To investigate the effect of crustal heterogeneities inherited from previous tectonic phases on magma-poor rifting processes, we performed numerical experiments of lithospheric extension with initial conditions that included strength variations from inherited crustal fabrics. Crustal fabrics were introduced in the model by using an element-wise bimineralic composition in which mineral phases were distributed in a way that was compatible with the orientation and distribution of kilometric-scale heterogeneities observed in seismic reflection data. Our numerical models show that strength variations from inherited crustal fabrics strongly influence the mechanisms of deformation in the stretching and thinning phases of rifting. The strength variations also generate alternative models for the evolution of faulting during distributed stretching and localized thinning phases that are usually associated with detachment or sequential faulting models. During the stretching phase, inherited strength variations control the distribution and the processes of deformation. Vertical fabrics favor the formation of horst-and-graben structures. Horizontal and dipping fabrics favor the formation of detachment faults and core complexes. During the thinning phase, processes differ depending on the orientation of the crustal fabrics and involve either a combination of detachment faults and sequential normal faults or an alternative model in which deformation remains decoupled between the upper crust and lithospheric mantle, with the formation of high-angle faults in the upper crust and a low-angle detachment fault in the upper mantle. As a consequence, strength variations inherited from crustal fabrics also control the resulting geometry of the margin and the width of the necking and hyperextended domains. Finally, our models demonstrate that inherited crustal fabrics do not control breakup and mantle exhumation. These processes are ubiquitously associated with the development of new detachment faults exhuming mantle to the seafloor.

Passive margins define about half of Earth’s coastlines and have been the focus of many geological and geophysical studies in the last decades. While great progress has been made in understanding the mechanics of extensional deformation, many fundamental questions remain about the effects of inherited geological conditions on localization processes at rifted margins. Several studies suggest that inheritance is a key control on the development of rift structures (Dunbar and Sawyer, 1989; Ring, 1994; Piqué and Laville, 1996; Corti et al., 2007; Clerc et al., 2015; Manatschal et al., 2015). Inheritances are the result of the successive tectonic events that affect the continental lithosphere during its complex geological history. Although they are interrelated, geologists usually distinguish three types of inheritances: compositional, structural, and thermal. In the literature, most of the studies focus on the effect of structural inheritances (Ring, 1994; Corti et al., 2004, 2007; van Wijk, 2005; Autin et al., 2013; Chenin and Beaumont, 2013) and thermal inheritances (Buck, 1991; Brune et al., 2014, 2017; Svartman Dias et al., 2015) on rifting localization. Structural inheritances are defined as mechanically weak shear zones inherited from previous orogenic events. Studies suggest that they can control the localization of deformation from the beginning of rifting and rejuvenate lithospheric structures that are properly oriented with respect to the direction of extension (Harry and Sawyer, 1992; Ring, 1994; Corti et al., 2004, 2007; Autin et al., 2013; Chenin and Beaumont, 2013). However, according to Manatschal et al. (2015), structural inheritances do not significantly control the location of breakup. Thermal inheritances can cause variations in the degree of coupling between crustal and mantle deformation, which in turn controls the long-term evolution and architecture of rifts (e.g., Manatschal et al., 2015). The rifting of old, cold lithosphere, with strong coupling between the upper brittle crust and mantle, results in the formation of narrow rifts, whereas rifting of a young, warm lithosphere with a thick decoupled lower crust results in the formation of a wide rift (Bassi, 1991; Bassi et al., 1993; Bassi, 1995; Buck, 1991; Brune et al., 2014, 2017; Svartman Dias et al., 2015). Field and seismic observations clearly demonstrate that the composition of the crust is compositionally heterogeneous (Smithson, 1978; Rudnick and Fountain, 1995). However, due to their apparent complexity, little attention has been given to the effects of compositional inheritances on the rifting process. Indeed, in most numerical experiments of lithospheric extension, the composition of the crust and mantle is assumed to be layered and homogeneous and composed of wet or dry plagioclase, quartz, or olivine (Buck, 1991; Lavier and Buck, 2002; Huismans and Beaumont, 2003, 2007, 2011; Huismans et al., 2005; van Wijk and Blackman, 2005; Gueydan et al., 2008; Rosenbaum et al., 2010; Duretz et al., 2016). To take into account the heterogeneities of the lithosphere, some numerical models use a laboratory-determined flow law for polymineralic rock like granite, quartz-diorite, diabase, or gabbro (Dunbar and Sawyer, 1989; Lavier and Manatschal, 2006; van Wijk and Blackman, 2005). By using a bulk strength envelope for the polymineralic aggregate, these studies do not explicitly take into consideration the interaction between the different minerals and imply, as for monomineralic assemblages, that rheology is either elastoplastic, in order to simulate a brittle upper crust and upper lithospheric mantle, or viscous/viscoelastic to simulate a ductile middle to lower crust and lower lithospheric mantle. Consequently, in numerical studies, the role of compositional inheritance has mainly been tested by comparing models in which the globally averaged crustal or mantle compositions vary. For example, Svartman Dias et al. (2015) compared models in which the crust is made of either dry quartz or plagioclase, and the mantle composition is wet or dry olivine. The main problem with such approaches is that the overall lithospheric composition remains homogeneous and layer-caked, and deformation at the brittle-ductile transition is constrained to occur at the sharp transition between brittle and ductile material.

