Mechanisms of crustal growth in large igneous provinces: The north Atlantic province as a case study
Published:January 01, 2007
- PDF LinkChapter PDF
Laurent Geoffroy, Charles Aubourg, Jean-Paul Callot, Jean-Alix Barrat, 2007. "Mechanisms of crustal growth in large igneous provinces: The north Atlantic province as a case study", Plates, Plumes and Planetary Processes, Gillian R. Foulger, Donna M. Jurdy
Download citation file:
The mechanisms of magma crust accretion at large igneous provinces (LIPs) are questioned using arguments based on the north Atlantic case. Published and new data on the calculated flow vectors within dike swarms feeding the early traps and subsequent seaward-dipping reflector lavas suggest that most of the mafic magmas forming the north Atlantic LIP transited through a small number of igneous centers. The magma was injected centrifugally in dike swarms at some distance away from individual igneous centers along the trend of the maximum horizontal stress acting in the crust, feeding lava piles via dikes intersecting the ground surface. This mechanism is similar to that observed in present-day Iceland and, more generally, in mafic volcano-tectonic systems. The absence of generalized vertical magma transit in a LIP has major geodynamic consequences. We cannot link the surface extent of LIP magmas to the dimensions of the mantle melting zone as proposed in former plume head models. The distribution of LIP magmas at the surface is primarily controlled by the regional stress field acting within the upper crust, but is also affected by magma viscosity. The igneous centers feeding LIPs most likely represent the crustal expression of small-scale convective cells of the buoyant mantle naturally located beneath the mechanical lithosphere.
This contribution focuses on the localization of upper-mantle melting at the origin of large igneous provinces (LIPs). We need first to summarize the different views on LIP origin, which is a subject of great debate (see the Web site www.MantlePlumes.org).
Two distinctive stages of development are recognized in LIPs: the widespread emplacement of flood basalts (the “trap stage”) and a (possible) consecutive “volcanic margin stage” during which synextension magmatism is concentrated along the breakup zone (e.g., White and McKenzie, 1989; Eldholm et al., 1995; Courtillot et al., 1999; Geoffroy, 2005). It is important to bear in mind the existence of these two developmental stages insofar as the geographic distribution and volumes of magma are different in each of them.
Traps or plateau basalts are flat-lying accumulations of mafic lavas that are emplaced during a relatively short time span. Some authors have used sedimentary records (White and Lovell, 1997) and wide-angle seismic surveys (Al-Kindi et al., 2003) to argue that magma underplating may also occur during this early stage. During trap emplacement, the amount of tectonic extension is usually small, and dilatation through dike injection seems to predominate over extension associated with normal faults (e.g., Doubre and Geoffroy, 2003).
Many authors have explained the large uplifted oceanic and/or continental areas covered with plateau basalts, as well as oceanic hotspots located at the crest of broad seafloor swells, by the presence of more or less axisymmetric hot mantle plume heads beneath the lithosphere (e.g., Morgan, 1971; Courtney and White, 1986; Olson and Nam, 1986). White and McKenzie (1989) summarize the geological and geophysical features that are generally linked with postulated mantle plumes beneath the lithosphere. Mantle plumes are primarily thought to represent hot gravitational instabilities formed at the core-mantle boundary due to a core-mantle thermal boundary layer (e.g., Anderson, 2004). In parallel, the enriched trace-element geochemistry of early traps at LIPs and oceanic islands basalts (OIB), as well as their rare gas isotopic ratios, is commonly explained by the postulated primitive composition of a lower-mantle reservoir (e.g., Courtillot et al., 2003). Many models of plume head–lithosphere interaction have been discussed, such as: (1) sudden impact of a very hot plume (e.g., Richards et al., 1989), (2) progressive thermal erosion of the basal lithosphere by a long-lived incubating plume head (e.g., Kent et al., 1992), (3) interaction of a plume head with a lithosphere of variable thickness (e.g., White and McKenzie, 1989; Thompson and Gibson, 1991), (4) small-scale convection within the plume head (e.g., Fleitout et al., 1986).
However, a range of new concepts and experiments has challenged the mantle plume theory (see the extensive references at www.MantlePlumes.org). The reinterpretation of mantle seismic tomography raises questions about deep-seated mantle plumes, as exemplified by the Icelandic case (e.g., Foulger, 2002). The chemistry of early flood basalts and OIB could also be explained by melting of a much shallower compositionally heterogeneous mantle (e.g., Gallagher and Hawkesworth, 1992; Anderson, 1994). Notably, melting of eclogites (old subducted slabs) is proposed as a possible component in igneous provinces developed along ancient orogenic sutures (e.g., Foulger and Anderson, 2005). Hotter than normal mantle is also debated as a cause of LIP magmatism (e.g., Green et al., 2001). Different top-to-down processes have been proposed as an alternative to hot plumes, most of them invoking, although at quite different scales, upward counterflow processes in the mantle due to the gravitational instability of cold lithospheric roots (e.g., King and Anderson, 1998; Lustrino, 2005).
Most volcanic passive margins (VPMs, Fig. 1) are consecutive to the emplacement of traps and associated with anomalously thick oceanic crust following continental breakup. The main crustal characteristics of VPMs are listed in White et al. (1987), Eldholm et al. (1995), Bauer et al. (2000), and Geoffroy (2005) and include data from both geophysical and geological surveys (Fig. 1). These margins are associated with significant magma accretion during lithosphere extension or rifting and subsequent breakup. Interestingly, the top-to-down trilogy of basalts, sheeted complex, and gabbros postulated for the VPM crust is strongly analogous with the structure of oceanic crust (Fig. 2). From the surface down to the Moho, the VPM crust (Fig. 1) is composed of (1) several wedges of seaward-dipping volcanic rocks (otherwise known as seismic SDR), (2) intensively intruded and stretched continental crust, and (3) large volumes of material of high compressional wave velocity (V p) usually interpreted as underplated high-Mg gabbros. SDR are made up of aerial or subaerial lavas, but also contain volcanic ejecta (e.g., hyaloclastites, tuffs, etc.). The tectonic significance of SDR has been discussed, principally in Eldholm et al. (1995) and Geoffroy (2005). It is noteworthy that an inner SDR prism is usually located immediately above the stretched and intruded continental crust (e.g., Roberts et al., 1979; Planke et al., 2000; Geoffroy, 2005; Fig. 1).
Mantle plume specialists believe that VPMs originate from continental breakup over the hot plume head or residual tail that remains after trap emplacement (e.g., White et al., 1987; White and McKenzie, 1989; Courtillot et al., 1999). However, small-scale convection due to the rifting process itself and/or pre-existing lateral variations in lithosphere thickness (e.g., Mutter et al., 1988; Keen and Boutilier, 1995; King and Anderson, 1998) could also account for the huge volumes of magma associated with VPMs without invoking any excess in mantle temperature. From numerical models of rapidly stretched continental lithosphere, Van Wijk et al. (2001) point out that the characteristic melt thickness of VPMs may be obtained solely through adiabatic melting of the sublithospheric buoyant mantle immediately before plate breakup. Anderson (1994 and 1995) proposed that the huge volumes produced at VPMs are the result of pull-apart processes over the fertile part of the mantle located beneath the mechanical lithosphere (called the “perisphere” or thermal boundary layer in the present article). All these non-plume models assume that plate tectonics and plate geometry (craton edges) are sufficient by themselves to account for the origin of VPMs and LIPs as a whole.
Melting Localization at LIPs
A fundamental aspect of LIP buildup concerns the localization of melting areas beneath or within the lithosphere. This problem has commonly been addressed through magma geochemistry. However, as outlined earlier, it is not always straightforward to characterize the mantle reservoirs involved (i.e., lithospheric or asthenospheric mantle) because of the contamination of the melts by various crustal lithologies as illustrated by the Scottish Tertiary volcanics (e.g., Dickin et al., 1987, among others). Another approach to discussing the origin of the igneous activity is to establish the pattern of magma flow within the oceanic or continental crust. This was performed at the scale of a Proterozoic LIP by Ernst and Baragar (1992). From an AMS (anisotropy of magnetic susceptibility) study in dikes, these authors argued that the McKenzie giant dike swarm (and related LIP) was fed centrifugally and horizontally in the continental crust from a central source (toward which the dikes converge) that was itself fed vertically from a mantle plume. Later, in the section on flow vectors, we return to the AMS methodology followed by these authors to infer the flow pattern in dikes. We should note that Phanerozoic LIPs are generally not associated with giant dikes (see the section on dike swarms and stress fields), and that no studies on the more recent LIP are specifically concerned with magma transfer from the mantle to the upper crust. Instead, in most LIP plume models, it is assumed either implicitly or explicitly that “the extent of [flood basalts] gives a good indication of the area underlain by the mushroom head of hot mantle carried up by the plume” (White and McKenzie, 1989). The proposed diameters for melting plume heads could thus reach values of up to 2000 km (White and McKenzie, 1989; Hill et al., 1992).
