indicates 4–6% crystallization relative to the initial melt.
 Despite the numerous assumptions in the AFC model, we find it possible to identify realistic melt—wall rock combinations that reproduce the relationships between Sr-Nd isotope ratios and Nb-Th-La of the Utgard sills. During Stage 1, the Lower Utgard Sill may have been contaminated by crustal rocks with slightly higher 87S/86Sr ratio than the assimilant that contaminated the Upper Sill. An alternative possibility is that this difference is inherited from a heterogeneous mantle source.
 In contrast to the Utgard sills, the 642E US lavas show wide ranges in both 87Sr/86Sr, (Nb/Th)N and (Nb/La)N ratios (Figures 9 and 10) (data from Meyer et al. ). Some of the samples fall to the high-87Sr/86Sr side of the Iceland basalt field. However, the ranges of the 642E US overlap the compositions of the Utgard sills. The scatter in 87Sr/86Sr, (Nb/Th)N and (Nb/La)N ratios in the 642E US lavas must be due to contamination by rocks with other chemical characteristics than those tested by EC-RAFC modeling in this study. The origin of the 642E lavas is unclear, but it remains possible that these lavas have a mantle source similar to that of the Utgard sills with respect to Sr-Nd isotopes. The “contaminated” 642E LS appears to follow a kinked trend in the Sr-Nd isotope diagram (Figure 9), essentially following the afc1 trend in the high-143Nd/144Nd domain, whereas samples at lower 143Nd/144Nd trend toward high 87Sr/86Sr ratios (trend2). In Figure 10, the 642E LS series does not define clear trends, but show lower (Nb/Th)N ratios, and, on average, lower (Nb/La)N ratios than the 642E US. Very low (Nb/Th)N and (Nb/La)N ratios are typical of the upper continental crust. This suggests two stages of contamination caused by crustal rocks with different chemical signatures or interaction with a heterogeneous column of crustal rocks during ascent.
 The short trend formed by the NE Greenland LS in Figure 9b (along trend1) suggests that minor contamination has indeed taken place also in these lavas, although, as concluded by Thirlwall et al.  the contamination is not significant. The US, in contrast, was clearly affected by AFC processes involving significant assimilation of melts formed from crustal rocks (trend3). These lavas appear to have formed from a mantle source similar to that of the Utgard sills. The crustal contaminants that fit the Utgard sills and the “contaminated” Vøring Margin lavas (at the high-87Sr/86Sr end of trend1 and trend2, respectively); however, cannot easily reproduce the NE Greenland trend. It thus seems likely that the crustal rocks that interacted most strongly with the NE Greenland magmas are different from those that affected the Norwegian margin.
4.3. Partial Melting
 We have tested partial melting in the mantle on the basis of (Ce/Sm)N-(Sm/Yb)N relationships (Figure 12). (Ce/Sm)N and (Sm/Yb)N ratios are very sensitive to degree of partial melting, in addition to temperature, pressure, and the phase assemblage in the source rock, but are not significantly affected by moderate degrees of fractional crystallization. The Utgard sills, however, have undergone extensive fractional crystallization. The partition coefficient clinopyroxene/melt is lower for Ce than for Sm and that for Yb may be slightly higher than, or similar to, that for Sm [e.g., Green et al., 2000]. Extensive fractional crystallization will therefore cause some increase in the Ce/Sm ratio, whereas the Sm/Yb ratio will stay constant or increase. The positions of the Utgard sills in Figure 12 are therefore likely to have shifted somewhat in opposite direction to that caused by partial melting. The estimated trends shown in Figure 12 also depend on the trace element composition of the source rock, partition coefficients, and, to some degree, the proportions in which the different phases go into the melt (melting mode). Because of the many uncertainties, we have chosen not to indicate degrees of partial melting (F) along the trends; however, different Fs may be inferred from the arrows which point toward increasing degrees of partial melting. The PM source was used for modeling; ideally, we would have used IRZ (Iceland Rift Zone), but REE data other than Nd are not available. The IRZ source has a somewhat higher Nd concentration (1.61 ppm) [Debaille et al., 2009] than the PM source (1.25 ppm) [McDonough and Sun, 1995] but this difference does not preclude using PM to represent the general features of the trends depicted in Figure 12. Partial melting in the stability field of spinel peridotite produces melts in which (Ce/Sm)N ratios decrease with increasing degree of melting, whereas their (Sm/Yb)N ratio stays about the same (trend A). Partial melting in the garnet stability field produces melts that are more strongly enriched in (Sm/Yb)N than in (Ce/Sm)N. Furthermore, the degree of enrichment in (Ce/Sm)N relative to (Sm/Yb)N decreases with increasing degree of partial melting, increasing proportion of garnet in the source, increasing pressure, and decreasing fertility of the source [e.g., Walter, 1998; Simon et al., 2007]. The effects of increasing proportions of garnet in the source are shown by the trends B, C, and D (3, 6, and 10% modal garnet, respectively), the difference between partial melts formed from a depleted relative to a fertile source is reflected in the differences between solid and dotted lines. The parameters used in the calculations are listed in Table 5.