Observations of the brittle-ductile transition show strong evidence of semibrittle deformation at the scale of rock or outcrop. One can observe that over the brittle-ductile transition, porphyroclasts remain slightly deformed or exhibit localized fractures, while the surrounding matrix shows evidence of ductile deformation (e.g., Wakefield, 1977; Mitra, 1978; White et al., 1980; Handy, 1990, 1994; Jammes et al., 2015). At the mesoscale (meters to kilometers), undeformed lenses of material surrounded by mylonitic shear zones lead to the formation of large anastomosing patterns or meter-scale boudinage structure. Such structures are well described at the fossil brittle-ductile transition exposed in Cap de Creus, Spain (Carreras, 2001; Fusseis et al., 2006) or in seismic images of the middle to lower crust along the Uruguayan margin (Clerc et al., 2015). Such observations demonstrate that a natural strength contrast between mineral phases or mineral aggregates exists, not only at the microscale, but also between meter- to kilometer-scale units of significantly different composition. The rheology and deformation processes at lithospheric scale are consequently controlled by the interaction between metric to kilometric blocks of strong or weak average composition—an observation that emphasizes the importance of compositional inheritances in tectonic deformation processes. In a previous study, to test the effect of mesoscale compositional heterogeneities on deformation processes, we performed numerical experiments of rifting using an explicit bimineralic composition in the crust and/or the mantle with a random distribution of mineral phases among particles located in each element of the model (Jammes et al., 2015; Jammes and Lavier, 2016). This complex rheology was introduced to take into account the interaction between kilometric blocks of strong material (approximated by the rheology of plagioclase for the crust or olivine minerals for the mantle lithosphere) and weak material (approximated by the rheology of quartz for the crust or orthopyroxene minerals for the mantle lithosphere), representing, for example, the presence of mafic clusters in a globally felsic crust. By comparing bimineralic models to numerical simulations using an implicit bimineralic composite (an average viscous flow law for a two-phase aggregate; Tullis et al., 1991) in the crust and the mantle, Jammes and Lavier (2016) demonstrated that an explicit bimineralic approach approximating rheological heterogeneities in the crust and mantle succeeds in reproducing the following structural features related to the formation of magma-poor rifted margins: (1) the absence of a sharp deformation zone at the brittle-ductile transition; (2) the initiation of the rifting process as a wide delocalized rift system with multiple normal faults dipping in both directions; (3) the development of anastomosing shear zones in the middle/lower crust and the upper lithospheric mantle similar to the crustal-scale anastomosing patterns observed in the field (Carreras, 2001; Fusseis et al., 2006) or in seismic data (Clerc et al., 2015); and (4) the preservation of undeformed lenses of material leading to lithospheric-scale boudinage structures and resulting in the formation of continental ribbons, as observed along the Iberian-Newfoundland margin.

Following these results, we believe that using an explicit bimineralic assemblage is a better approximation of the rheological complexity of the lithosphere and yields a better understanding of rifting processes. However, this previous work (Jammes et al., 2015; Jammes and Lavier, 2016) used a random distribution of heterogeneities unconstrained by any observations. Heterogeneities in the crust are not completely random; they preserve a structural pattern (fabric) inherited from a complex tectonic history. Here, we designed numerical experiments with initial conditions that included strength variations inherited from crustal fabrics. Crustal fabrics were parameterized as variations in the orientation and mineralic composition of the crust and were derived from two-dimensional (2-D) seismic observations. The objective of this study was to understand how strength variations from inherited crustal fabrics influence the mechanisms of deformation during rifting processes and how they affect the resulting geometry of the margins. Magmatic intrusions are also known to play an important role during extensional processes, but their basic physics still need to be clarified before they can be consistently included in the mechanics of rifting (e.g., Qin and Buck, 2005; Davis and Lavier, 2017). We therefore focused our study on the effect of inherited strength variations on rifting processes in magma-poor rift settings.