The question of LIP feeding is also addressed by authors defending models that do not involve mantle plumes. Denying the existence of dike swarms that radiate from a single point, they propose that traps at LIPs are fed upward through regional dike swarms whose location is controlled solely by plate-related lithospheric stresses (e.g., Favela and Anderson, 1999; McHone et al., 2004), not by a plume-related stress field (i.e., radial due to a lithospheric swell, as in Ernst et al., 1995).
Thus, the most commonly held view of LIP formation involves a vertical transfer of magma from extensive zones of melting in the mantle to the upper crust (intrusions) or the Earth's surface (lavas, volcanic ejecta). This transfer is assumed to be direct (primary magmas) or indirect (magmas differentiated in crustal reservoirs). Inferring that the area covered by flood basalts corresponds more or less to the extent of mantle melting at depth is an important assumption, because it determines the whole geodynamic model for the origin of LIPs. In the present study, we make use of the north Atlantic case to question this assumption by establishing the actual mechanisms of magma accretion during LIP growth.
THE NORTH ATLANTIC PROVINCE: WHAT KIND OF MAGMA FEEDING SYSTEM?
Volume Estimates for the North Atlantic LIP
In the north Atlantic volcanic province (Figs. 3 and 4), the earliest oceanic magnetoanomaly is dated as C24r, i.e., ca. 56 to 54 Ma according to the Berggren et al. (1995) timescale. The emplacement of Paleocene flood basalts spanned ∼6 m.y., from 62 to 57 or 56 Ma (Hitchen and Richtie, 1993). Ar-Ar ages from the East Greenland Coastal Dike swarm (Lenoir et al., 2003) indicate that the breakup is bracketed between 54 and 51 Ma, which is consistent with a sudden Eocene breakup event, coeval with the earliest oceanic accretion.
Coffin and Eldholm (1994) and Eldholm and Grue (1994) estimated the minimum volumes of magma in the north Atlantic volcanic province (including initial traps and VPMs) at no more than 6 × 106 km3, with rates of magma production reaching 2 km3/yr. However, such an estimate is difficult to establish and should be considered only as a maximum value. For instance, most authors favoring the plume hypothesis unhesitatingly assume that west Greenland, east Greenland, the Faeroes, and the British Tertiary igneous province are parts of the same flood basalt province (e.g., Saunders et al., 1997; Larsen et al., 1999). This shortcut hypothesis is debatable, because there is no continuity of outcrop between west and east Greenland. Estimating the volume of underplated mafic magmas from the high-velocity zone (HVZ; Figs. 1 and 2) may also lead to unreliable results. Most authors agree that the HVZ represents substantial amounts of magma accreted (“underplated”) at the Moho (e.g., Eldholm and Grue, 1994; Holbrook et al., 2001). However, Gernigon et al. (2004) have challenged the HVZ magma interpretation beneath the Voring Mesozoic basin. Should their observations be correct and applicable to other sedimentary basins, this would considerably decrease the estimated magma volume in the north Atlantic volcanic province.
Igneous Center Distribution in the North Atlantic Igneous Province
An important characteristic of LIPs is the ubiquity of igneous centers punctuating the nonrifted and rifted continental or transitional crust. In the uppermost crust, igneous centers are represented by magma chambers and overlying hypovolcanic intrusions, making up the roots of large polygenic volcanoes (e.g., Vann, 1978; Irvine et al., 1998; Bauer et al., 2000; Chandrasekhar et al., 2002). In the north Atlantic, these igneous centers are well known from direct observation and/or potential field data (Fig. 4). To a first approximation, the large subcircular or elliptical magnetic and gravity anomalies associated with the igneous centers can be modeled by cylinder-shaped bodies of mafic to ultramafic rocks extending down to the Moho (e.g., Bott and Tuson, 1973; Bott and Tantrigoda, 1987).
The internal structure of these bodies is unknown. For instance, Bauer et al. (2000) proposed a crustal-scale interconnected network of mafic planar intrusions for the Messum Igneous Complex in Namibia. In the north Atlantic volcanic province, at least thirty-eight igneous centers were probably active during the trap stage (Fig. 4). Although not all igneous centers have yet been recognized and/or dated, their distribution appears to be (1) 2-D in map view and (2) unrelated to the thickness of the crust (Fig. 4). According to Callot (2002), the spacing between off-shore trap-stage igneous centers would vary from ∼75 ± 30 km in the Hatton area to 100 ± 40 km in the Rockall–Faeroe Bank area (Fig. 4). This spacing decreases locally and significantly in the British Tertiary igneous province (35 ± 3 km for the Sy-Mu-Am-Mu group (Callot, 2002). In the latter case, the distribution of igneous centers is evidently controlled by the location of the main Caledonian-inherited discontinuities that were reactivated during the Tertiary (Fig. 4B; e.g., Roberts, 1974).
Many igneous centers are also associated with the north Atlantic volcanic province breakup process, so they are seen to punctuate the volcanic margins (e.g., Barton and White, 1997; Korenaga et al., 2000; Callot, 2002; Callot et al., 2002; Callot and Geoffroy, 2004). Apart from the eroded along-strike exposures of the innermost parts of a VPM, as observed on the southeast Greenland coast (Figs. 4 and 5; Myers, 1980; Bromann-Klausen and Larsen, 2002), the igneous centers associated with the breakup stage are less easy to distinguish physically due to their lower gravity and magnetic contrasts with the enclosing transitional or igneous crust (Fig. 5). However, the presence of relative gravity highs (Fig. 5) and magnetic anomalies (Gac and Geoffroy, 2005) suggest that igneous centers in the VPM transitional crust display an aligned or zigzag 1-D pattern or else a 2-D arrangement within a narrow band (Callot et al., 2002). Callot (2002) measured a 155 ± 7 km spacing of igneous centers in on-shore east Greenland (an area of low to moderate crustal thinning), decreasing to 58 ± 12 km off-shore, where the lithospheric thinning was greatest (Fig. 5 and related caption). This latter value compares well with the wavelength of gravity and magnetic segmentation observed along the U.S. East Coast VPM at the continent-ocean transition (Behn and Lin, 2000).
It is noteworthy that, for at least two north Atlantic volcanic province igneous centers (Rosemary Bank and Anton Dohrn; Fig. 4), the magmatic activity is likely to be Late Cretaceous in age (e.g., Jones et al., 1974; Hitchen and Richtie, 1993), thus predating the postulated Paleocene emplacement of the so-called Icelandic mantle plume. In addition, magmatic activity at igneous centers is often a persistent phenomenon. In some centers that were active during trap emplacement, highly differentiated magma continued to be intruded during and even after the breakup process, sometimes at great distances from the volcanic margins (for example, the end-Eocene granites of the Mourne, Skye, and Lundy igneous centers; Fig. 4) (Hitchen and Richtie, 1993; Saunders et al., 1997).
There is general agreement that most LIP volcanism is of subaerial and fissural type. Low-viscosity tholeiitic or intermediate lava flows were fed by dikes intersecting the ground surface as in present-day Hawaii, Afar, or Icelandic volcanic systems (e.g., Self et al., 1997). Basaltic tuffs result from the dikes themselves (monogenic cones along fissures) or can be produced by ash eruptions from igneous centers related to polygenic volcanoes. It is thus evident that igneous centers and dike swarms play an essential role in distributing magmas within LIPs, not only at the ground surface (lavas and ejecta), but also within the intruded continental crust (magma crystallizing within the dikes themselves and in the igneous centers, hypovolcanic complexes, and magma chambers).