Table 4. Compositions of End-Members Used in EC-RAFC Modelling With References
N-MORB: Sun and McDonough et al. .
DMM: depleted MORB-mantle: Workman and Hart .
 Figure 5 shows that the Utgard sills have inclined patterns in the HREE domain [Dy/Yb]N = 1.4–1.5 and 1.3 in the Upper and Lower Sills, respectively. This implies garnet in the mantle source. The same is true for the NE Greenland LS lavas ([Dy/Yb]N = 1.2–1.4), whereas most 642E US lavas fall in the range (Dy/Yb)N=1.0–1.3. In Figure 12, both the 642E US and the NE Greenland LS define trends of strongly increasing (Sm/Yb)N with mildly increasing (Ce/Sm)N, but they also have off-shoots toward higher (Ce/Sm)N ratios. This is in agreement with the conclusion of Thirlwall et al.  that the NE Greenland LS lavas formed by mixing between partial melts from both garnet peridotite and spinel peridotite (trend E; Figure 12). A similar model is compatible with the 642E US trend, but the low (Dy/Yb)N ratios in many samples suggest a larger proportion of spinel-facies melts on the Norwegian side. The Lower Utgard Sill may have formed by a higher degree of partial melting than the Upper Sill, and/or contains a larger proportion of spinel-facies melts (melting at a somewhat shallower level). It is unlikely that the minor amount of crustal contamination estimated by the EC-RAFC modeling has significant effects on the Ce/Sm and Sm/Yb ratios.
 The Upper Utgard Sill has a higher (Sm/Yb)N ratio than the NE Greenland and 642E lavas (Figure 12). This may be explained by either of the following models: (i) partial melting in the garnet stability field only or (ii) mixing between melts formed from both garnet and spinel peridotites with a higher proportion of garnet-facies melts than in 642E and NE Greenland lavas. Both models imply partial melting at greater depths in the mantle. Melting at deeper levels is in agreement with the timing and position of the Upper Utgard Sill landward relative to the 642E lavas. The ages of the Utgard sills (55.6 ± 0.3 and 56.3 ± 0.4 Ma) [Svensen et al., 2010] imply emplacement during the early stages of the breakup-related volcanism, and the location of the Utgard borehole (6607/5-2) is at about 190 km from the breakup zone. Melting at relatively shallow levels to form the 642E magmas is in agreement with the general view that the lavas on the outer part of the Møre and Vøring margins extruded during the last stages of rifting and earliest stages of seafloor spreading when the lithosphere in the area had been stretched and thinned [e.g., Skogseid and Eldholm, 1987, 1989; Meyer et al., 2009]. We therefore surmise that the lithosphere in the area of the Utgard borehole was considerably thicker than that at the 642E location (Figures 1 and 2). A thicker lithosphere implies a lower geothermal gradient and thus less melting at shallow depths.