After reviewing prerift crustal structures, rifting mechanisms in magma-poor margins, and geometry, we present and compare the results of our numerical study to observations of magma-poor margins and the most recent models attempting to explain the mechanics of rifting. We show that strength variations inherited from crustal fabrics strongly affect rifting processes, including: (1) the distribution of strain during the initial phase of deformation, (2) the mechanism of thinning during the rifting process, and (3) the mechanism leading to mantle exhumation and/or the formation of oceanic crust.

Field and seismic observations clearly demonstrate that the composition of the crust is highly heterogeneous at the mineralic scale, but also at the lithospheric scale, where kilometer-scale units of significantly different composition may be juxtaposed (Smithson, 1978; Rudnick and Fountain, 1995). Structural inheritances, defined as mechanically weak zones inherited from previous orogenic events, also contribute in defining the complexity of the crustal structure. They may affect the entire lithospheric thickness, but brittle structures are limited to the shallowest part of the crust (Ring, 1994; Corti et al., 2004, 2007), while anastomosing ductile shear zones affect the deeper part of the crust. The two seismic profiles presented in Figure 1 are onshore deep seismic profiles illustrating the complexity of crustal composition and structure of eastern Canadian continental crust (for location, see Fig. 2). Located onshore, these profiles depict continental crust that is not strongly affected by extensional processes. We can therefore use them as analogues of prerift crustal structure.

These two profiles (Figs. 1A and 1B) show strong variability in seismic reflectivity. Using theoretical models and numerical simulation, Hurich and Smithson (1987) and Hurich (1996) demonstrated that the scale of compositional variability can be measured effectively in seismic reflection profiles. These results are consistent with other studies demonstrating that deep seismic reflectivity is strongly controlled by lithological variation (Green et al., 1990; Ji et al., 1997). For ductile shear zones, studies show that they may be transparent, unless they are compositionally layered (Hurich et al., 1985) or contain significant volumes of foliated phyllosilicates (Jones and Nur, 1984). As a consequence, the seismic variability observed in Figure 1 can be appropriately parameterized as compositional heterogeneities, interpreted as crustal structures of variable orientation and scale. The apparent dips of the crustal structures vary from horizontal to dipping (20° to 50° dip) to vertical, whereas the scales of structures range from 5 km to 25 km long. The aim of our numerical experiment was to understand the effects of the orientation of strength variations inherited from crustal fabrics on rifting processes in a magma-poor setting. We therefore focused on end-member models in which crustal fabrics are 10–15 km long and horizontal, vertical, or dipping (30° dip).

Forty years ago, two end-member mechanisms were proposed to explain extension of the lithosphere in magma-poor settings: pure shear with uniform stretching (McKenzie, 1978), and simple shear, where extension is accommodated by a low-angle detachment fault (Wernicke and Burchfiel, 1982). Since then, progress has been achieved in understanding extensional deformation, and it is now commonly accepted that rifting in magma-poor settings is a polyphase process that involves different deformation mechanisms (i.e., both pure shear and simple shear processes; Lavier and Manatschal, 2006; Reston, 2009; Péron-Pinvidic and Manatschal, 2009; Péron-Pinvidic et al., 2017). Lavier and Manatschal (2006) showed that three main phases of deformation during the rifting process can be identified: distributed stretching, localized thinning, and exhumation. Deformation processes during each phase are characterized by the degree of coupling of deformation between the crust and mantle lithosphere (Reston, 2009; Péron-Pinvidic and Manatschal, 2009; Jammes et al., 2010). In the initial stretching phase, the deformation is decoupled and mostly controlled by pure shear processes: Brittle faults initiated in the upper crust are rooted in the ductile middle and lower crust and remain decoupled from the deformation accommodated in the brittle upper mantle. During this phase, rifting appears to be a largely symmetric process on a crustal scale. The subsequent localized thinning phase involves stretching of the crust, resulting in its complete embrittlement, and faulting in the brittle mantle lithosphere that can connect with the faults formed in the remaining brittle crust. During the final exhumation phase, circulation of water through the fractures into the mantle lithosphere leads to its serpentinization. As a result, large asymmetric detachment faults develop in the weak mantle serpentinites, producing a late-stage asymmetry and exhuming serpentinized mantle on the seafloor (Reston, 2009; Péron-Pinvidic and Manatschal, 2009; Jammes et al., 2010). This late asymmetric rifting process results in the formation of an “upper-plate margin” in the hanging wall and a “lower-plate margin” in the footwall (Lister et al., 1991). The upper-plate margin is described as a narrow, sharp margin, constituted of faulted upper crustal blocks (Reston, 2009; Péron-Pinvidic and Manatschal, 2009; Péron-Pinvidic et al., 2017), whereas the lower-plate margin is highly structured, faulted, and hyperextended as a result of movements along exhuming detachment faults overlain by tilted crustal blocks and extensional allochthons (Reston, 2009; Faleide et al., 2010; Huismans and Beaumont, 2011; Péron-Pinvidic et al., 2017). While published magma-poor rifting models globally agree on these first-order rifting mechanics, many differences between the models remain at second order.