Dike Swarms and Stress Fields
In the north Atlantic volcanic province continental domain, only a small number of off-shore and on-shore dike swarms have been identified from aeromagnetic surveys and direct observations, respectively (Fig. 3) (e.g., Larsen, 1978; Myers, 1980; Speight et al., 1982; Kirton and Donato, 1985; Bromann-Klausen and Larsen, 2002). For some off-shore areas, however, high-resolution aeromagnetic data may be missing or remain untreated. None of the recognized dike swarms correspond to giant dike swarms (i.e., swarms of dikes exceeding ∼30 m in thickness): in the whole north Atlantic volcanic province area, outcropping dikes (even through eroded basement) exhibit an average thickness rarely exceeding 2 m during the trap stage (e.g., Speight et al., 1982) and ranging from 3 m (highly stretched crust) to 8 m (weakly stretched crust) along the VPM (Bromann-Klausen and Larsen, 2002). Notable exceptions include individual intrusions such as the ∼20-m-thick Cleveland Dike (Scotland; McDonald et al., 1988) and the ∼600-m-thick Kraemer Island dikelike intrusion (east Greenland, Kangerlussuaq area).
Dike swarms trend parallel or subparallel to the maximum horizontal stress (e.g., Anderson, 1951), so the general pattern of the north Atlantic volcanic province dike swarms may reflect the stress field in this area during the Paleogene (Fig. 3A). However, this stress field was not uniform. A regional northwest-southeast-trending maximum stress during the Paleocene was associated with prebreakup trap emplacement in the British Tertiary igneous province (e.g., Vann, 1978) and in the Faeroes area (Geoffroy et al., 1994; Fig. 3B). At the scale of the north Atlantic volcanic province, the maximum horizontal stress was apparently radial and focused on the Kangerdlussuaq area, which contains outcrops of a large system of igneous intrusions (Fig. 3A). It should be noted that most of the observed dike swarms in the north Atlantic volcanic province are centered on individual igneous centers (Fig. 3B). This is well established both in the trap area (e.g., Vann, 1978; Speight et al., 1982) and on the exposed parts of the VPM (Myers, 1980; Bromann-Klausen and Larsen, 2002). In all cases studied, the finite horizontal magma dilatation associated with these swarms increases toward the igneous centers.
Therefore, in the north Atlantic volcanic province (and generally in LIPs), magma transport in the brittle crust follows specific flow paths. The covering of vast continental areas by repetitive lava flows and volcanic products coming from a small number of fissure systems is not a hypothesis but a fact. The formation of dikes also plays a significant role in the magmatic accretion of the transitional crust (Figs. 1 and 2). Although this mechanism is now well accepted, it still needs to be firmly integrated into a mantle or lithosphere model for Phanerozoic LIPs, because the LIP magma feeding system is evidently connected with the distribution of mantle melting at depth. Some authors have proposed that dikes are fed vertically from mantle ridge structures (or linear thinned zones) that are undergoing partial melting and follow the same trend as the swarms (e.g., Speight et al., 1982). Such a view is also implicit in the work of Al-Kindi et al. (2003). Others, such as White and McKenzie (1989), suggest that the magma migrates upward from a more or less homogeneously subcircular melting mantle plume head (see a variation of this plume model by introducing preexisting lithospheric thin spots into Thompson and Gibson, 1991, and Nielsen et al., 2002).
In the present study, we discuss these views using a statistical analysis of the flow vectors in selected dike swarms emplaced during both the trap and the SDR stages of north Atlantic volcanic province evolution.
MEASURING FLOW VECTORS IN DIKES AND DIKE SWARMS
Magma Flow Vectors Estimated by AMS
In the field, it is often difficult to determine with precision the fossilized flow vector in dikes. This is due to the scarcity of observable flow indicators both inside (e.g., oriented phenocrysts, elongated gas vesicles) and along the walls of the intrusions (e.g., mechanical lineations; Baer and Reches, 1987). Our method for studying magma flow in the north Atlantic volcanic province dikes is based on AMS. The AMS technique consists of determining the maximum, intermediate, and minimum principal axes (K1, K2, and K3, respectively) of the magnetic susceptibility ellipsoid of a rock sample submitted to a weak magnetic field (for an explanation of the technique, see Rochette et al., 1991). The application of AMS to the petrofabric study of basaltic dikes has been extensively discussed (e.g., Ellwood, 1978; Knight and Walker, 1988; Hargraves et al., 1991; Rochette et al., 1991; Ernst and Baragar, 1992; Staudigel et al., 1992; Baer, 1995; Varga et al., 1998; Aubourg et al., 2002). Briefly, the magnetic foliation in basalts (i.e., the plane containing axes K1 and K2, perpendicular to K3) is considered to reflect the distribution of ferromagnetic oxides in the rock mass. Depending notably on the time of appearance of these grains during differentiation of the magma, their distribution is thought to correspond directly (e.g., Borradaile, 1988) or indirectly (Hargraves et al., 1991) to the fossilized magmatic foliation (i.e., a flow plane; see Nicolas, 1992). A number of authors assume that the AMS axis K1 (i.e., the magnetic lineation) yields the orientation of the long axes of multidomain magnetic grains (i.e., Borradaile, 1988). In dikes, K1 would thus indicate the trend (but not the absolute direction) of the magma flow, i.e., the magmatic lineation (e.g., Staudigel et al., 1992; Varga et al., 1998). The commonly observed obliquity of K1 axes relative to the walls of a dike was termed “imbrication fabric” by Knight and Walker (1988). This imbrication has been used by some authors to determine the absolute direction of flow (e.g., Blanchard et al.,1979; Knight and Walker, 1988; Staudigel et al., 1992; Baer, 1995). This fabric would result from the downstream and oblique distribution of phenocrysts due to the strong velocity gradients existing in the magma near the walls of a dike (Fig. 6A).
Flow Vectors Determined Solely from K3 Axes
It is difficult to use the AMS technique in basaltic rocks because independent measurements of the magma flow direction from thin sections in the (K1, K2) planes demonstrate that the K2 axis may also be subparallel to the alignment of phenocrysts in the rock mass (Ellwood, 1978; Moreira et al., 1999; Geoffroy et al., 2002; Callot and Geoffroy, 2004). This observation implies that K1 cannot be used indiscriminately as an indicator of the flow lineation in basaltic intrusions or lavas. In some cases, K1 would represent the intersection axis between shear-type and foliation-type magma planes (Callot and Guichet, 2003). Geoffroy et al. (2002) have proposed avoiding the incorrect use of K1 axes as flow indicators by considering just the angle—when it can be distinguished—between magnetic planes with respect to each wall of the dikes (Fig. 6). In other words, we should use solely the K3 axes (poles of magnetic foliation) and the poles of the dike walls to determine a mean flow vector for the intrusion in order to avoid any misinterpretation of K1 or K2 as flow axes (Fig. 6). For example, Ernst and Baragar (1992) conclude that there was a centrifugal flow of magma within the Proterozoic McKenzie giant dike swarm away from the area of convergence of the dike trends (where they infer vertical flow). Although this is a major result, it is based solely on K1 statistics. Their data should be reworked using the K3 methodology described earlier.
In addition, it may be appropriate to consider not just individual dikes but sets of parallel intrusions of similar thickness and composition, assumed (or demonstrated) to be of the same age (e.g., Callot et al., 2001). In such cases, the working hypothesis is that the set of intrusions were (1) emplaced at the same depth, (2) derived from the same reservoir, and (3) governed by the same dynamics. We can then analyze the statistical grouping of the K3 axes from the whole-core data obtained from the two opposite walls of the dikes (which yield statistical imbrications at the walls), with all data represented in terms of “dike coordinates” (see Rochette et al., 1991). We follow the same reasoning and computation to determine the mean flow orientation within the swarm as applied in the case of a single intrusion.