4.4. Amount of Lower Crustal Cumulates and Underplating
 The discussion above and in supporting information indicates that fractional crystallization and assimilation during Stage 1 must have taken place in magma chambers in the lower crust. This means that the magmas stayed in the lower crustal magma chambers long enough to heat the country rocks to solidus temperatures, and that this was accompanied by cooling and crystallization in the magmas. Experimental studies of crystallization sequences in different types of basaltic melts have shown that plagioclase (low density) is an early crystallizing phase at low pressures; at high pressures (0.8–1.0 GPa), crystallization is dominated by dense minerals (olivine + pyroxenes + spinel), whereas plagioclase only starts crystallizing at temperatures near the solidus; at pressures >1.0 GPa spinel forms instead of plagioclase [e.g., Green and Ringwood, 1967; Presnall et al., 1978, 2002; Villiger et al., 2004; Falloon et al., 2008]. The Upper Utgard Sill shows trends of strongly decreasing concentrations in Ni, Sc, and Cr with decreasing MgO (Figure 4). This is evidence of extensive fractionation of olivine + pyroxenes, and may have involved Cr-spinel (but not magnetite). The Upper Utgard Sill is also significantly richer in Sr than the NE Greenland and 642E lavas (Figure 4). This means that although the Utgard sills show small negative Sr and Eu anomalies (Figure 5) crystallization of plagioclase must have been insignificant before intrusion into the upper crust where the petrography shows plagioclase to be a major crystallizing phase. We conclude that a large portion of the fractional crystallization that gave rise to the final Utgard magmas occurred in the lower crust (Stage 1). In experiments on tholeiitic melts at 1.0 GPa, Villiger et al.  found plagioclase to start forming only after >50% crystallization (equilibrium or fractional) of olivine, clinopyroxene, orthopyroxene, and spinel. In transitional magmas, like the Utgard melts, crystallization of clinopyroxene is even more dominant than in tholeiitic melts. Based on these experiments, we propose that crystallization in the Utgard melts took place partly in the lower crust, Stage 1, and partly in situ in the upper crust, Stage 2. Crystallization in the lower crust comprises two steps, first removal of olivine + pyroxenes + spinel forming dense olivine pyroxenite cumulates (Stage 1a), followed by removal of ± olivine + pyroxenes + plagioclase ± spinel forming gabbros (Stage 1b). The evolutionary sequence outlined above is shown schematically in Figure 13.
 This conclusion bears on the interpretation of the extensive high-velocity, high-density lower crustal body (LCB) identified in the deep crust beneath the mid-Norwegian margin (Figure 2) and frequently discussed. Proposed models include underplating by dense magmatic material during rifting and breakup [e.g., Mjelde et al., 2002; Voss and Jokat, 2007], stacked sill complexes in the lower continental crust [White et al., 2008], high-grade metamorphic rocks formed during the collapse of the Caledonian mountain range [Gernigon et al., 2003, 2004; Ebbing et al., 2006], older, high-pressure granulite/eclogite bodies [e.g., Gernigon et al., 2004, 2006; Mjelde et al., 2009], and serpentinized mantle [e.g., Reynisson et al. 2010]. A detailed review of the different hypotheses was recently presented by Reynisson et al. .
 Our results imply that for each volume unit of sill or lava on the Norwegian margin, there is a large mass of cumulates left in the lower crust. Based on density data, we have made some very simplified estimates of the average thickness of cumulate bodies that may be associated with the Utgard sills and with other sills and lavas in the Vøring Basin, assuming that the cumulate bodies have the same areal extent as the sills/lava bodies with which they are associated. The parameters used are discussed below and the results listed in Table 6.
Table 6. Estimates, Based on Densities, of the Thicknesses Per Areal Unit (e.g., km2) of Cumulates in the Lower Crust Corresponding to the Thicknesses of Sills or Lavas Per Square Unit in the Upper Crust
Thickness per areal unit of sill intrusion or lava in the upper crust.
Berndt et al.  identified 6–7 sills with a maximum thickness of 100 m in the Hel Graben. We show estimates assuming six sills each 50 m thick (Min. thick.), and seven sills each 100 m thick (Max.thick.).
Less evolved (Figure 4), therefore a different lava/cumulate proportion.
Vierick et al. .
Upper Series 770 m thick, Lower Series 140 m thick.
Minimum value, see text for explanation.
Eldholm et al.  and Skogseid and Eldholm .
 We showed above that the amount of fractional crystallization varies in the models examined. The EC-RAFC modeling indicates that 70% of the original mass of the parent magma precipitated as cumulates (Mc) to produce the Utgard magmas (Stage 1 + Stage 2). Estimates for Mc depend on the temperature difference (ΔT) between the initial and the solidus temperature of the assimilant. We used ΔT = 10°C in the EC-RAFC modeling, higher ΔT gives more extensive fractional crystallization (higher Mc) before the onset of assimilation. Estimated degree of fractional crystallization based on Nb concentrations indicate that Stage 1 involves ∼90% fractional crystallization to form the least evolved Upper Sill sample and ∼80% for the least evolved Lower Sill. In the following calculations, we have used much more conservative degrees of fractional crystallization, 60 and 55% crystallization for Stage 1 in the Upper and Lower Utgard Sills, respectively.