In the magma-poor margin literature, two main polyphase rifting models are discussed. In these two models, deformation mechanisms differ mostly during the pre-exhumation phases and more precisely during the thinning phase. In the model deriving from the work of Pérez-Gussinyé et al. (2003), Pérez-Gussinyé (2013), Reston (2005, 2007, 2009), and Ranero and Pérez-Gussinyé (2010), extreme crustal thinning is interpreted to occur by sequential normal faulting. On the other hand, in the model deriving from the work of Whitmarsh et al. (2001), Manatschal et al. (2001), Manatschal (2004), Lavier and Manatschal (2006), and Péron-Pinvidic and Manatschal (2009), extreme crustal thinning is interpreted to occur by detachment faulting. In the following, we will describe these two models in order to use them as a basis for discussion in the description and interpretation of our results.

Sequential Normal Faulting Models (Figs. 3A–3D)

In this model (Reston, 2007, 2009), the mechanism of pre-exhumation deformation is divided into two main phases. During the first phase of deformation, extension is distributed over a broad area along high-angle faults of different orientations. During the second phase of deformation, extension localizes, and boudinage of the lower crust occurs, while upper-crustal deformation is accommodated along a series of high-angle faults mostly dipping in the same direction (Fig. 3A). Fault slip and rotation cause further thinning of the crust on the first generation of faults until they lock (first generation of faults lock up). Locking occurs when the dip and static friction angle of the faults are not compatible with slip. As a result, a second generation of faults forms at more favorable high-angle dips, thinning the crust further. The previously formed tilted blocks are then truncated by high-angle normal faults (Fig. 3B). The same pattern is then repeated, resulting in a gradual thinning of the crust toward the future ocean (Pérez-Gussinyé and Reston, 2001; Reston, 2007, 2009). When complete embrittlement of the crust has been achieved, water can diffuse through the deformed crust, leading to the serpentinization of the underlying mantle. Mantle unroofing is the result of continued extension as large asymmetric detachment faults can develop in the serpentinized mantle during the exhumation phase (Fig. 3C). As extension continues, crustal separation occurs along the detachment fault, leading to the initiation of seafloor spreading and the formation of new oceanic crust (Fig. 3D).

Detachment Faulting Model (Figs. 3i–3iv)

In this model, the stretching phase is characterized by high-angle faults associated with classical half-graben subsidence. During this phase, deformation is distributed over a broad region in which continental crust is slightly stretched (Fig. 3i). As extension continues, deformation localizes along two conjugate faults that are decoupled from the mantle along a midcrustal décollement. These faults delimit an upper-crustal block (identified as the H block in Lavier and Manatschal, 2006) and evolve from high-angle faults to low-angle detachment faults, exhuming deep crustal and/or mantle material underneath the H block (Fig. 3ii). Through this process, detachment faulting results in local exhumation of basement rocks and extreme and localized thinning of the crust at the transition between the proximal and distal margins (e.g., the pronounced necking zone observed on the Newfoundland margin; Péron-Pinvidic and Manatschal, 2009). This extreme thinning is associated with an embrittlement of the crust that can lead to the exhumation phase. Detachment faults can crosscut the remaining crust and exhume serpentinized mantle rocks at the seafloor (Fig. 3iii). Final seafloor spreading is defined by the irrevocable localization of thermal and mechanical processes in a narrow zone corresponding to a protoridge (Fig. 3iv).

In these two models, the final exhumation phase is controlled by the initiation of a large detachment fault that exhumes serpentinized mantle to the seafloor, resulting in an asymmetric conjugate margin. The upper-plate margin corresponds to the left-side margin of both models, and the lower-plate margin corresponds to the right-side margin (Fig. 3). As previously described, the upper plate is sharp and narrow and composed of faulted upper-crustal blocks resulting from the sequential faulting models (Fig. 3D) or from the dismantling of the H block (detachment faulting model, Fig. 3iv). On the other side, the lower plate is more extended as a result of detachment faults overlain by tilted crustal blocks and extensional allochthons.