MAGMA FLOW DURING THE TRAP EMPLACEMENT STAGE: THE CASE OF THE ISLE OF SKYE
Dike Swarms on the Isle of Skye
The Paleocene tholeiitic dike swarms of the British Tertiary igneous province follow a northwest-southeast to NNW-SSE trend and are related to Paleocene to Eocene igneous centers (e.g., Vann, 1978; Speight et al., 1982) (Figs. 3 and 7). Two sub-parallel sets of Paleocene dikes are known on the Isle of Skye (Mattey et al., 1977). Alkaline dikes seem uniformly distributed over the island and are associated with a small finite dilatation (not exceeding 1%). They probably fed the Skye Main Lava Series (Mattey et al., 1977). These dikes are postdated by a prominent tholeiitic swarm, focused on the Skye igneous center, that is associated with significant northeast-southwest-trending magma dilatation (up to 20%) displaying a dual positive gradient (Fig. 7) not only toward the symmetry axis of the swarm but also toward the igneous center (Speight et al., 1982). According to Bell (1976) and Mattey et al. (1977), this major swarm fed the (nowadays, mostly eroded) tholeiitic traps (the “Preshal Mohr lavas”).
AMS studies have already been conducted in the Skye acid ring–dikes (Geoffroy et al., 1997) and the mafic cone sheet (Herrero-Bervera et al., 2001). Both of these studies concluded that magma flow within the annular intrusions of the igneous center was probably subvertical and governed by bottom-to-top pressure gradients from a central crustal magma reservoir. Herrero-Bervera et al. (2001) also investigated the flow pattern within nine intrusions belonging to the regional dike swarm, concluding that there had been some lateral magma flow within the mafic dikes. However, both their methodology and their AMS interpretation were questioned by Aubourg and Geoffroy (2003).
AMS Study of the Tholeiitic Swarm: A Technical Approach
We present here, for the first time, a study carried out in 1995 on magma flow in the Isle of Skye dike swarm (Geoffroy and Aubourg, 1997). We sampled 522 samples in the walls of 30 basaltic dikes (1J to 30J). To avoid turbulent flow, we cored dikes with a thickness not exceeding 1.65 m (average thickness 0.9 m). We preferentially cored dike margins, where cooling had been more rapid, to obtain the largest flow velocity gradients and avoid postinjection rearrangement of the flow fabric. We selected only the basal 2.2 cm of the cores for measurement to minimize any effects due to weathering. The dikes were sampled from six sites (Fig. 7A–F) located at different distances from the igneous center along the general northwest-southeast trend of the swarm. Although all the dikes are tholeiitic, two of them (5J and 17J) nevertheless display high K2O contents (>1 wt%) (Table 1). The tholeiitic dikes clearly belong to the Preshal Mohr type of basalts (Mattey et al., 1977; Kent and Fitton, 2000). The well-defined trends in Figure 8 suggest a single parental melt composition, with magmatic processes dominated by crystal fractionation, possibly within the same reservoir. The magnetic susceptibilities range from 10−4 to 10−1 SI, which clearly indicates the predominance of magnetite in the rock mass. The rock magnetic fabric is dominantly planar, with an average magnetic foliation ratio exceeding the average magnetic lineation in 74% of the dike walls. In most dikes, there is a closer clustering of the K3 axes (Table 2) compared to the much more scattered K1 axes at both walls of the intrusions. Oriented thin sections made from five key samples demonstrate that both K1 and K2 could represent the flow lineation (Geoffroy and Aubourg, 1997), which justifies our choice in using only the magnetic foliation to determine the flow vector orientation.
Because extreme caution is needed in applying AMS to mafic rocks, we used a very critical approach to interpret our results (see details in the notes to Table 2). In our study, twenty-four dikes out of a total of thirty provided results that could be interpreted in terms of orientation of magnetic foliation relative to at least one of the dike walls (Table 2 and Fig. 7). However, there is frequently a large discrepancy between the flow vectors computed from walls on either side of a dike, at the worst yielding nearly opposite directions (Table 2 and Fig. 7). We were able to determine consistent flow vectors (see notes to Table 2) for both of the walls in only eight dikes (1J, 6J, 20J, 21J, 22J, 24J, 26J, and 28J), while considering as (possibly) valid the single-wall data from 5J, 11J, 12J, and 18J (Table 2).
Figure 7 presents all the interpretable data for both walls of each dike at all the studied sites. Figure 9 presents a statistical analysis of K3 and K1 axes grouped by area: all data from the northwest or southeast of Skye are pooled (Groups 1 and 2, respectively) with the exception of data from site Ardvasar (A). In Figure 9, K1 and K3 statistics are expressed in dike coordinates (Rochette et al., 1991; this means than all AMS data are rotated with the dike plane oriented vertically, assuming an arbitrary north-south trend).
Analysis and Interpretation of Results
We summarize the Skye data as follows (Table 2; Figs. 7, 9, and 10). Analysis of the individual data reveals that, apart from dikes 1J and 21J, the flow vectors are all downward-plunging on the Isle of Skye. Southeast of the igneous center (sites A–D), the flow pattern is complex. The flow vector is oriented outward from the igneous center for two dikes (1J and 18J), while it is directed inward in three cases (5J, 6J, and 12J) and downward in two others (11J and 20J). Northwest of the igneous center, the flow vectors are plunging both downward and outward (24J, 26J, and 28J), with the exception of the inward- and downward-plunging 22J (Figs. 7, 9, and 10).
The statistical study of data for the area northwest of the igneous center (Fig. 9, Group 1) shows that (in dike coordinates) the magnetic foliation (K3) yields a very clear imbrication at the “eastern” walls (Φ angle –7° with respect to the “north”), but this becomes less well defined at the western walls (Φ angle +4° with respect to the “north”). This indicates that the dominant flow in this area is lateral and directed toward the northwest (see Fig. 6). A clear imbrication is also encountered southeast of the igneous center (Fig. 9, Group. 2) at the eastern walls (Φ angle +18° with respect to “north” on the diagram), with the statistical foliation parallel to the dike at the western walls (Φ angle 0°). This indicates a general southeastward lateral magma flow, i.e., away from the igneous center. Finally, at the Ardvasar site (Fig. 7) the statistical orientation of K3 at the dike walls is clouded by the coexistence of lateral magma flows directed (geographically) toward both the northwest and the southeast (two clear maxima are observed for K3 at each wall, apart from the east-west trend in Fig. 9).
To summarize, the majority of flow vectors in basaltic dikes from Skye are subhorizontal to downward plunging, which argues strongly in favor of lateral feeding from one or several high-level magma chambers (Fig. 10). The generalized northwestward and southeastward flows in areas northwest and southeast of the Skye igneous center, respectively, can be interpreted in terms of centrifugal lateral feeding of dikes (belonging to the major tholeiitic swarm) from a magma chamber located at the Skye igneous center (Fig. 10). Because both north-westward and southeastward lateral flows are encountered south of Skye, we propose to interpret this result as indicating either (1) a double-feeding source or (2) the existence of postinjection back-flow fabrics in these dikes, as reported elsewhere (Philpotts and Asher, 1994). According to the first hypothesis, and because the Skye regional swarm is connected to the Mull igneous center via a sigmoidal but continuous dilatation axis (Fig. 3B), we tentatively propose that the northwestward flows recorded on southern Skye could have originated from the contemporaneous Mull igneous center. Alternatively, we could also invoke the existence of a small magma chamber south of Skye (see the high level of finite dilatation calculated from dikes south of Skye in Fig. 7).
More specifically, we should note that the steepest plunges of flow vectors are often observed on dikes with the shallowest dips (e.g., dike 11J). In addition, most dikes that yield opposite directions of flow from one wall to the other (i.e., “class 4” dikes; see Table 2) also exhibit the shallowest dips (dikes 16J and 17J, for example). These particular cases could well be explained by a normal shear transposition of the flow-related fabric during solidification of the magma. Also, some of the results from northwest Skye (e.g., 27J and 28J) probably represent the effect of a lateral intrusive flow combined with a lateral Couette-type displacement, respectively sinistral and dextral, in excellent agreement with the NNW-SSE orientation of the maximum principal horizontal stress during the Tertiary in this region (England, 1988; Geoffroy et al., 1993; see Fig. 3). Another interesting point is the clear downward plunge of many flow vectors at some distance from the proposed feeding center (Figs. 7 and 10). The occurrence of a downward flow component in individual dikes has already been suggested from the analysis of flow markers (Baer and Reches, 1987) and petrofabrics (Shelley, 1985; Aubourg et al., 2002). Northwest of the Skye igneous center, ∼50 km away from the magma chamber, our data suggest that the magma flow vector could have been systematically downward in most of the dike swarm at depths of probably less than 3 km below the Paleocene topographic surface. Because magma is injected from chambers at a level of neutral buoyancy (Rhyan, 1987), such a pattern could reflect the increase in magma density due to cooling within dikes farther away from their feeder source. Decreasing both lateral pressure gradients and negative buoyancy of the magma with respect to its host rock would promote convection of the magma within the dike fissure. Another hypothesis would involve levels of neutral buoyancy inclined away from the summit of the Skye polygenic volcano, but this seems to conflict with the strong plunges of flow vectors northwest of Skye.
