Comparison of lithosphere structure across the Orphan Basin–Flemish Cap and Irish Atlantic conjugate continental margins from constrained 3D gravity inversions

Regionally constrained 3D gravity inversion results on the Orphan Basin–Flemish Cap and the Irish Atlantic conjugate continental margins are compared to investigate crustal structure, early rifting history and geological evolution of this part of the North Atlantic. The full-crustal density anomaly distributions provide some of the first depth images of how rifted structures compare along and across these conjugate margins. Broad similarities in crustal structure are identified with some noticeable differences, linked to rifting and crustal stretching processes. Extreme crustal thinning (stretching factors >3.5) is indicated beneath much of the southern Porcupine Basin, the western half of West Orphan Basin, the eastern half of Jeanne d’Arc Basin, the southeastern half of East Orphan Basin and in pockets beneath Rockall Basin. This appears to have resulted in the serpentinization (and possible exhumation) of mantle lithosphere on the Irish Atlantic and Flemish Cap margins but not beneath Orphan Basin. A simple evolution model is proposed for the early stages of rifting between the margins. It is suggested that ancient orogenic sutures played an important role in controlling the northward migration of rifting and the rotation and displacement of Flemish Cap out of Orphan Basin. Supplementary material: Enlarged maps from this paper are available at www.geolsoc.org.uk/SUP18527.

Non-volcanic or magma-poor margins represent an ideal setting for studying rifting dynamics as they lack the complexity introduced by magmatic bodies, which can alter or obscure the extensional crustal structures of interest. This study focuses on comparisons across the non-volcanic or magma-poor Orphan Basin-Flemish Cap and Irish Atlantic conjugate continental margins (Fig. 1), which have been largely overlooked compared with the Newfoundland-Iberia and Flemish Cap-Galicia Bank conjugate continental margins to the south. This oversight is due in large part to the limited number of deep seismic profiles available with which to reconnect the ancient conjugate margins and investigate variations in their along-strike structures.
A recent regionally constrained 3D gravity inversion study over the Irish Atlantic continental margin (Rockall, Porcupine and Goban Spur) showed excellent agreement between available seismic constraints of crustal structure and major structures resolved in the inverted 3D density anomaly model for the margin (Welford et al. 2010a). The resulting model allowed for crustal characteristics such as Moho structure and crustal thickness to be tracked and extrapolated across the margin into regions lacking deep seismic constraints. In the present study, equivalently constrained 3D gravity inversion results have been computed for the Orphan Basin-Flemish Cap region using identical parameters to those used for the Irish study, allowing structures across the two margins to be compared.
Although largely extensional in origin, the tectonic evolution of the Orphan Basin-Flemish Cap and Irish Atlantic conjugate continental margins is also postulated to have involved contemporaneous rotation and displacement of Flemish Cap out of the Orphan Basin (Enachescu et al. 2005;Sibuet et al. 2007b). Owing to this added dimension of complexity, asymmetry in the conjugate pair and regional variations in crustal thinning are expected.
The purpose of this study is to build upon previous tectonostratigraphical and crustal work on both margins to reconcile their respective rifting histories and to test hypothesized connections between linked basins and underlying pre-rift basement terranes. The results may also lead to an improved understanding of the petroleum habitat of the region and aid in connecting exploration plays across the North Atlantic to their conjugate equivalents.

Tectonic setting
The closing of the Iapetus Ocean during the Caledonian-Appalachian orogeny in Palaeozoic time stitched together the distinct basement terranes that make up the offshore Newfoundland and Irish margins considered in this study (Haworth & Keen 1979;Williams 1984Williams , 1995Chew 2009). The continental crust of the Grand Banks and Flemish Cap, offshore Newfoundland, consists of basement rocks from the Gondwanan Avalon terrane (Lilly 1965;King et al. 1985). On the conjugate Irish margin continental terranes of Proterozoic, Avalonian and Variscan age have been identified based on geological extrapolation from the Irish mainland (Chew 2009;Chew & Stillman 2009), geological provenance studies (Tyrrell et al. 2007(Tyrrell et al. , 2010 and interpretation of seismic reflection data . Structures and tectonic fabrics within these basement terranes generally follow the NE-SW strike of the Caledonian-Appalachian Orogen, and/or the east-west strike of the Variscan orogen, particularly in the southeastern part of the study area. Despite the mosaic of underlying basement terranes, there appears to be no clear regional correlation between crustal thickness and the nature or boundaries of the basement terranes on the Irish margin. A 30 km thick crust is observed beneath Ireland across a range of basement types and structures (Lowe & Jacob 1989;Landes et al. 2005;Ziegler & Dèzes 2006). The crust is similarly 30 km thick beneath the Porcupine Bank (O'Reilly et al. 2006) and the Rockall Bank (Vogt et al. 1998). There is no evidence for significant crustal extension in these areas and 30 km is therefore regarded as the unstretched crustal thickness in the Irish margin. In contrast, the Newfoundland margin exhibits greater variability of 'unstretched' crustal thickness (30 km beneath Flemish Cap (Funck et al. 2003); 36 km beneath the Grand Banks (Lau et al. 2006a); 40 km for the onshore Avalon zone, 45 km for the Proterozoic Grenville province of western Newfoundland and 30-37 km for the onshore crust between those two regions (Hall et al. 1998)). Although this variability does generally correlate with basement terranes and their inferred boundaries onshore Newfoundland, the contrast in crustal thickness between the Grand Banks and Flemish Cap occurs within the same inferred basement terrane, suggesting some rheological or deformational variability within terranes.