 Fractional crystallization in the Utgard magmas during Stage 1 (see supporting information) was dominated by olivine + clinopyroxene ± orthopyroxene. Olivine and pyroxenes in the asthenospheric mantle typically have Mg/(Mg+Fe2+) ratios (mg#) of 0.85–0.91 [e.g., Simon et al., 2008]. The first olivine and pyroxene to form in primary mantle melts will have the same mg# ratio as the mantle host, but this will decrease during fractional crystallization. Assuming an average Mg/(Mg+Fe2+) ratio of 0.8 for olivine and pyroxenes in the Utgard cumulates, density data on mineral end members [Robie et al., 1966] give average densities for olivine, clinopyroxene, and orthopyroxene of 3.45, 3.33, and 3.40 g/cm3, respectively. The presence of Al-bearing end-members in clinopyroxene does not change the given average significantly. Experimental data [Presnall et al., 1978] imply that at 1.0 GPa mantle melts will have a first stage of olivine crystallization followed by coprecipitation of olivine + clinopyroxene in the proportion 20:80. We have chosen an average olivine:clinopyroxene ratio of 30:70 for these two stages and ignored orthopyroxene as the proportion of orthopyroxene is difficult to assess. Such an olivine-clinopyroxene mixture gives an average density of 3.37. Lower average Mg/(Mg+Fe2+) ratios, a significant proportion of orthopyroxene and/or the presence of titanomagnetite (∼5 g/cm3), Al-Cr spinel (3.6–4.4 g/cm3), and/or garnet (3.4–4.3 g/cm3) in the cumulate increase the average density. However, the cumulates will also contain gabbroic material formed from melts trapped in interstices between cumulate minerals. Hyndman and Drury  report averages of 2.795 g/cm3 for basalts and 2.957 g/cm3 for gabbros in the oceanic crust. We have chosen the intermediate value of 2.8 g/cm3 for trapped material in the cumulates. Tegner et al.  found densities of 3.0–3.3 g/cm3, and the proportion of trapped liquid to be 3–47% in low-pressure cumulates with high proportions of plagioclase (2.6–2.8 g/cm3) [Robie et al., 1966] and Fe-Ti-oxides in the Skaergaard intrusion (east Greenland). The density of a cumulate decreases with increasing proportion of trapped liquid. We have chosen 20% trapped liquid for our estimates. This suggests an average density of roughly 3.24 g/cm3 for lower crustal cumulates (cumulate minerals + interstitial material) produced from a primary magma that gave rise to the Utgard sills (Table 6).
Berndt et al.  found a mean density of 2.75 g/cm3 (and a mean velocity of 7.0 km/s) in the Upper Sill. The mass proportion of 40% dolerite with a density of 2.75 g/cm3 relative to 60% cumulates with an average density of 3.24 g/cm3 gives a volume proportion of 1.28 and 1.22 units (for the Upper and Lower sills, respectively) of cumulates relative to 1 unit sill (Table 6). Assuming the same areal extent, the formation of the two Utgard sills with a combined thickness >140 m must have given rise to a >177 m thick layer of cumulate crystals. Including 20% trapped liquid, the total thickness of cumulates becomes >212 m. Furthermore, as our estimated cumulate density is higher than the average density of 3.10 g/cm3 estimated by geophysics for the LCB beneath the Vøring Plateau [Ebbing et al., 2006; Reynisson et al., 2010], the cumulates could be mixed with significant volumes of rocks with densities <3.10 g/cm3, for example, old continental crust. Ebbing et al. [2006, and references therein] gave densities of 2.95–3.00 g/cm3 for the old continental crust in the Vøring Plateau; we chose the value of 2.95 g/cm3 (Table 6) for our estimates. With an average density of 3.25 g/cm3 for the Utgard cumulates, cumulates and old crustal basement would have to be mixed in the approximate proportion 2:1 in order to match a bulk density of 3.10 g/cm3. This implies that the Utgard sills may account for a >320 m thick layer of mixed cumulate + crustal basement in the LCB (Table 6). In addition, the fractional crystallization in Stage 1b has given rise to some gabbros.
 Seismic surveys have not found the LCB beneath the 6607/5-2 Hole in the Utgard High (Figure 2). However, this hole only penetrates the easternmost part of the Utgard sills, and the LCB extends eastwards to just west of the hole and is thus present beneath most of the E-W range of these sills. We therefore do not see any disagreement between the seismic data and our results.