MAGMA FLOWS AT THE VPM STAGE: THE EAST GREENLAND CASE
Between ∼66°N and 68°N, the southeast Greenland coast partly exposes the western VPM that formed during the Eocene when Greenland and Europe split away to form the Reykjanes basin (Figs. 3A and 5). Three fieldwork campaigns in 1998, 1999, and 2000 were chiefly aimed at establishing the mean magma flows within a mafic dike swarm that crosscuts the transitional crust. This major dike swarm trends northeast to NNE, with a clear dilatation gradient across-strike of the margin (northwest to southeast; Fig. 5). The gradient also increases toward the coastal igneous centers that punctuate the margin (Myers, 1980; Bromann-Klausen and Larsen, 2002; Callot, 2002). The coastal outcrop area represents the flexed transitional crust located beneath an inner SDR wedge, which is nowadays eroded (Geoffroy, 2005). Many dikes were passively tilted during SDR formation. While some of them were injected during the flexing, another set of vertical intrusions postdates the crustal flexing (e.g., Karson and Brooks, 1999; Bromann-Klausen and Larsen, 2002; Lenoir et al., 2003).
We focused especially on the dike swarm centered on the Imlik-Kialineq igneous center (Fig. 5). This swarm is located at the southwest edge of the intrusive complex. A total of 44 dikes were sampled over a distance of 125 km, representing a total of 1172 drilling cores, making this analysis one of the most extensive ever carried out at the scale of a dike swarm. Based on a quantitative comparison between K1 and the observed textural fabric (i.e., paramagnetic phenocrysts) in thin sections from 52 cores, we concluded that neither K1 nor K2 could represent a valid estimate of the flow vector orientation (Geoffroy et al., 2002; Callot and Geoffroy, 2004). We thus drilled specifically chilled margins, considering only the imbricate foliation fabrics reliable for inferring flow vectors.
Dike flow vectors could be interpreted for 24 of the 44 intrusions studied (Fig. 11). In all cases but two, the individual flow vector is directed to the southwest.
Magma flow vectors at the scale of the dike swarm studied are remarkably consistent with a subhorizontal magma flow toward the southwest.
We thus have little doubt that the overlying SDR (volcanic formations) along this VPM were fed laterally (i.e., along strike) from the upper-crustal igneous centers, not vertically as initially thought.
MAGMA FEEDING MODEL FOR THE NORTH ATLANTIC LIP
The Accretion Center Model
Thus, in the north Atlantic volcanic province, we can infer that both traps and SDR are fed laterally by magmas collected in central crustal reservoirs. By itself, this result is not surprising (but needed to be confirmed), because this type of lateral feeding mechanism has long been established at slow-spreading or moderate-spreading oceanic accretion axes (e.g., Staudigel et al., 1992), in Iceland (e.g., Sigurdsson, 1987) or in Hawaii (e.g., Fiske and Jackson, 1972; Knight and Walker, 1988; Tilling and Dvorak, 1993; Parfitt et al., 2002). Such a mechanism seems to be predominant in mafic volcano tectonic systems (Parfitt et al., 2002). The lateral feeding model is also in good agreement with the general observation that lavas forming traps and SDR are differentiated by fractional crystallization in high-level crustal magma chambers (e.g., Cox, 1980; Andreasen et al., 2004). The mechanisms that control dike nucleation in magma chambers have been thoroughly investigated and do not need to be further discussed here (e.g., McLeod and Tait, 1999). Figure 12 presents a horizontal plan illustrating the concept of a single LIP “accretion center” at the depth of the magma chamber. This accretion center model defines the elementary volcano tectonic segmentation in LIP-related volcanic rifts and margins (Geoffroy, 2005; see also Ebinger and Casey, 2001). One may debate the importance of lateral transport of low-viscosity magmas along cracks in controlling the regional distribution of traps and SDR. It is not easy to determine the along-strike length of individual dikes because dikes, like any tabular intrusion, are segmented in 3-D. Nevertheless, magma has been shown to flow laterally as far as 100 km in the Hawaii dikes (e.g., Parfitt et al., 2002). We suggested earlier (Fig. 10) that some of the dikes on the Isle of Skye are fed by the Mull igneous center, corresponding to ∼200 km of lateral flow (Fig. 3B). McDonald et al. (1988) concluded from geochemical evidence that the Cleveland Dike (Fig. 3B) was fed laterally in a single pulse from the Mull igneous center, which would represent up to 430 km of subhorizontal flow. Because low-viscosity lavas may flow over great distances from their eruptive fissures (Self et al., 1997; see also Fig. 12 for the role of faults), the observations made earlier imply that the areal extent of LIP lavas is controlled by crustal processes rather than the mantle.
Could the Whole of the North Atlantic LIP Magmas Be Drained through Individual Magma Centers?
Igneous centers are thus the key to understanding the distribution of melting in the mantle underlying LIPs. Although the approach is highly speculative, it is possible to estimate the volume of magma that has transited through individual north Atlantic igneous centers. We estimate that a minimum of thirty-eight igneous centers were active during the trap stage in the north Atlantic (Fig. 4). The number of igneous centers active during the Eocene breakup stage is presently unknown (see Chapter 1). Basing our estimate solely on the trap stage, we obtain a volume of 106 km3, which seems a very large upper-limit value for the magma extruded and intruded during the period from 62 to 58 Ma (Paleocene–Earliest Eocene) (see White and McKenzie, 1989; Eldholm and Grue, 1994). The average upper-limit output rate at an individual igneous center would thus be ∼7 × 10−3 km3/yr. This value could be compared with the volume of magma represented by a dike intrusion phase to estimate an average eruption rate. However, as is explained further later, such an exercise may be meaningless.
One of the best-documented cases is from Hawaii, which provides an output of 3.3 × 10−3 km3 (Cervelli et al., 2002). A similar value (7.5 × 10−3 km3) is obtained from GPS measurement in the Galapagos, following the 1995 Fernandina flank eruption (Jónsson et al., 1999). The former estimates relate to both thin (<1 m) and nonfeeder intrusions. Although in a different geodynamic context, the lateral dike intrusion event monitored in the Izu Islands during the year 2000 corresponds to an estimated magma volume of ∼1 km3 (Nishimura et al., 2001). The volume of the Cleveland Dike (thickness ∼20 m) appears to attain 85 km3 according to McDonald et al. (1988). Self et al. (1997) report a volume of 1300 km3 for a single lava flow in the Columbia River LIP, thus implicitly setting a lower-limit value for the associated feeder dike. With such a range of values (over six orders of magnitude), it is not possible to evaluate the average volume of magma coming from a single igneous center. However, to obtain a gross estimate of the dike intrusion frequency and related magma volumes, we can tentatively refer to the two most intensively investigated mafic central volcanoes, i.e., Kilauea in Hawaii (e.g., Tilling and Dvorak, 1993) and Krafla in Iceland (e.g., Sigurdsson, 1987; Hofton and Foulger, 1996), which are situated in intraplate and plate boundary settings, respectively. In both cases, the magma supply at igneous centers seems highly dependent on the progressive buildup of stress within the surrounding crust, irrespective of whether these stresses are due to gravity (Hawaii) and/or plate tectonics (Iceland).