Separation of Iberia and Eurasia from North America to form the North Atlantic Ocean was preceded by an initial rifting episode during the Triassic, which created many of the half-graben basins found on the Grand Banks (e.g. Carson-Bonnition, Jeanne d'Arc, Whale, Horseshoe; Tankard & Welsink 1988), in the East Orphan Basin (Enachescu et al. 2004c;Enachescu 2006), on the Irish margin (e.g. Celtic Sea, Porcupine, Western Approaches; Naylor & Shannon 2009) and on the Galicia Bank margin (e.g. Inner Galicia, Lusitanian; Murillas et al. 1990). There is also seismic evidence that rifting occurred in the Rockall Basin as early as the Permo-Triassic (Shannon et al. 1999;Štolfová & Shannon 2009;Tyrrell et al. 2010;Naylor & Shannon 2011). However, the Flemish Cap was not affected until the Early Jurassic (Sibuet et al. 2007b).
The formation of these early Mesozoic basins was controlled largely by pre-existing basement structures and tectonic fabrics (Shannon 1991), with most basins on the Newfoundland and Irish offshore margins generally parallel to the NE-SW Caledonian-Appalachian structures. On the southern Irish margin, basins such as the North Celtic Sea Basin and those in the eastern part of the Goban Spur province display an ENE-WSW alignment where Caledonian and Variscan fabrics are subparallel (Naylor et al. 2002;. A second phase of rifting from the Late Jurassic to the Early Cretaceous, which progressed diachronously from south to north, led to the east-west separation of the Grand Banks from central Iberia and then to the SE-NW separation of the southeastern margin of Flemish Cap from Galicia Bank (Williams 1984;Tucholke et al. 1989;Grant & McAlpine 1990). The latter separation was preceded by, and continued contemporaneously with, the opening and subdivision of Orphan Basin to the NW of Flemish Cap into an interpreted older East Orphan Basin and a younger West Orphan Basin (Enachescu et al. 2004c). The anticlockwise progression in orientation from NE-SW to north-south of the faults and basement ridges within the Orphan sub-basins reflects the westward propagation of rifting (Enachescu et al. 2004a,c;Enachescu 2006 (Enachescu et al. 2004a,b,c;Sibuet et al. 2007b) but this conclusion may simply reflect a lack of deep well control in the basin.
Geodynamic changes at the Jurassic-Cretaceous boundary resulted in major basinal changes. A phase of Cretaceous (Aptian-Albian) rifting was part of a major plate reorganization  Hall et al. (1998) and Tyrrell et al. (2007). Locations of magnetic anomaly 34 (from Srivastava et al. 1988Srivastava et al. , 1990 (Naylor & Shannon 2011) and differed not only in orientation but also in style from the two previous rift phases. The north-south Jurassic rift system was overprinted and sometimes 'beheaded' by a set of large NE-SW-oriented Early Cretaceous rifts that accumulated very thick, typically marine strata, such as in the Porcupine and Rockall basins.
NE-SW extension, starting in the Late Cretaceous Tucholke et al. 1989;Hopper et al. 2006;Tucholke & Sibuet 2007), separated the NE margins of Flemish Cap and Orphan Basin from the Goban Spur and the Porcupine High, offshore Ireland. This was interpreted by Doré et al. (1999) as an oblique reopening of the Caledonian-Appalachian suture and fold system. As a result of the reorientation of extension directions, the East Orphan Basin was again reactivated and widened. On the Irish margin the Early Cenozoic coincided with the onset of the North Atlantic Igneous Province. Widespread igneous activity took place across a broad region, with the formation of lavas, sills and dykes and possible magmatic underplating at the continent-ocean boundary. Uplift and episodic influx from continental margins into the adjacent sedimentary basins occurred. Phases of epeirogenic movemements, in early, mid-and late Cenozoic times, have been documented along the Irish margin (Praeg et al. 2005), together with at least one major phase of compressive tectonism as sea-floor spreading progressed.
The postulated clockwise rotation of Flemish Cap (43° relative to Iberia) during the Late Triassic to Early Cretaceous, and its 200-300 km southeastward displacement (relative to North America) from Late Jurassic to early Aptian time (Enachescu 2006;Sibuet et al. 2007b) is supported by seismic reflection datasets showing the development of arcuate faults and basement ridges in the East Orphan Basin (Gacal-Isler 2009) as well as shear zones along the northeastern margin of the cap (Welford et al. 2010c). At present, it is not clear how this transtensional dimension of rift complexity may have affected the conjugate Irish Atlantic margin.

The Irish Atlantic margin
The Irish continental shelf and many of the basins therein have been the focus of seismic reflection, seismic refraction and potential field investigations for several decades (Holder & Bott 1971;Whitmarsh et al. 1974;Bunch 1979;Jacob et al. 1985;Makris et al. 1988;Roberts et al. 1988;Lowe & Jacob 1989;O'Reilly et al. 1991O'Reilly et al. , 1995O'Reilly et al. , 2010Vogt et al. 1998;Landes et al. 2000Landes et al. , 2003Morewood et al. 2005;Hauser et al. 2008). Coupled with sparse well control, the resulting studies have allowed for the piecing together of the detailed tectonostratigraphic basin evolution of the margin (Shannon 1991;Naylor & Shannon 2009. The two largest and widest basins on the Irish margin are the Rockall and the Porcupine basins, with 6 km and 12 km thick sedimentary packages respectively O'Reilly et al. 2006). Underlain by very thin (<5 km in places) continental crust, as interpreted from the RAPIDS (Rockall and Porcupine Irish Deep Seismic) wide-angle seismic data (Fig. 2), these basins appear to have formed through differential stretching of the crust, and exhumation of serpentinized mantle has been interpreted beneath both basins O'Reilly et al. 1996O'Reilly et al. , 2006Reston et al. 2004). Although a few studies have attempted to compare and connect these and other structures and basins from the Irish margin across to the conjugate Newfoundland margin (Sinclair et al. 1994;Shannon et al. 1995), the main focus of previous bridging studies has been on the Goban Spur and its conjugate margin of Flemish Cap Peddy et al. 1989;Horsefield et al. 1994;Bullock & Minshull 2005).