 In the upper crust beneath the Hel Graben, Berndt et al.  (Figure 2) have identified six to seven sills, each with a maximum thickness of 50–100 m. Table 6 shows two estimates for the amount of mixed cumulates and basement associated with these sills, one based on the assumption of six 50 m thick sills, the other assuming seven 100 m thick sills. Using the same chemistry and proportion of trapped melt as for the Utgard sills, the Hel Graben sills may account for a 700–1600 m thick mixed cumulate and basement layer in the LCB.
 The Upper and Lower 642E lava series on the Vøring Platform are about 770 and 140 m thick, respectively [Viereck et al., 1998, 1989]. The average MgO contents in the Upper Series are considerably higher than in the Utgard sills (Figure 4). We therefore assume a significantly lower average degree of fractional crystallization (50%) for the 642E lavas than for the Utgard sills, and consequently a somewhat higher Mg/(Mg+Fe) ratio for the mafic phases (0.85). The higher MgO content of the lavas and cumulates give a lower average cumulate density, 3.23 g/cm3, and a volume proportion of lavas:cumulates of 1:0.87. On this basis, the estimated thickness of cumulates mixed with basement rocks associated with the drilled volcanics of the Vøring Plateau is ∼1400 m. The thickest lava sequence on the Vøring Plateau, however, has been estimated to exceed 6 km [Eldholm et al., 1987; Skogseid and Eldholm, 1987]. Assuming the same parameters as for the lavas recovered in Hole 642E, the mixture of cumulates and old basement rocks beneath the thickest lava sequence on the Vøring Plateau must be at least 8.8 km thick. This result is slightly higher than that based on wide-angle seismic data indicating the LCB in this area is ≤8 km [e.g., Mjelde et al., 2002], and somewhat lower than a model based on structural observations and thermokinematic modeling which gave a maximum thickness of 11 km for the LCB beneath the Vøring Margin [Gernigon et al., 2006].
 The estimates summarized in Table 6 strongly depend on the choice of input parameters (degree of fractional crystallization, proportion of trapped melt in the cumulates, sill thicknesses, etc.). However, we have systematically chosen parameters, e.g., significantly lower degree of fractional crystallization than indicated by EC-RAFC modeling and estimates based on Nb, high Mg/(Mg+Fe) ratios, and a high proportion of trapped melt in the cumulate minerals that give low estimates of cumulate densities and mixed cumulate-basement complexes. In spite of the fact that the estimates given in Table 6 represent oversimplifications, we regard our results as evidence that significant parts of the LCB can be explained as a heterogeneous mixture of dense cumulates associated with the opening-related magmatism and less dense rocks, such as old continental basement.
 A model involving a mixture of rocks with contrasting physical properties (for example, opening-related cumulates and old continental crust) is in agreement with the large variations in VP (7.1–7.8 km/s) and relatively low VP/VS ratios (1.75–1.78) documented within the LCB by Mjelde et al. . Our model is also in agreement with the conclusions of Wangen et al. . They studied the nature of the LCB beneath the western Vøring Margin on the basis of three scenarios related to the extension history (a) only Caledonian crust; (b) half Caledonian crust and half magmatic underplating; (c) only magmatic underplating, and found model (b) to be most likely. However, our results indicate that the LBC includes opening-related gabbros, and do not exclude the possibility that the LCB beneath the Vøring Plateau also involves other rock types, such as older igneous material, high-grade metamorphic rocks (granulites or eclogites), and/or serpentinized mantle [e.g., Gernigon et al., 2003, 2004, 2006; Ebbing et al., 2006; Mjelde et al., 2009; Reynisson et al., 2010].
5. Evolutionary History
 Based on major and trace element and Sr-Nd isotope data, we propose the following evolutionary history for the Utgard sills (summarized in Figure 13) and their relationships with the 642E and NE Greenland lava series.
 The primary Utgard melts originated by partial melting of an asthenospheric mantle source depleted with respect to Sr-Nd isotope and trace element compositions. Its chemical characteristics appear to be similar to those of the source that gave rise to the NE Greenland lavas (IRZ). The wide ranges in 87Sr/86Sr, (Nb/Th)N, and (Nb/La)N ratios among the 642E Upper Series lavas prevent conclusions about their mantle source. However, a similar mantle source as for the Utgard sills and the NE Greenland lavas is possible and likely. The slightly different Sr-Nd isotope compositions of the two Utgard sills most likely reflect heterogeneities in the mantle source.