The most recent intrusive activity on Kilauea (since 1956) appears to fit with at least one dike intrusion every 4 yr, with periods of much higher activity (see Tilling and Dvorak, 1993). In the case of Krafla, it seems that periods of quiescence lasting 100–150 yr (periods of tectonic stress concentration) alternate with episodic faulting or diking events (the last one spanning 6 yr from 1975), during which about twenty dikes were injected laterally (Sigurdsson, 1987) parallel to the trend of the maximum horizontal stress. We should note that, in both Hawaii and Iceland, the volume of magma intrusion in dikes during an intrusive or eruptive event largely exceeds the volume represented by magma chamber deflation. In this way, diking events reflect the continuous feeding of the upper-crustal reservoirs by the mantle. The total flow out the Krafla reservoir during the last period of activity was ∼1.08 km3, which corresponds to an average of 8 × 10−3 km3/yr over a period of 125 yr. The mean output rate of Hawaii gives a strikingly similar value when averaged since 1840 (Tilling and Dvorak, 1993). These values compare well with the estimated maximum output rate of ∼7 × 10−3 km3/yr for an individual north Atlantic volcanic province igneous center at the trap stage, suggesting that these igneous centers could be good candidates for the feeding of the whole LIP.
IMPLICATIONS FOR MANTLE MELTING
In this section we discuss the implications of our results on mantle models. We use the term lithosphere to refer to the mechanical entity, including the crust and part of the upper mantle, that is able to sustain stress over geological periods (e.g., Anderson, 1995). This lithosphere is thermally conductive. It is separated from the large-scale convecting mantle by a thermal boundary layer (TBL) in which temperature tends asymptotically to a convective-type gradient. This boundary layer is considered either stable or convective on a small scale (e.g., Jaupart and Mareschal, 1999; Morency et al., 2002).
Although other melting materials could be involved (see the first section of this article), the adiabatic decompressive melting of rising mantle is generally acknowledged as the primary source of magmas at LIPs and oceanic ridges. In such cases, the area of mantle melting is evidently primarily controlled by the effective mantle temperature, pressure, and volatile contents.
Our data suggest that no generalized vertical magma transfer occurs in LIPs (apart from beneath the crustal igneous centers themselves). This leads to major geodynamic consequences. The question we then need to address is the meaning of igneous centers in relation to the pattern of mantle melting. With such a localized distribution of feeders for magma crust accretion in LIPs, how could the mantle be homogeneously melting (as in the initial plume head model)?
We should bear in mind that, at the trap stage (see the section on igneous center distribution), the distribution of igneous centers seems partly independent of the premagmatic rift zones in the north Atlantic (Fig. 4). However, major discontinuities (e.g., Late Caledonian strike-slip faults, but also Mesozoic normal faults reactivating Caledonian thrusts) clearly have some influence on the igneous centers' distribution. During the VPM stage, this cannot be the case, because the igneous centers are regularly distributed within the necked and segmented crust (the area associated with SDR development; see Geoffroy, 2005; Fig. 5). In this latter context, their spacing seems related to the amount of lithosphere necking associated with the breakup. At both stages of LIP evolution, the nonrandom distribution of igneous centers strongly suggests the existence of some kind of small-scale fluid instability within the lithosphere. The related “fluidlike” material could be present as melts (hypothesis 1: low-viscosity magma diapirs) or, as discussed further later, in the solid state (hypothesis 2: small-scale mantle diapirs).
Are Magma Diapirs Possible?
Hypothesis 1 could be compatible with the following scenario: mafic to ultramafic magma rises homogeneously through the mantle lithosphere, collecting as a continuous sill-like layer(s) at the Moho, where it partially differentiates, then ascends as diapirs to form igneous centers (Fig. 13A). It is likely that magma collects at levels of neutral buoyancy, thus explaining the presence of the HVZ at Moho depth under LIPs (Figs. 1 and 2; e.g., Fyfe, 1992; Holbrook et al., 2001). However, it is extremely improbable that Rayleigh-Taylor instabilities could develop in a basaltic layer (sill-like?), and indeed this should be ruled out. First, this would imply that the bulk density of the magma decreases more rapidly through fractional crystallization than the density increases due to cooling. Second, the fluid behavior of a magma, whether of Newtonian or power-law type, depends on several factors, including its temperature and crystal content (e.g., Weinberg and Podlachikov, 1995). A mafic magma extracted from a reservoir is more likely to behave as a Newtonian fluid, with a viscosity not exceeding 102 Pa/s (e.g., Spera, 1980). On the other hand, the lower-crust viscosity, even in high heatflow areas, is not expected to be lower than 1018 Pa/s. This constitutes a very strong obstacle for the development of Rayleigh-Taylor instabilities for a low-viscosity fluid. Nevertheless, some authors accept that a high-temperature diapir may decrease the host-rock viscosity at its edges (Spera, 1980; Rubin, 1993). This phenomenon could be enhanced by partial fusion of the country rock. In addition, if the magma diapir behaves as a power-law fluid enclosed in a power-law “ductile” crust, the buoyant stress of the diapir may also decrease the wallrock viscosity (Weinberg and Podlachikov, 1995). However, whatever the true fluid behavior of the magma, the viscosity of the Q-rich lower crust would probably not fall beneath 1016 Pa/s (see, for example, Weinberg and Podlachidov, 1995). Such a high viscosity ratio between the magma and the host rock suggests that this latter behaves elastically and would fracture (Rubin, 1993). This may be related to the accepted geological observation that the emplacement of mafic plutons is always fracture-associated and never involves processes of diapiric intrusion (e.g., Shaw, 1980). Moreover, we note that even the existence of acid diapirs can be questioned (e.g., Clemens and Mawer, 1992). Finally, we should add that buoyant rising magma diapirs are expected to slow down, cool, and finally solidify beneath the brittle-ductile transition in the crust. Such a scenario would completely contradict the postulated existence of large mafic magma chambers in the LIP upper crust.
We thus believe that the only plausible explanation for the distribution and role of LIP igneous centers as magma feeders is that melting is focused within the TBL mantle itself (hypothesis 2). This may occur in two cases: (A) if the TBL is a nonconvective steady-state hot layer, but melt products or melting are localized in specific areas, or (B) if the TBL exhibits small-scale 3-D convection, and melting occurs specifically at the top of the uprising cells.
Steady-State TBL Hypothesis
We first discuss the concept of a steady-state buoyant TBL with inhomogeneous melting.
Many authors have discussed the general issue of melt extraction and migration in an adiabatically flowing melting mantle, especially at spreading ridges (e.g., Spiegelman and Reynolds, 1999). Both magma percolation through a solid compacting matrix and magma hydrofracturing have been investigated, with the latter mechanism favored (e.g., Shaw, 1980; Spera, 1980; Nicolas, 1990). Most authors acknowledge that, overall, melt buoyancy contributes in a major way to magma flow in the mantle compared with compaction and dynamic pressures (e.g., Schmeling, 2000). Therefore, we do not expect large amounts of lateral magma flow within the mantle. However, provided the magma is extracted, a highly viscous dehydrated layer could be developed at the top of the melting zone that may act as a barrier channeling the magma flow from below (e.g., Morgan, 1987; Choblet and Parmentier, 2001). This led Madge et al. (1997) to explain along-axis variations in the thickness of igneous crust at slowly spreading ridges by the lateral upslope migration of melts. Such a process is assumed to occur at the top of large-scale undulations in a passive buoyant mantle whose topology is inherited from the 3-D lithosphere structure. In particular, areas with excess conductive cooling would act as melt deflectors. While this model contradicts the diapiric model of oceanic accretion proposed by Lin et al. (1990), it might be compatible with the existence of a regional and continuous low-resistivity layer in northeast Iceland, as shown from both magnetotelluric and electrical measurements, with a minimum depth beneath the Krafla and Grimsvotn igneous centers (e.g., Bjornssön, 1985). This layer was interpreted as corresponding to a partially molten mantle (5–20% partial melt) at the roof of uprising asthenosphere domes. However, the model of Madge et al. (1997) can hardly be applied to the north Atlantic continental lithosphere case, at least at the trap stage. Indeed, there appears to be no clear correlation at this stage between the thickness of basaltic products and the initial (Cretaceous, i.e., premagmatic) thermal state of the very heterogeneous lithosphere.