The northeastern Flemish Cap and Orphan Basin margins
Multichannel seismic reflection coverage of the northeastern margin of Flemish Cap has been provided by surveys from the 1985 Frontier Geoscience Project (FGP) conducted by the Geological Survey of Canada (GSC), and the 1992 Erable Project conducted by the GSC and Ifremer (Srivastava & Sibuet 1992;Welford et al. 2010c). The main profile from the FGP project that sampled the northeastern margin of Flemish Cap is 85-3 ( Fig. 2), which extends SW-NE across the margin of the cap (de Voogd & Keen 1987;Keen et al. 1987a;Keen & de Voogd 1988). A portion of FGP profile 85-3 was also investigated with a coincident seismic refraction survey (Reid & Keen 1990) and most of the profile was the focus of the 2002 wide-angle seismic reflection/refraction Flemish Cap Margin Transect (FLAME; Jackson et al. 2002;Gerlings et al. 2011).
From the FLAME, FGP and Erable profiles, a narrow transition zone between continental and oceanic crust has been interpreted for the northeastern margin of Flemish Cap (Welford et al. 2010c;Gerlings et al. 2011). On the basis of modelling of seismic refraction data, this transition zone is consistent with exhumed serpentinized mantle similar to that interpreted beneath the thinned continental crust (Reid & Keen 1990;Gerlings et al. 2011).
Seismic reflection surveying of the Orphan Basin has been conducted by the petroleum exploration industry (Smee et al. 2003) and along two crustal-scale FGP profiles, 84-3 and 87-4 ( Fig. 2; Keen et al. 1987b;Keen & de Voogd 1988;Welford et al. 2010c). Profile 84-3 transects the Orphan Basin and extends SW-NE from the Bonavista Platform of the Grand Banks, across the Orphan Basin to beyond Orphan Knoll, an isolated continental fragment lying inboard of the continent-ocean boundary (Keen et al. 1987b;Keen & de Voogd 1988). Coincident wide-angle seismic reflection/ refraction modelling along this profile has revealed that the Orphan Basin is underlain by stretched continental crust extending more than 400 km to Orphan Knoll (Chian et al. 2001).

Inversion procedure
The inversion procedure undertaken for the present Orphan Basin-Flemish Cap study area was identical to that used for the Irish Atlantic margin study by Welford et al. (2010a), with the only differences being the overall extent of the inverted areas (outlined in Fig. 1) and the sources of the constraints. For both margins, the GRAV3D inversion algorithm (Li & Oldenburg 1996) was used to invert the observed regional free air gravity data and obtain a smooth depth-weighted subsurface 3D density anomaly model, subject to a prescribed reference model. Using independent bathymetric and depth to basement constraints, the reference model and inversion parameters were set to ensure that the density of seawater did not vary during the inversion and that sedimentary basin densities remained within depth-dependent limits consistent with Athy's Law and those observed on comparable passive margins (Albertz et al. 2010). On the basis of these constraints, the lowest density values within the basins were limited to the top 2 km, with densities rapidly increasing and approaching densities typical of basement rocks below, and with no hard density boundary imposed at the base of the sedimentary succession. Beneath the sedimentary basins, the inversion was given great flexibility in determining how it distributed the remaining mass needed to reproduce the gravity observations. Further details on how the reference models were constructed have been given by Welford et al. (2010a).
For the Irish margin, bathymetry was obtained from the General Bathymetric Chart of the Oceans (GEBCO) global 30 arc-second gridded bathymetric dataset (http://www.gebco.net/data_and_ products/gridded_bathymetry_data). Depth to basement for the Irish margin was computed using the bathymetry and minimum sediment thickness estimates from the National Geophysical Data Center (NGDC) of the National Oceanic and Atmospheric Administration (NOAA) Satellite Information Service (http:// www.ngdc.noaa.gov/mgg/sedthick/) with a localized adjustment to the thickness estimates to the south of Goban Spur, where more recent seismic constraints were available (Welford et al. 2010a). Equivalent constraints for the Orphan Basin and Flemish Cap region were obtained from the Canadian Hydrographic Service and from the Geological Survey of Canada (Grant 1988). The free air gravity anomalies used for both inversions were obtained from the DNSC08 gravity anomaly compilation from the National Space Institute of the Technical University of Denmark (Andersen et al. 2008). These satellite altimetry data represent an updated and augmented version of an earlier compilation by Sandwell & Smith (1997) of the results from both the Geosat Geodetic Mission and the ERS 1 Geodetic Phase mission.
All inversion parameters such as mesh cell size (5 km × 5 km × 500 m), maximum mesh depth (35 km), smoothing factor (25 km horizontally by 5 km vertically), target data misfit (two times the number of data points), reference density (2850 kg m −3 ), reference model setup and inversion bounds were kept the same so that the inversion results from both margins could be compared directly. As explored by Welford & Hall (2007) and Welford et al. (2010a), the correspondence between available seismically constrained boundaries and those inferred from the gravity inverted results is dependent on the combination of maximum mesh depth and reference density used. The reader is directed to the study by Welford et al. (2010a) for a more detailed explanation of the inversion method used and the justification for the parameters chosen.