 The primary melts that gave rise to the Utgard sills formed either by partial melting in the garnet stability field or by mixing between melts from both garnet-facies and spinel-facies peridotites, similar to the model proposed for the NE Greenland lavas [Thirlwall et al., 1994]. In any case, the melting dynamics associated with the Upper Utgard Sill appears to have involved a higher garnet proportion in the source, showing that the Utgard magmas formed at greater depth than the 642E and NE Greenland lavas. The Utgard sills were emplaced at a great distance from the continent-ocean transition during the early stages of the breakup-related volcanism, whereas the 642E lavas extruded on the outer Vøring Plateau during the last stages of rifting and earliest stages of seafloor spreading. This indicates that at their times of emplacement the lithosphere was significantly thicker beneath the Utgard sills than beneath the 642E lavas.
 The magmas that gave rise to the Utgard sills appear to have ascended through the subcontinental lithospheric mantle (SCLM; Figure 13) without significant interaction with the wall rocks. When entering the less dense lower crust, the melts lost much of their buoyancy and ponded to form magma chambers where the melts were subjected to AFC processes (Stage 1). Our modeling indicates this stage involved extensive fractional crystallization in the lower crust, mainly involving removal of dense the phases olivine and pyroxenes. Fractional crystallization in the lower crust also gave rise to minor amounts of gabbros. The total extent of fractional crystallization during Stage 1 is estimated to be at least 80% relative to the initial magma mass. Furthermore, processes in the lower crust included minor assimilation of crustal melts (<0.5%). The anatectic crustal melts appear to have had 87Sr/86Sr ratios ≤0.715, (Nb/Th)N of ∼1 and (Nb/La)N between 0.5 and 1.0 (Figures 9 and 10). Rocks with similar properties are found in the West Norwegian Gneiss region.
 The evolved residual melts finally ascended to the upper crust where they formed the Utgard sills. A new stage of fractional crystallization (Stage 2: 4–6% relative to the original magma mass; Figure 13) occurred in situ.
 In the NE Greenland LS lavas fractional crystallization is, on average, less extensive than in the Utgard sills and crustal contamination is insignificant. The NE Greenland US is significantly contaminated by crustal rocks although different from those that contaminated the Utgard sills. The 642E LS lavas show a two-stage contamination history. The first stage involves strongly decreasing 143Nd/144Nd with moderately increasing 87Sr/86Sr, compatible with a lower crustal assimilant. The second stage involves typical upper crustal contaminants (high 87Sr/86Sr and moderately low 143Nd/144Nd, (Nb/Th)N<<1, (Nb/La)N<<1; Figures 9 and 10). The wide ranges in Sr isotope ratios and ratios between incompatible elements among the 642E lavas suggest contamination by assimilants with a variety of chemical characteristics.
 The extensive fractional crystallization that affected the parent melts of the Utgard sills imply that significant proportions of the parent magma was left as cumulates in the deep crust (underplating). Assuming an areal extent similar to that of the Utgard sills, these cumulates, with an estimated average density of 3.23–3.25 g/cm3, form a >210 m thick layer. In order to obtain the average density of 3.10 g/cm3 estimated for the LCB by geophysics, the cumulates may be mixed with less dense rocks. A mixture between cumulates and old continental lower crust (2.95 g/cm3) makes a >320 m thick layer beneath the Utgard sills with the average density of 3.10 g/cm3. Significant volumes of dense cumulates in the lower crust must also exist beneath other opening-related magmatic complexes on the Norwegian margin. A layer with an average density of 3.10 g/cm3, consisting of cumulates and old continental lower crust, would have a thickness between 700 and 1600 m beneath the Hel Graben, >1.4 km beneath Hole ODP Leg 104 Hole 642E, and >8.8 km beneath the thickest part of the lava sequence on the Vøring Plateau, slightly higher than the estimate of Mjelde et al.  (Table 6). We thus argue that opening-related cumulates make up a significant part of the LCB. The LCB also includes gabbros formed during the last part of the crystallization in the lower crust (Stage 1b; Figure 13), and may comprise other rock types such as old sill complexes, eclogites, and/or serpentinized peridotites.
 We are indebted to Romain Meyer who gave us access to new geochemical data on lavas from ODP Leg 104 Hole 642E lavas before publication. We thank the Norwegian Petroleum Directorate for access to samples from the Utgard borehole. This work was financed by a YFF grant to H. Svensen and a SFF grant to PGP (Physics of Geological Processes), both from the Norwegian Research Council. The paper benefited from the constructive reviews of Godfrey Fitton, Reidar G. Trønnes, and Tod Waight.