Melting instabilities could also develop laterally within a thermally destabilized subhorizontal asthenosphere that is close to melting or partially molten. This scenario has been discussed by Tackley and Stevenson (1993) and Schmeling (2000), who investigated the lateral development of such instabilities using Newtonian and non-Newtonian viscosities, respectively. Although such instability propagation could explain the alignment of volcanoes with a progressive variation of ages, it cannot account for the observed 2-D distribution of igneous centers of similar ages within the north Atlantic LIP (Fig. 4).
Convective Destabilization of the TBL
We now explore the most plausible mechanism, i.e., partial melting of the TBL at the top of buoyant small-scale convection cells (Fig. 13). The spacing of igneous centers beneath the thinned north Atlantic volcanic province lithosphere at the VPM is shorter than for igneous centers punctuating the thicker trap lithosphere, thus suggesting a relationship between igneous centers and mantle dynamics. The TBL is increasingly considered to be undergoing natural small-scale convection, even in the absence of any additional heat supply (i.e., without invoking a mantle plume). This has been highlighted using different mantle rheologies and boundary conditions in a large number of experimental (e.g., Davaille and Jaupart, 1993, 1994) and numerical studies (e.g., Dumoulin et al., 1999; Callot, 2002; Morency et al., 2002).
During the trap stage, we should point out that the distribution of igneous centers is more or less homogeneous in 2-D map view (Fig. 4), and not specifically associated with, for example, thinned Mesozoic crust. Therefore, at this stage the upwelling of the melting mantle does not appear to be primarily dependent on previous tectonic stretching and thinning of the lithosphere. Such a conclusion has also been drawn from the time evolution of igneous geochemistry in the British Tertiary igneous province by Thompson and Morrison (1988) and Kerr (1994). These authors proposed a progressive and localized penetration of melting mantle into the continental lithosphere beneath the Skye and Mull igneous centers to explain the chemistry of the successive magma series.
To test the hypothesis that igneous center spacing is correlated with small-scale 3-D convection in the TBL, we would need to compare the characteristic wavelength λ of this small-scale convection with the average spacing of igneous centers in the north Atlantic (Callot, 2002). Theoretically, λ should be close to the thickness of the convective layer itself, because the convective cell aspect ratio generally lies between 1.00 and 1.35 (e.g., Houseman et al., 1981). Therefore, to resolve this issue, we need to evaluate the thickness of the convective TBL beneath a 60 Ma, relatively heterogeneous lithosphere. Although this thickness cannot be accessed directly, e.g., from geophysical data, it could be determined indirectly from experimental data. For example, Davaille and Jaupart (1993, 1994) propose an equation in which λ is inversely proportional to the surface heat flux Qs (at the time of the convection). From the data of Morency et al. (2002), we can also derive different relations (depending on the type of mantle viscosity) between λ and Qm, the mantle heatflow beneath the conductive lithosphere.
Theoretically, we could also estimate (1) the crustal heat production (and its time evolution; see Artemieva and Mooney, 2002) and (2) the cooling and recovery of the continental lithosphere. These estimates could be used to correct present-day surface heatflow (or lithosphere thickness) in the north Atlantic volcanic province and correlate it with the past wavelengths of small-scale convection cells (Callot, 2002). The final step in testing the hypothesis would be to compare the theoretical wavelength with the actual spacing of igneous centers. While this point is not fully investigated here, we nevertheless give some first-order idea of the validity of small-scale convection at LIPs. It is clear that the previous reasoning should be primarily applied to comparing trap areas that undergo little lithosphere thinning before the onset of trap formation. It is indeed difficult to estimate the true premagmatic lithosphere thickness if strong extension occurred (this depends notably on the rate of lateral cooling during extension). This should rule out any direct application to the north Atlantic volcanic province because of its relatively complex Mesozoic evolution (e.g., Van Wijk and Cloetingh, 2002; Scheck-Wenderoth et al., 2006). However, according to the previous argument, and as a first approximation, the thicker the present-day lithosphere (or the lower the surface heat flux), the longer the average spacing between igneous centers. Callot (2002) tested this hypothesis on a range of trap areas worldwide (Deccan, Siberia, Parana-Etendeka, etc.) by making use of available geological, seismological, and heatflow data, but without correcting for the cooling and evolution of the lithosphere since the associated trap emplacement (however, in many cases we may consider the error in thickness as falling within the uncertainties of estimation of present-day lithosphere thickness).
Taking account of the evident serious limitations previously outlined, Figure 14 suggests a positive correlation between lithosphere thickness (and indirectly TBL small-scale convection) and the spacing of igneous centers, even when considering the off-shore Hebrides data (heterogeneous in thickness lithosphere). Note that the main discrepancy with the general correlation shown in Figure 14 concerns the British Tertiary igneous province igneous centers, where the close spacing is fault controlled. It is noteworthy that many igneous centers lie along inherited lithospheric-scale or crustal-scale discontinuities (especially Late Caledonian subvertical shear zones) (Fig. 4). At the same time, we should also take into account the existence of a transient Paleocene lithospheric stress field with the maximum horizontal stress converging toward a single point (Fig. 3). This could suggest a relationship during trap formation between a sudden change in regional-scale stress field and the enhancement of TBL destabilization, especially along the Caledonian discontinuities that were slightly reactivated during the Paleocene.
The distribution of igneous centers during the breakup stage seems much more focalized along the thinned and stretched Eocene VPM than at the trap stage (Fig. 5). We note a strong decrease in the spacing of igneous centers (or igneous centers postulated from off-shore gravity and magnetism) from the onshore internal margin (thicker crust and lithosphere) and the offshore or distal margin (thinnest crust or lithosphere with outer SDR). The data presented here strongly support a mechanism of VPM accretion similar to that at slowly spreading ridges, which partly explains the strong analogy in structure between the two crusts (Fig. 2). We point out that the wavelength of magma segmentation along the east Greenland VPM (located off-shore) is very similar to that observed along the adjacent Reykjanes Ridge (Gac and Geoffroy, 2005). Similar observations have been made elsewhere (Behn and Lin, 2000). Some authors have proposed that magma may be focused at the top of small-scale convective cells, not only at slowly spreading ridges (Lin et al., 1990), but also at volcanic passive margins (e.g., Mutter et al., 1988; Keen and Boutilier, 1999). Kelemen and Holbrook (1995) and Holbrook et al. (2001) present arguments in favor of a strongly active upwelling mantle at VPMs, with active rates up to four times faster than passive rates (stretching-related). According to Huismans et al. (2001) as well as Van Wijk et al. (2001), a significant component of active mantle upwelling (and consecutive melting) may naturally occur beneath rifts at the end of passive stretching. However, Nielsen et al. (2002) argue for a slight temperature excess in the mantle. In any case, considering the true 3-D architecture of a VPM, the active mantle upwelling should be regarded as small-scale 3-D (channeled along the breakup zone) and certainly not 2-D axisymmetric (see Geoffroy, 2005).
We show in this article that magma distribution at the LIP ground surface has little to do with the extent of mantle melting at depth. At both stages of LIP evolution, the magma is channeled through pinpoint crustal pathways that extend downward to localized melting zones in the mantle. We thus propose a magma feeding model for LIPs that is quite distinct from previous views of homogeneous mantle melting over plume heads (Fig. 15A) or homogeneous melting over deep “mantle ridges” (Fig. 15B). At both LIP stages, the best model describing the pattern described during traps emplacement invokes the existence of small-scale convection within the mantle (Fig. 15C). Small-scale convection in the TBL is not a specific LIP-related phenomena and may correspond to a generalized process beneath the mechanically rigid lithosphere (e.g., Korenaga and Jordan, 2002; Morency et al., 2002). It is probably a natural consequence of the negative buoyancy of the bottom of the lithosphere. However, such small-scale convection would have to be enhanced to explain the quite sudden mantle melting during the trap stage in LIPs (and not elsewhere or at any other time). A transient excess in TBL temperature (e.g., plume head emplacement) could cause enhanced convection. However, we suggest that other controls, such as a transient Paleocene compressive stress field (see Doubre and Geoffroy, 2003) acting in a lithosphere of highly variable thickness, could indirectly trigger mantle melting. This could be tentatively explained by diapiric destabilization at the top of the existing buoyant small-scale convecting cells, especially along reactivated lithospheric sutures. This latter explanation probably fits best with the available observations as well as the time and space constraints.