Inversion results
Using the reconstruction of the Newfoundland and Irish Atlantic conjugate margins at the initiation of sea-floor spreading (magnetic anomaly 34) adapted from Srivastava et al. (1988) as the unifying framework (Fig. 2), six maps are presented in Figs 3-5 showing some of the inversion constraints and results from both margins. These maps are enhanced with simplified structural components on both margins to help highlight possible trends. The depth to basement constraints obtained for both margins correspond to the depth to acoustic basement as identified on multichannel seismic reflection data. As such, it may therefore not correspond to crystalline basement in all places. Nonetheless, Fig. 2. Bathymetric reconstruction of the Newfoundland and Irish Atlantic conjugate margins at magnetic anomaly 34 adapted from Srivastava et al. (1988).
Coordinates are relative to the presentday Newfoundland margin. Bathymetric contours are in metres. The locations and lateral extents of the major sedimentary basins are shown in tan colour (Enachescu et al. 2004c;Sibuet et al. 2007b;Naylor & Shannon 2011). Superimposed structural components on the Newfoundland margin were adapted from the structural map presented by Sibuet et al. (2007b) and originally modified from Enachescu et al. (2004a,b,c  several broad trends can be observed in the depth to basement map across the conjugate margins (Fig. 3a). Depth to basement is generally consistent at 5 km close to magnetic chron 34 on both margins but shows greater variability inboard. On the Irish margin, deepening and extreme deepening of the basement are observed in the Rockall Basin and the southern part of the Porcupine Basin respectively.
In the Orphan Basin, the depth to basement map is significantly more complex, with the base of the West Orphan Basin lying much deeper than the East Orphan Basin. To the south, the depth to basement in the Jeanne d'Arc Basin is similar to that in the West Orphan Basin, with the two basins appearing to be offset eastwest across the Cumberland Belt Transfer Zone. A similar trend was observed from sediment differences computed from an earlier gravity inversion study over the basins (Welford & Hall 2007). However, the significance of the Cumberland Belt Transfer Zone is debatable given the presence of Jurassic-aged sediment-filled troughs that appear to connect across the belt (Enachescu 2006).
Residual total magnetic anomalies on the reconstructed map (Fig. 3b) again show similar amplitude and size patterns in Fig. 4. Maps of juxtaposed present-day (a) observed and (b) predicted free air gravity anomaly on both margins on a reconstruction at the initiation of sea-floor spreading (at magnetic chron 34). The observed free air gravity anomaly data for each margin were obtained from the DNSC08 gravity anomaly compilation from the National Space Institute of the Technical University of Denmark (Andersen et al. 2008). Contours on both plots correspond to present-day bathymetry (1000 m contour interval). proximity to magnetic chron 34 whereas the patterns become more dissimilar inboard. The Porcupine Basin is characterized by a magnetic low along its axis whereas the Orphan Basin exhibits curved anomalies. In the West Orphan Basin, the curved anomaly trend runs oblique to many of the north-south to NNE-SSW faults and basement ridges in the basin (Sibuet et al. 2007b); whereas the trend of the magnetic low in the East Orphan Basin more closely resembles the fault trends (Gacal-Isler 2009). Similar to the Porcupine Basin, a negative magnetic anomaly roughly follows the axis of the Rockall Basin, whereas the southwestern part of the Porcupine Bank is characterized by positive anomalies that are similar in amplitude and size to those seen over the Orphan Knoll. These large irregular positive anomalies resemble those observed on the southeastern half of the Flemish Cap and outboard of Goban Spur, which may indicate similarities in the nature, thickness or rheology of the crustal basement.
Large positive free air gravity anomalies (Fig. 4a) characterize Flemish Cap, the western half of the West Orphan Basin, Orphan Knoll, Porcupine Bank and Goban Spur, whereas strong negative anomalies correspond to the eastern half of the West Orphan Basin, the Jeanne d'Arc Basin and portions of the East Orphan Basin. In contrast, portions of the Rockall Basin and the East Orphan Basin show relatively neutral gravity anomalies. The Porcupine Basin shows a complex gravity anomaly pattern that is divided by an inferred NW-SE-trending synrift fault system . To the north of the fault system, the basin contains a gravity anomaly high that correlates broadly with the Porcupine Arch (Naylor et al. 2002) and may correspond to serpentinized mantle beneath excessively thin crust O'Reilly et al. 2006). To the south of the inferred fault system, the Porcupine Basin more closely resembles the Rockall Basin in terms of neutral gravity anomaly and highly extended crustal structure with a rib of thicker crust in the centre of the basin (O'Reilly et al. 1996(O'Reilly et al. , 2006Shannon et al. 1999).
Whereas the high-low gravity couplet observed from the shelf break into the Orphan Basin on the Newfoundland margin is expected owing to the offset interplay between shallowing high-density mantle material and the thickening overlying low-density water column, the gravity high in the western half of the West Orphan Basin actually occurs off the shelf break in deep water. This gravity high zone, which runs from NW of Flemish Cap to the Charlie-Gibbs Fracture Zone, is also significantly wider than would typically be expected for a rifted margin and may correspond to an interpreted failed rift zone where higher density mantle material lies closer to the surface (Chian et al. 2001). The overall gravity low in the Orphan Basin near the White Sail Fault suggests sagging in the centre of the basin, possibly owing to a weak upper mantle. Comparison of the observed gravity anomalies (Fig. 4a) and those predicted from the inverted density anomaly models (Fig. 4b) generally shows an excellent correspondence with few, if any, significant differences. Moho depth proxies inferred by selecting a density anomaly isosurface corresponding to 170 kg m −3 (equivalent to a density of 3020 kg m −3 for a reference density of 2850 kg m −3 ) from the density anomaly models from the two margins are plotted in Figure 5a. On the Irish margin, the Moho proxy obtained from the regional gravity inversion agrees (to within 5 km) with over 80% of the Moho constraints obtained from a recent compilation of seismic refraction results (Kelly et al. 2007). Bearing in mind that the Moho depth will be underestimated in areas with anomalously high-density lower crust (generally owing to magmatism), the combined Moho depth map for the two margins (Fig. 5a) shows the deepest Moho beneath the Grand Banks, the Irish Mainland Platform and the Porcupine Bank, with a slightly shallower Moho beneath Flemish Cap and Goban Spur. It is shallower still beneath the Orphan, Rockall and Porcupine basins, and the shallowest Moho occurs near magnetic chron 34, the inferred locus of the initiation of sea-floor spreading (Srivastava et al. 1988).