The majority of magma-poor margins display a number of common features, such as extreme crustal thinning from 30 km to a few kilometers thick over length scales of 100–200 km, normal faults in the upper crust, and a zone of serpentinized mantle exposed under an extremely thin crust (Reston, 2009). They also exhibit significant variability in their architecture, which has not yet been explained by the two faulting models previously discussed. This variability is particularly pronounced in the width and characteristics of the necking and hyperextended domains. When defined for magma-poor continental margins, the necking domain corresponds to a wedge-shaped structure in which the continental crust is abruptly thinned from 30 km to <10 km thick, whereas the hyperextended domain corresponds to a crustal wedge tapering to 0 km thickness, which may be succeeded oceanward by a zone of exhumed mantle (Péron-Pinvidic et al., 2013, 2017; Sutra et al., 2013; Tugend et al., 2015). Formation of the hyperextended domain is followed by the eventual exhumation of serpentinized mantle lithosphere, which itself constitutes the most distal exhumation domain. In Figure 4, three conjugate margins located between Europe and Canada are presented with their respective necking and hyperextended domains (for locations, see Fig. 2). From north to south, examples include the Labrador–western Greenland conjugate margin (Chian et al., 1995); the northern Newfoundland–Iberia margin (Funck et al., 2003; Zelt et al., 2003; Sutra et al., 2013), and the southern Newfoundland–Iberia margin (Van Avendonk et al., 2006; Lau et al., 2006; Dean et al., 2000; Sutra et al., 2013). The Labrador and the Iberian margins are usually identified as lower-plate margins, whereas the western Greenland and Newfoundland margins are described as upper-plate margins (Péron-Pinvidic and Manatschal, 2009; Reston, 2009; Sutra et al., 2013) and should be compared separately.

Comparison between Upper-Plate Margins

Both the western Greenland (Fig. 4A) and the southern Newfoundland margins (SCREECH 3 profile; Fig. 4C) present a sharp necking zone and a relatively wide hyperextended domain in which the lower crust seems to be missing beneath a thin remnant of stretched upper crust (Reston, 2009). While the architecture of both margins seems to be similar, they differ in scale, since the hyperextended domain is ∼115 km long in the southern Newfoundland margin and only 56 km long in the western Greenland margin. In contrast, the northern Newfoundland margin (SCREECH 1 profile; Fig. 4B) presents a different geometry, with a more gradual necking zone leading to a narrow hyperextended domain (31 km long). It appears therefore that if the characteristics of the western Greenland margin correspond to an archetypical upper-plate margin (narrow, sharp, and devoid of structure), the same is not the case for the wide southern Newfoundland margin or for the northern Newfoundland margin.

Comparison between Lower-Plate Margins

The necking domain is widely distributed in the northern Iberian margin (ISE 1 profile; Fig. 4B) and is narrowly distributed in the southern Iberian margin (IAM 9 profile; Fig. 4C), thus showing significant variations in width and total extension along strike. In contrast, the width of hyperextended domain is surprisingly constant along the entire margin (41 km and 49 km respectively; Sutra et al., 2013). In comparison, the Labrador margin (Fig. 4A) presents the opposite geometry, with a wide hyperextended domain (122 km long) and a narrowly distributed necking domain (∼68 km). It appears therefore that only the Labrador margin seems to correspond to the archetype of a lower-plate margin, with a narrow necking domain and a wide hyperextended domain.

These comparisons demonstrate that not only does a first-order asymmetry exist between conjugate sides of a magma-poor margin, but also there is significant variability in terms of the width of the necking and hyperextended domains and in the architecture of the “upper-plate” and “lower-plate” margins. Although sequential and detachment faulting models can explain the first-order asymmetry, they do not provide any explanation for the variability in the width of the “upper-plate” and “lower-plate” margins (Fig. 4). Numerous studies have demonstrated that margin architecture is largely the mechanical consequence of the variability in crustal thickness, lithospheric thermal structure, rheological properties of the crust and mantle, finite strain, and extension rates (England, 1983; Kusznir and Park, 1987; Bassi, 1991, 1995; Buck, 1991; Buck et al., 1999; Huismans et al., 2005; Lavier and Manatschal, 2006; Gueydan et al., 2008; Huismans and Beaumont, 2011; Brune et al., 2014; Svartman Dias et al., 2015). The presence of preexisting fabrics is also thought to be an additional control (Dunbar and Sawyer, 1989; Ring, 1994; Piqué and Laville, 1996; Corti et al., 2007; Clerc et al., 2015; Manatschal et al., 2015), but their effect on the development of rift structures is not very well constrained. Our numerical experiments tested whether or not preexisting fabrics can affect deformation processes, and if they do, whether sequential normal faulting or detachment faulting is favored. The numerical experiments can also explain the variability in “upper-plate” and “lower-plate” margin architecture in terms of the width of the necking and hyperextended domains.