There is a strong structural analogy between VPMs and oceanic crust, which is based on their layering (Fig. 2), along-axis segmentation (Lin et al., 1990; Behn and Lin, 2000; Callot et al., 2002; Gac and Geoffroy, 2005), and crustal growth mechanisms (this study; for spreading ridges see Staudigel et al., 1992; Madge et al., 1997). This is probably due to similar mantle and crustal growth processes (Geoffroy, 2005). To improve our understanding of VPMs, we need to ask why some continental rifts function as oceanic rifts, although with enhanced magmatism (Fig. 2), while other rift systems do not. Both types of rift system (i.e., amagmatic and magmatic) are formed under extensional stress regimes and may develop at the margins of cratonic areas; the east Greenland VPM is an example of a magmatic system, whereas the Baikal rift is an amagmatic system (e.g., Pavlenkova et al., 2002). This suggests that lithosphere edge effects (e.g., Anderson, 1994; Sheth, 1999) are not the sole indirect cause of magmatism at VPMs. We argued elsewhere that the pattern of lithospheric deformation at VPMs during breakup is quite dependent on the weakening of the upper-lithosphere mantle by low-viscosity anomalies located within the mantle lithosphere (the soft-point model of continental breakup; see, for example, Callot et al., 2002, and Geoffroy, 2005). These low-viscosity anomalies explain both the 3-D strain localization and the rift propagation at VPMs (Callot et al., 2002; Geoffroy, 2005). They fit very well with the postulated zones of mantle melting at depth presented here, thus giving a consistent model of combined magma, rheologic, and tectonic evolution at VPMs (Geoffroy, 2005). Here again, the 3-D sublithospheric convection pattern provides a key to understanding the origin and evolution of VPMs. Probably all rift systems are associated with small-scale convection in the TBL (see Korenaga and Jordan, 2002). The differences in along-strike segmentation between nonvolcanic (e.g., Rhine Graben) and volcanic rifts (e.g., the Ethiopian rift) suggest that the wavelength of small-scale convection is different in the two cases, thus also implying differences in mantle viscosity (see the discussion of the 3-D pattern of oceanic accretion in Choblet and Parmentier, 2001). Our study does not provide a solution for this particular issue. However, we suggest that during the latest stages of continental breakup in a LIP the small-scale convection suddenly turns from a poorly organized sublithospheric pattern (large wavelengths) to a more regular spreading-type regime (smaller wavelengths; see Fig. 14).
This article is contribution no. 2173 of the GDR Marges. The research was co-supported by Euromargin (LEC-01) and Institut Paul-Emile Victor (IPEV) (Program 290). Don Anderson, Gillian Foulger, Donna Jurdy, and Jolante Van Wijk are warmly thanked for their constructive and interesting comments. M. Carpenter is thanked for his efficient correction of the English language.
22 January 2007, Jolante W. van Wijk and Laurent Gernigon
The data presented in the chapter by Geoffroy et al. (this volume) suggest that a significant part of the magmas forming the central North Atlantic Igneous Province (NAIP) were injected in a limited number of igneous centers, feeding dikes (sub-horizontally) over (sometimes) large distances. More work is needed, but if the suggested pattern of the calculated dike flow vectors as presented in this study is correct, existing models for emplacement processes may need to be reevaluated.
Prior work on emplacement processes often assumed that magma is transferred vertically and that, therefore, the surface extent of flood basalts gives a good indication of the extent of mantle melting at depth. Geoffroy and co-workers, however, find that magma feeds a limited number of igneous centers and then travels from these centers outward, subhorizontally. They propose, for example, that seaward-dipping magmatic reflector sequences could be fed laterally from central crustal reservoirs instead of vertically as initially thought. This conclusion confirms an earlier discussion advanced by Klausen and Larsen (2002) on the east Greenland coast parallel dike swarm. Geoffroy et al. (this volume) provide new quantitative results and show that this lateral feeding model proves to be valid for other areas in the north Atlantic as well.
The authors propose that no direct relationship needs to exist between the magma distribution at the surface of the Earth and the extent of mantle melting at depth. The distribution of igneous centers in the crust is more indicative of the extent of mantle melting. Several lines of arguments exist on how melt accumulates in igneous centers, and the authors favor a process in which melt migration horizontally in the mantle over large distances is unlikely, such that the igneous centers thus form above or close to the location of mantle melting. To explain the distribution of igneous centers, small-scale convection-induced magmatism just below the lithosphere is proposed. Geoffroy et al. (this volume) do not support diapiric structures inside the continental lithosphere to explain the igneous feeding systems. Nevertheless, other studies have supported such a mechanism for different study areas and under specific conditions; Gerya et al. (2004), for example, describe and model diapiric features in the Bushveld Complex explained by the temporal inversion of a vertical temperature gradient as a result of the emplacement of large quantities of hot and mafic magma onto cold material. Drury et al. (2001) simulate diapiric intrusions inside a cold cratonic area.
If the melting region in the mantle is indeed reflected by the distribution of igneous centers, a large, quite uniform thermal plume head is not supported by these observations, according to the authors.
The suggested link between these sublithospheric instabilities (small-scale convection) and the igneous centers is clearly at a developing stage at this moment, and no geophysical evidence exists yet that allows us to prove the presence of these mantle instabilities during Paleocene time. Emplacement and formation of huge igneous complexes at depth is far from being well understood. Recent potential-field investigations in the Rockall–UK area (Edwards, 2002) for example, suggest that massive high-density igneous complexes vary in shape and can be present deep in the crust or restricted to only the upper part of the crust. We cannot exclude that different modes of emplacement may exist (Edwards, 2002). More geophysical data are needed to detect and investigate igneous intrusions in other parts of the NAIP as well.
The authors discuss about three dozen igneous centers located in the central NAIP. Most of them are Paleocene in age, but some remain undated, and some datings are not well constrained (e.g., Hitchen, 2004; Meyer et al., this volume). To our knowledge, these magmatic features far from the breakup axis have not been recognized in the rest of the NAIP area to the same extent (Fig. D-1). This could change in the future when more datasets are processed, but for now it is unknown whether the Rockall–UK area style of massive igneous accumulation is representative of the entire NAIP. On the Norwegian margin, for example, Berndt et al. (2000) identified the Hel Graben Sill Complex, but this feeder system does not seem similar to the feeders investigated by Geoffroy et al. (this volume).
Along-margin variations in melt accumulation could be expected, as the tectonic evolution (and stress field) of different margin segments has varied during the long rifting history that ultimately resulted in breakup. Lithosphere-scale processes (such as the regional stress field, as mentioned by Geoffroy et al., this volume) affected by inherited structures play a role in how magma transfer and accumulation will develop. Deep shear zones such as the Great Glen fault can probably affect the upper mantle as suggested by deep seismic data cutting the Moho (e.g., Klemperer and Hobbs, 1991) or/and act as conduits for melt transport. Figure D-1 illustrates the distribution of major (recognized) fracture and shear zones in the north Atlantic, suggesting a geographical and structural relation with breakup-related intrusions.
At this point, many open questions remain, including these:
How has melt accumulated in the NAIP? Are there regional variations?
Does the suggested subhorizontal melt transfer mechanism proposed by Geoffroy et al. (this volume) apply to the entire NAIP?
If there indeed exists a dense concentration of igneous feeding centers in the Rockall–UK area, how can this be explained? Why are similar features not clearly observed north of the Faeroes-Shetland basin?
In general, how do melt migration and accumulation work in continental lithosphere, and what is the relationship/connection with mantle melting?
It is clear that more research is needed to establish the extent, style, and distribution of igneous feeding centers along the margins of the NAIP. A more complete picture would enable testing the model proposed by Geoffroy et al. (this volume) for magma transfer in large igneous provinces.
Figures & Tables
Plates, Plumes and Planetary Processes
- Atlantic Ocean
- continental crust
- dike swarms
- Great Britain
- Highland region Scotland
- Iceland Plume
- igneous rocks
- Inner Hebrides
- Inverness-shire Scotland
- Isle of Skye
- large igneous provinces
- mafic magmas
- North Atlantic
- oceanic crust
- preferred orientation
- stress fields
- structural analysis
- transition zones
- trap rocks
- United Kingdom
- volcanic rocks
- Western Europe
- North Atlantic igneous province