Using the depth to basement and inferred Moho depth maps (Figs 3a and 5a), crustal thickness can be computed for the two margins (Fig. 5b). This map shows that crustal thickness is generally consistent at 5-10 km across the East and West Orphan basins. There is a moderate thickening of up to 15 km between the two basins in proximity to the White Sail Fault Zone, which extends out to Orphan Knoll. On the Irish margin, the Rockall Basin and the southern half of the Porcupine Basin have comparable crustal thicknesses to those in the Orphan sub-basins (Fig. 5b). Flemish Cap, the Porcupine High and the inboard portion of the Goban Spur are of similar thickness, whereas the outboard portion of the Goban Spur is significantly thinner. The crustal thickness variations across the Goban Spur may reflect the development of a set of generally small and elongate fault-bounded basins and highs during rifting and the opening of the North Atlantic (Fig. 2). They may also reflect, in part, variations in the original crustal thicknesses that developed during the Caledonian and Variscan orogenies, prior to Pangaean breakup.

Conjugate margin comparisons
Many of the trends observed in the reconstructed maps (Figs 3-5) can also be observed in depth slices extracted from both inverted models and matched up at magnetic chron 34 (Fig. 6). Three crosssections were selected to sample and illustrate the conjugate margins from relatively unstretched crust on one side to relatively unstretched crust on the other. Variations in the geometry of inferred upper, middle and lower crustal layers, defined by density differences, are described and interpreted.
Transect A-A′ (Fig. 6) crosses from the Bonavista Platform, through the West Orphan Basin and Orphan Knoll (a foundered continental fragment (Keen et al. 1987b;Chian et al. 2001)), across the Porcupine Bank and onto the Irish Mainland Platform. This transect reveals the significant difference in relatively undeformed crustal thickness between the Bonavista Platform, where it exceeds 35 km, and both the Porcupine Bank and the Irish Mainland Platform, which are closer to the regional average of 30 km for unstretched crust onshore in Ireland (Lowe & Jacob 1989).
Compared with these two relatively unstretched blocks on the Irish margin, the crust beneath the Orphan Basin on the Newfoundland margin is c. 5-10 km thinner. Moho depths between the two margins at magnetic chron 34 match very well, whereas small discrepancies in bathymetry and gravity anomaly are observed, which may be due to the promixity to the Charlie-Gibbs Fracture Zone, where minor effects of later transtensional movement have not been accounted for in the reconstruction. In terms of the inferred thickness of crustal layers, these are fairly evenly distributed beneath the Bonavista Platform, the Porcupine Bank and the Irish Mainland Platform. Densities typical of the upper crust are largely absent beneath the Orphan Basin and the Orphan Knoll. Significantly, the highly thinned crust immediately outboard of the Bonavista Platform on transect A-A′ (Fig. 6) has also been detected from seismic refraction modelling studies (Chian et al. 2001), where it was interpreted as the product of the development of a failed rift.
Transect B-B′ (Fig. 6) crosses from the Bonavista Platform, through the East Orphan Basin and the southern part of the Porcupine Basin, onto the Celtic Platform. Again, a significant difference is observed between unstretched crustal thickness of the Bonavista Platform (c. 35 km) and the Celtic Platform (c. 25 km), whereas the thinning beneath the basins appears more uniformly distributed. Slight discrepancies are observed between the matchups of Moho depth, bathymetry, top of lower crust and gravity anomaly at magnetic chron 34 but overall the slices display similar crustal density distributions. Whereas the inferred lower crust maintains a uniform thickness across the entire cross-section, upper crust thins out beneath most of the East Orphan Basin and immediately outboard of the Celtic Platform (Fig. 6). In contrast, inferred middle crustal densities thin out near magnetic chron 34. It is interesting to note that the inferred failed rift identified along transect A-A′ (Chian et al. 2001) appears wider and less obvious in terms of crustal thickness along transect B-B′.