Experiments were performed with an extended version of the numerical code PARAVOZ for elasto-visco-plastic material (EVP), called geoFLAC (Fast Lagrangian Analysis of Continua; Poliakov et al., 1993; Tan et al., 2012; Svartman Dias et al., 2015). At low temperatures, the material behaves elastically until it reaches its yield strength, as described by a Mohr-Coulomb failure criterion. Subsequently, the material flows plastically (Choi et al., 2013). When temperatures are high enough to activate dislocation creep, materials deform by Maxwell viscoelastic thermally activated creep, approximated as a nonlinear temperature- and strain rate–dependent flow (Choi et al., 2013; Svartman Dias et al., 2015). The mechanism of deformation that requires less energy or effective stress (square root of the second invariant of the stress tensor) is favored. As in our previous study (Jammes and Lavier, 2016), the explicit polymineralic rheological composition was generated by distributing the different mineral phases in each element among Lagrangian particles. In each element, one fraction of particles is assigned one mineral phase while the remnant fraction is assigned the other. The fraction of each mineral phase corresponds to the percentage of each mineral phase in the elements. The friction, cohesion, and viscosity in each element correspond to their geometric average, weighted by the ratio of the particles (Jammes et al., 2015; Jammes and Lavier, 2016). For each phase, the parameters assigned to the viscous flow law were obtained from laboratory measurements. We used “wet” quartz (Brace and Kohlstedt, 1980) and plagioclase (Shelton and Tullis, 1981) for the crust and “dry olivine” (Goetze and Poirier, 1978) and “orthopyroxene” (Raleigh et al., 1971) for the mantle (see Table S1 for the parameters1). In the literature (Fountain and Christensen, 1989), the composition of a granite or granodiorite is described as being composed on average of 20% to 40% weak minerals (quartz and biotite) and 80% to 60% strong minerals (plagioclase, feldspar, and amphibole). Following these conventions, we used in this study 20% quartz and 80% plagioclase for the crustal composition and 70% dry olivine and 30% orthopyroxene for the lithospheric mantle peridotite. In the crust, mineral phases are either randomly distributed amongst the particles of each element or statistically distributed using an algorithm initially developed to create a 2-D synthetic velocity field (Holliger and Levander, 1992; Holliger et al., 1993; Goff et al., 1994). In this study, mineral phases were distributed with a horizontal scale of 7 km and vertical scale of 1 km. This distribution was then rotated by 30° or 90° to be compatible with the orientation of the structures observed in the seismic reflection data (Fig. 2). In the lithospheric mantle, mineral phases were randomly distributed. Due to a lack of constraints, the asthenospheric mantle is considered as depleted and was modelled as only dry olivine.

In each model, the domain was 250 km in depth and 400 km in length. Velocity boundary conditions were imposed on both sides of the models (velocity = 0.5 cm yr–1, in extension), the top surface was free, while at the base of the model, a Winkler foundation was imposed to maintain isostatic equilibrium. The crust was initially 35 km thick. The temperature at the crust-mantle boundary was taken to be at 500 °C, increasing to 1330 °C at 100 km depth, and was set at 10 °C at the surface. Strain softening was introduced in the models to account for the formation of faults and occurred when the plastic strain was higher than 0.1, corresponding to a linear decrease of cohesion (from = 40 MPa to = 4 MPa) and friction angle (from φ = 30° to φ = 15°) with the plastic strain (Lavier et al., 2000; Lavier and Manatschal, 2006). All the parameters used are summarized in Table S1 (see footnote 1).

Parameters such as the percentage, the scale, and the orientation of heterogeneities should be tested to understand the overall effect of strength variation from inherited crustal fabrics and the relative importance of crustal versus mantle lithosphere fabrics. However, few constraints are available on mantle lithosphere fabrics. In this study, we focused on the effect of the orientation of kilometer-scale crustal fabrics on rifting processes; when possible, further studies will complete our analysis in contrast with the mantle lithosphere fabrics.


Three end-member models are presented in this paper and compared to a reference model with random strength distribution (Fig. 5A, model 1; Fig. S1 [footnote 1]). In all the other models, the phases were randomly distributed in the lithospheric mantle, but in the crust, they were either distributed to form a horizontal fabric (Fig. 5B, model 2; Fig. S2), a vertical fabric (Fig. 5C, model 3; Fig. S3), or a 30°-dipping fabric (Fig. 5D, model 4; Fig. S4). Each model is presented in the Supplemental File (Figs. S1–S4 [see footnote 1]), but the main results are summarized in Figure 5.