Transect C-C′ (Fig. 6) extends across the inferred conjugate pair of Flemish Cap and Goban Spur, although the appropriateness of this link is put into question by the postulated rotation of Flemish Cap out of the Orphan Basin (Sibuet et al. 2007b). Nevertheless, the crustal thickness of Flemish Cap matches relatively well with the inboard limit of Goban Spur at 20-25 km, which lends support to the hypothesis of an ancient linkage between the two. Outboard, Flemish Cap thins more abruptly than Goban Spur, with upper and middle crustal densities pinching out more rapidly than on the Irish margin. The lack of faulting observed on Flemish Cap (Funck et al. 2003;Hopper et al. 2006), and the broad area of faulting leading to the development of a set of elongate basins and highs documented for the Goban Spur Naylor & Shannon 2009), may have provided a more distributed zone to accommodate thinning on the Irish margin. Moho depth and gravity anomaly at magnetic chron 34 match very well along this transect (Fig. 6) whereas a significant jump is observed in the bathymetry. Along-margin cross-sections extracted from the density anomaly models (Fig. 7) illustrate the first-order differences between the Newfoundland and Irish margins, as well as illustrating alongstrike variations on both margins. On the Irish margin, the crustal thickness appears bimodal. The crust is uniformly thick beneath the Rockall Bank, the Porcupine Bank and the Goban Spur, whereas it is very thin beneath the Rockall and Porcupine basins. Densities typical of the upper crust are pinched out below the basins but are more evenly distributed relative to middle and lower crustal densities beneath the unstretched blocks. On the Newfoundland margin, whereas the unstretched Flemish Cap resembles its Irish equivalents in terms of crustal thickness and crustal layer thicknesses, the Orphan Basin appears to be more laterally stretched out with greater pinching out of upper crustallike densities towards Flemish Cap in the East Orphan Basin. From these cross-sections (Fig. 7), it is apparent that thinning was focused and compartmentalized into discrete large basins located between more competent crustal blocks (Rockall Bank, Porcupine Bank and Goban Spur) on the Irish margin whereas thinning was more uniformly distributed within the Orphan Basin on the Newfoundland margin. On both margins, the competent crustal blocks (Flemish Cap, Porcupine Bank and Goban Spur) display broadly similar crustal density distributions, which are consistent with a likely earlier structural linkage and common history.

Discussion
To compare variations in the amount and distribution of extension experienced on the Flemish Cap-Orphan Basin and Irish Atlantic conjugate margins, the crustal thicknesses derived from the inversions (Fig. 5b) are plotted in Figure 8a as a percentage of the unstretched thickness of the Flemish Cap and the Irish crust (30 km; Funck et al. 2003 andLowe &Jacob 1989 respectively). Whereas the unstretched crustal thickness for the Irish crust is generally consistent at 30 km for the entire margin (Lowe & Jacob 1989;Vogt et al. 1998;Landes et al. 2005;O'Reilly et al. 2006;Ziegler & Dèzes 2006), the broader crustal thickness variability for the offshore Newfoundland margin (30 and 36 km beneath Flemish Cap and the Grand Banks respectively; Funck et al. 2003 andLau et al. 2006a), means that the crustal thickness percentages on the Grand Banks, which are not the focus of this comparison study, occasionally exceed 100% in Figure 8a.
Using the crustal thicknesses, the extension or stretching factor, β, which corresponds to the ratio of extended to unextended length, is computed across the conjugate margins by taking a 1 km long block of 30 km thick crust and computing the extended length for the varying crustal thicknesses by assuming that the cross-sectional area remains constant. The computed stretching factors are plotted in Figure 8b and to aid in the interpretation of the figure, two key contours are highlighted. The first, where β = 2 and indicated by the dotted white line, corresponds to the stretching factor above which polyphase faulting can obscure older faulting and make it difficult to quantify the total amount of extension based on seismically imaged faults alone (Reston 2007). The second contour, where β = 3.5 and indicated by the dotted dark grey line, corresponds to the stretching factor above which embrittlement of the entire crust is thought to be possible and serpentinization of the uppermost mantle can occur based on 1D numerical modelling studies Pérez-Gussinyé et al. 2003).

Variations in rifting style
It is evident from Figure 8 that the greatest amount of thinning is beneath the southern part of the Porcupine Basin (less than 20% of original crustal thickness), whereas thinning in the West Orphan, the East Orphan, the Jeanne d'Arc and the Rockall basins is slightly less (<20% of original crustal thickness). Such localization of significant thinning is also highlighted in the stretching factor plot (Fig. 8b) and coincides with β values greater than 3.5, as outlined by the dotted dark grey contour. It may represent regions where serpentinization of the uppermost mantle has occurred in response to rheological embrittlement. Seismic velocity modelling evidence for such serpentinization processes has been documented beneath the Rockall Basin (O'Reilly et al. 1996), where stretching factors of the order of 5-6 have been determined (Morewood et al. 2005), and the Porcupine Basin (Reston et al. , 2004, where some serpentinized mantle is interpreted to have been exhumed (O'Reilly et al. 2006). However, the only crustal-scale seismic refraction profiling results available for the Orphan Basin cross the southern limit of large β values in the West Orphan Basin and show significant thinning of the crust to 6-8 km, but typical unaltered mantle velocities are derived beneath the thinned crust (Chian et al. 2001).
Using geodynamic modelling, Huismans & Beaumont (2011) have demonstrated that depth-dependent extension is necessary to reproduce many of the basic properties of non-volcanic rifted margins, and that the style of such depth-dependent extension can be explained by different end-member rifting models. For exhumation of serpentinized mantle to occur at the sea floor as is widely observed between Newfoundland and Iberia (Péron-Pinvidic et al. 2007;Welford et al. 2010b), the crust must rupture before the mantle lithosphere does. This requires a strong lithosphere (Huismans & Beaumont 2011). Conversely, to obtain wide zones of thinned continental crust without mantle exhumation, the mantle lithosphere must rupture before the crust. This requires decoupling of the upper and lower lithosphere along a weak crustal detachment layer. With time, the ruptured continental mantle in this second rifting model is advected away to be replaced either by hot asthenosphere, which will eventually cool to form oceanic lithosphere, or by hot depleted lower cratonic lithosphere (Huismans & Beaumont 2011).
Because Figure 8b gives only an estimate of overall crustal thinning, it is impossible to quantify the proportion of the tectonic extension accommodated within the various crustal layers. Based on the seismic refraction modelling results for the Rockall and the Porcupine basins, where evidence of serpentinized mantle has been detected beneath the thinned crust (O'Reilly et al. 1996(O'Reilly et al. , 2006Reston et al. 2001Reston et al. , 2004, a strong lithosphere seems to be required based on recent geodynamic models (Huismans & Beaumont 2011). Conversely, the lack of evidence for serpentinized mantle beneath the Orphan Basin based on more limited seismic refraction modelling (Chian et al. 2001) requires a weak crustal layer (lower crust?) and rupture within the mantle lithosphere. Hence,   8. Maps of juxtaposed present-day (a) crustal thickness (derived from the regional 3D gravity inversions) as a percentage of the unstretched thickness of both Flemish Cap and the Irish crust (30 km) and (b) stretching factor, β, on both margins on a reconstruction at the initiation of sea-floor spreading (at magnetic chron 34). In (b), the β = 2 contour corresponds to the stretching factor above which polyphase faulting becomes important (Reston 2007) and the β = 3.5 contour corresponds to the stretching factor above which embrittlement of the entire crust is possible Pérez-Gussinyé et al. 2003). These contours are highlighted by the white and dark grey dotted lines respectively. the typical mantle velocities observed beneath the thin crust of the Orphan Basin (Chian et al. 2001) may indicate that rifting moved outboard of the basin before the mantle lithosphere could be completely advected away as in the Huismans & Beaumont (2011) model. The lack of serpentinized mantle beneath the Orphan Basin does not necessarily mean that its crust differed greatly from the crust on the Irish margin but rather that the Irish crust failed more abruptly as the serpentinized mantle beneath it acted as a ductile detachment. Relatively less weak lower crust may exist on both margins, with the main difference between the rifting styles of the two margins being due to the localization of rifting and the ability of the rifting to generate crustal-scale faults that could allow serpentinization of the uppermost mantle.
The results from this study suggest a fundamental difference in the rheological properties of the lithosphere between the Orphan Basin-Flemish Cap region and Irish Atlantic margin. Although the rheological variations may be due to differences in the original crustal compositions of the margins, it is also possible that the rheological variations developed owing to the manner in which rifting was focused on each margin. Ultimately, focused rifting on the Irish margin led to deeper fractures, highly thinned crust and serpentinization of the mantle between strong, rheologically more robust, blocks. On the Newfoundland margin, extension was accommodated over a broader area so the faults may not have penetrated as deep and the weakest crustal layer (lower crust?) may have controlled the formation of the massive Orphan Basin without any significant serpentinization in the uppermost mantle.

Rifting evolution
Several key findings are apparent from Figure 8a. The first is that Flemish Cap (10% thinned) and the outboard part of the Goban Spur (50-60% thinned) are distinctly different in their style of thinning. Assuming that these differences did not exist prior to Pangaean breakup, two opposing scenarios can be envisaged. If the Flemish Cap and Goban Spur were originally conjugate to each other, then the NE-SW extension between them was largely accommodated by rifting and crustal thinning in the outboard portion of the Goban Spur with Flemish Cap acting as a strong and rigid microplate (Hopper et al. 2007;Sibuet et al. 2007b). Alternatively, the two margins may not have been originally conjugate to each other. This scenario is consistent with the rotation of Flemish Cap out of the Orphan Basin (Sibuet et al. 2007b) and also with the indications that they consist of different basement domains ( Fig. 1) with Flemish Cap being Avalonian in terrane characteristics (King et al. 1985) and Goban Spur being of Variscan affinity (Tyrrell et al. 2010). Although the basement domain constraints are admittedly sparse, the rotational model is preferred, as it is supported by seismic evidence of transtensional faulting along the northeastern margin of Flemish Cap (Welford et al. 2010c).
As discussed above, Figure 8a highlights the lack of any substantial thinning beneath Flemish Cap, Rockall Bank, Porcupine Bank and the adjacent continental platforms, relative to the intervening crust below the sedimentary basins themselves. A schematic, preliminary illustration of the movement of these relatively unstretched crustal blocks is given in Figure 9, where the positions of these blocks are marched back from the initiation of sea-floor spreading (c) to the start of rifting (a). This speculative and qualitative reconstruction is formulated on the basis that Flemish Cap has been rotated 43° and displaced 200-300 km out of Orphan Basin as argued by Sibuet et al. (2007b). The NW-SE transfer zones proposed by Readman et al. (2005) to explain the rifting evolution of the Porcupine Basin are overlain on Figure 9c to attempt to understand their role in the broader rifting history of the two margins. For simplicity, and owing to the lack of other constraints, the transfer zones are kept fixed in orientation relative to the Porcupine Bank as it rotates back to close the Porcupine Basin at the initiation of rifting (Fig. 9a).
Some simple but important insights can be gleaned from Figure  9. At the initiation of rifting, Flemish Cap, Orphan Knoll, Porcupine Bank, Goban Spur and Galicia Bank were connected in the past, possibly in a configuration like that shown in Figure 9a. As continental rifting between Newfoundland and Iberia progressed northward from the south, the delayed breakup between Flemish Cap and Galicia Bank is attributed to extension being accommodated west of Flemish Cap, in the Orphan basins (Welford et al. 2010b). At some point in geological time, this rift basin and the roughly parallel Porcupine Basin rift failed, with rifting ultimately succeeding between Flemish Cap and Porcupine Bank.