Model 1 with Random Distribution (Fig. 5A; Fig. S1)

In this model, the initial phase of deformation is characterized by distributed normal faulting in the upper crust and distributed anastomosing shear zones in the lower crust and lithospheric mantle (Fig. 5A, part i; Fig. S1B [footnote 1]). The plastic strain (Fig. S1B) and strain rate field (Fig. 5A; Fig. S1B) show that brittle shear zones cut through the brittle upper crust and sole out at midcrustal levels. After 50 km of extension, two types of faults can be identified in the upper crust: 60°–50°-dipping normal faults and low-angle normal faults that sole out in the middle crust and exhume crustal material (see distribution of plastic strain in Fig. S1B). In the lower part of the crust and in the lithospheric mantle, extension is accommodated along distributed ductile shear zones, forming an anastomosing pattern preserving undeformed crustal or mantle blocks (Fig. 5A, part i). This combination of brittle and ductile deformation processes results initially in a uniformly stretched lithosphere. After ∼100 km of extension, strain weakening promotes localization of deformation in one of the basins. This basin becomes the main rift basin, under which the lithospheric mantle is progressively thinned by the action of persistent ductile shear zones forming an anastomosing pattern (see strain rate field in Fig. S1B). As a result, the asthenospheric mantle upwells (Fig. 5A, part ii). After 200 km of extension, deformation remains mostly localized in and next to the main rift basin (Fig. S1C). Left and right of the main basin, high-angle brittle detachment structures alternate and exhume crustal material (see distribution of plastic strain in Fig. S1C). However, deformation processes in the crust and mantle remain decoupled, and there is no detachment structure crosscutting the entire crust and potentially exhuming mantle to the seafloor. While extension continues, we can see a basinward progression of crustal detachment systems that progressively thin the crust and elevate the mantle (Fig. S1D). Final exhumation occurs after 360 km of extension (Fig. 5A, part iii; Fig. S1E [footnote 1]) with the development of a final low-angle detachment fault. As a result, the basin is extremely asymmetric, with a gradually thinning, 95-km-long necking zone and a 185-km-long hyperextended domain on the left side (forming the lower plate), and a sharp 66-km-long necking zone with a very narrow hyperextended domain (34 km) on the right side (forming the upper plate).

Model 2 with Horizontal Crustal Fabrics (Fig. 5B; Fig. S2)

With a horizontal crustal fabric, the initial phase of deformation is distributed along the entire width of the model, but deformation is accommodated along a few normal faults and low-angle detachment faults extending to the Moho. These faults develop in the upper crust and sole into the anastomosing shear zones created in the upper lithospheric mantle (Fig. 5B, part i). They result in the formation of several kilometer-deep basins associated with crustal thinning. After 100 km of extension, deformation localizes in one of the basins, and a crustal-scale detachment fault exhumes the deeper part of the crust. As a result, the crust is quickly thinned, and the mantle is elevated (Fig. S2B [footnote 1]). After 150 km of extension, a conjugate detachment fault develops and delimits a crustal block in the center of the rifted domain (Fig. 5B, part ii; Fig. S2C). The strain rate shows that at this stage, normal faults are developing in this crustal block, presaging its future dismantlement. After 200 km of extension, a gradual thinning of the crustal block by normal faulting occurs on the left side of the margin (Fig. S2D). Breakup occurs after 260 km of extension, through the development of an antithetic detachment fault exhuming the mantle lithosphere to the seafloor (Fig. 5B, part iii; Fig. S2E). The resulting margin is asymmetric, with the upper plate on the left side and the lower plate on the right side. The upper plate consists of a sharp 71-km-long necking zone and an 87-km-long hyperextended domain, and the lower plate is made of a sharp 72-km-long necking zone and a wide hyperextended domain (106 km). Another remarkable characteristic of this model is the presence of an aborted rift basin in the lower plate, resulting in boudinage at the scale of the lithosphere.

Model 3 with Vertical Crustal Fabrics (Fig. 5C; Fig. S3)

With a vertical crustal fabric, the initial phase of deformation is characterized by the formation of horsts and grabens delimited by normal faults rooted in anastomosing shear zones that develop in the upper part of the mantle lithosphere. The deformation remains distributed for 100 km of extension, leading to the formation of a wide domain of stretched crust (Fig. 5C, part i; Fig. S3B [footnote 1]). After 150 km of extension, deformation begins to localize in the mantle lithosphere, with the formation of a low-angle detachment fault exhuming mantle at the base of the crust (Fig. 5C, part ii; Fig. S3C). This is accompanied by the breakup of the crust by high-angle normal faults (Fig. S3D). Final breakup occurs after 260 km of extension with the development of a synthetic detachment fault exhuming mantle to the seafloor (Fig. 5C, part iii; Fig. S3E). As a result, the lower plate (on the left side) is composed of a gradual necking zone, 148 km long, and a narrow hyperextended domain, 65 km long. On the other side, the upper plate is composed of a narrow, abrupt necking zone (31 km long) and a narrow hyperextended domain (27 km long). However, the stretched domain, resulting in lithospheric boudinage, affects the entire width of the model.