On the Newfoundland margin, the northern limit of the failed rift zone coincides with the Charlie-Gibbs Fracture Zone, the oceanward extension of the Dover Fault, an ancient Pangaean suture, which separates the Gander and Avalon basement terranes onshore Newfoundland. On the Irish margin, the connection between the failed rift and major pre-existing crustal sutures is less obvious. At the initiation of sea-floor spreading in Figure 9c, the Charlie-Gibbs Fracture Zone could be extrapolated to the Variscan Front, whose oceanward continuation is uncertain. Alternatively, the schematic reconstruction may indicate that at the initiation of rifting (Fig. 9a), the Charlie-Gibbs Fracture Zone could be extended southwestwards from the Caledonian Iapetus Suture. Complicating the story even further are the NW-SE transfer zones proposed by Readman et al. (2005), which could have accommodated varying amounts of extension and transtension between the ancient sutures. The Clare Lineament (not shown), which would lie immediately to the south of the NW-SE transfer zones of Readman et al. (2005), has been proposed as the ancient precursor to the Charlie-Gibbs Fracture Zone (Tate 1992), but the lineament's very existence is questionable based on more recent improved gravity data.
Although this study does not allow for the detailed interplay between ancient sutures and transfer zones to be elucidated, the resolved lithospheric structures and their spatial correlations do suggest that the complex rifting evolution between the Flemish Cap-Orphan Basin and Irish Atlantic margins was strongly influenced by them. These correlations demonstrate the wellknown importance of such terrane boundaries in influencing the locus of global rifting in the North Atlantic (Naylor & Shannon 2011). Variations in the style of rifting that occur in a NE to SW direction across the North Atlantic region, between the conjugate margins, are also significant. They are probably associated with trans-Atlantic ('along-strike') variations in the rheological properties of the lithosphere within the major terranes shown in Figure 1.
Finally, the distribution of inferred serpentinized mantle (exhumed or in situ) beneath the Porcupine Basin (O'Reilly et al. 2006) and offshore Goban Spur (Bullock & Minshull 2005) and Flemish Cap (Gerlings et al. 2011) compared with the lack of evidence for serpentinization beneath Orphan Basin (Chian et al. 2001) reveals interesting regional variations in the broad rheological character of the reconstructed margin. These may be revealing strong crust or mantle lithosphere for the Irish margin and its connection to Flemish Cap and a weak crustal layer for the Orphan Basin, based on geodynamic modelling results (Huismans & Beaumont 2011). It is not clear whether these rheological variations were intrinsic characteristics of the pre-rift crust or if the distribution and depth extent of localized faulting generated weak serpentinized zones on the Irish margin that failed more abruptly compared with the Orphan Basin, where only the lower crust was weak but the upper mantle remained strong. Ultimately, it is possible, as Figure 9a suggests, that the lithosphere to the NE of the failed Orphan Basin rift was stronger and that Flemish Cap crust was originally part of the Irish Atlantic margin with the Porcupine Bank rather than Goban Spur as its conjugate at the initiation of rifting.

Conclusions
Regionally constrained 3D gravity inversions of the free air data over the conjugate Flemish Cap-Orphan Basin and Irish Atlantic continental margins have provided the first complementary crustalscale geophysical datasets with which to compare variations in crustal structure at locations both across-and along-strike of the conjugate margins. Projection of these results onto a palaeoreconstruction of the margins at the initiation of sea-floor spreading (magnetic chron 34) provides maps from which crustal thickness, density structure and Moho depth can be compared. Matched across-strike and along-strike slices through the density anomaly models provide additional views from which to compare and contrast the margins. Key findings include the following.
(1) The Irish Atlantic margin exhibits regionally compartmentalized rifting with extremely thinned crust beneath a small number of large sedimentary basins bounded by relatively unstretched crustal blocks. On the Newfoundland margin, the crust is more uniformly stretched across the broad Orphan Basin, adjacent to the relatively unstretched and intact Flemish Cap thick crustal block.
(2) Zones of extreme crustal thinning are observed beneath the southern part of the Porcupine Basin, the western half of the West Orphan Basin, the eastern half of the Jeanne d'Arc Basin, the southeastern half of the East Orphan Basin and in pockets within the Rockall Basin. Although serpentinization of the mantle lithosphere (and possible exhumation) has been inferred in most of these basins from seismic refraction modelling, there is no evidence for serpentinization beneath the Orphan Basin. Geodynamic modelling suggests that these differing rifting styles may correspond to rheologically strong crust-mantle lithosphere for the Irish Atlantic margin and for Flemish Cap, with a weaker crustal layer beneath Orphan Basin. These rheological differences may be inherited or they may have evolved from localized variations in faulting, with focused, full-crustal faulting on the Irish margin-Flemish Cap leading to the formation of a very weak subcrustal serpentinized zone whereas more distributed faulting in the Orphan Basin allowed for a wider zone of extension over a relatively less weak lower crustal layer. (3) A schematic tectonic reconstruction of the Flemish Cap-Orphan Basin and Irish Atlantic margins at the start of rifting, assuming that Flemish Cap was rotated and displaced out of Orphan Basin, supports the ancient connection between Flemish Cap, Orphan Knoll and the Porcupine Bank rather than between Flemish Cap and Goban Spur.
The complex rifting evolution of the Flemish Cap-Orphan Basin and Irish Atlantic margins appears to have been controlled by the interplay between ancient orogenic sutures and transfer zones as well as localized variations in lithospheric strength both across and along strike of the conjugate margins. The rheological variations may be intrinsic characteristics of the pre-rift crust or may have developed as a result of localized variations in the depth and distribution of faulting. These results underline the interpretative value of regionally constrained 3D gravity inversions for delineating crustal structures in regions lacking sufficient deep seismic control and for tying structures across reconstructed rifted continental margins.