Model 4 with 30°-Dipping Crustal Fabrics (Fig. 5D; Fig. S4)

With a 30°-dipping crustal fabric, the initial phase of deformation is characterized by the development of lithospheric-scale low-angle detachment along the weak heterogeneities that extend into the upper mantle (Fig. 5D, part i; Fig. S4B [footnote 1]). Crustal deformation and mantle deformation are coupled in the early phase of deformation. After around 150 km of extension, deformation preferentially focuses on one of the detachment faults and exhumes the mantle in its footwall almost to the surface (Fig. 5D, part ii; Fig. S4C). The breakup of the hanging-wall block of the lithospheric detachment results in the juxtaposition of crustal allochthons on top of the newly exhumed mantle (Fig. S4C). After 200 km of extension, deformation migrates slightly to the right with the initiation of a new detachment system (Fig. S4D). This results in mantle and asthenospheric exhumation after 260 km of extension (Fig. 5D, part iii; Fig. S4E). In this model, the lower plate is then located to the right of the upper plate. The lower plate consists of a wide, 89-km-long necking zone and a wide, 120-km-long hyperextended domain. On the other side, the upper plate thins along a large gradual necking zone (139 km long) and a narrow hyperextended domain (35km long). On both sides, the necking zone appears to be gradual toward the continent and abrupt oceanward.

Distribution of Strain during the Stretching Phase

Comparison of the models after 50 km and 100 km of extension (Fig. 5; Figs. S1–S4 [footnote 1]) gives insights into the effects of strength variations inherited from crustal fabrics on the evolution of strain distribution during the stretching phase. In the case of randomly distributed heterogeneities, the stretching phase is highly distributed. Extension is accommodated by normal faults formed every 20–30 km that dip in both directions, resulting in the formation of horsts and grabens. Similar, but wider structures are observed in model 3 (vertical fabrics), where normal faults are formed every 40–50 km. In models 2 and 4 (with horizontal and dipping fabrics, respectively), faults are more spaced out and evolve rapidly into crustal- and lithospheric-scale detachments. As a result, core-complex structures are generated instead of horsts and grabens.

Therefore, the distribution of compositional heterogeneities affects the distribution of the deformation. In model 1, heterogeneities are randomly distributed in the entire model; as a result, deformation is widely distributed among numerous normal faults. In models 2, 3, and 4, heterogeneities are spaced out according to a certain scale, influencing the distribution of the deformation among fewer and spaced out faults.

The orientation of heterogeneities also affects the distribution and style of deformation during the stretching phase. Our models show that horizontal and dipping crustal fabrics result in the development of core complexes, whereas vertical fabrics or randomly distributed heterogeneities result in the formation of horsts and grabens. In the case of dipping fabrics, it is clear that detachment faults initiate along the dip of the heterogeneities. As a result, all the low-angle faults formed dip in directions consistent with crustal fabrics. To confirm this result, a model with fabrics dipping in the opposite direction was also tested (Fig. S5 [footnote 1]). As expected, we observed the formation of low-angle detachment faults consistent with the orientations of the heterogeneities. In the case of horizontal fabrics, faults initiate along heterogeneities, but the absence of orientation allows for the development of low-angle faults dipping in both directions.

Finally, we find that compositional heterogeneities must have a high strength contrast with the surrounding rock in order to affect deformation processes during the stretching phase. In the case of a random distribution of heterogeneities (model 1), the location of the heterogeneity first nucleating deformation is difficult to identify. However, in models 2, 3, and 4, only middle- and lower-crustal heterogeneities control the distribution of deformation. In the upper part of the crust, both quartz and plagioclase are in the brittle regime (Jammes et al., 2015, Jammes and Lavier, 2016); for the case of equal cohesion and frictional resistance, there is no strength contrast between compositional heterogeneities and surrounding rock. As a result, upper-crustal heterogeneities do not control the localization process (models 2 and 4). However, below 10 km, quartz becomes ductile while plagioclase remains brittle. The middle and lower crust are therefore in the semibrittle deformation field, and a depth-dependent strength contrast arises between the weaker quartz-dominated heterogeneities and the stronger plagioclase-dominated rock. This strength contrast can explain why localization tends to initiate in the middle and lower crust. We believe that these models show that deep crustal fabrics are more likely to control deformation than shallow crustal fabrics. However, if shallow crustal fabrics also impart a decrease in cohesion and friction (not accounted for in our modeling), the strength contrast between the fabrics and surrounding rock could be great enough to influence the distribution of deformation. In summary, in the models, micromechanisms of deformation, which are here taken into account as parameters in the Mohr-Coulomb yield criterion (cohesion and static friction angle) and dislocation creep laws (viscosity), determine both brittle and ductile behavior of the mineral phases present in the model. It is clear, however, that the mesoscale (here, kilometer-scale) distribution of the mineral phases used to create the compositional heterogeneities plays a dominant role in determining the locus and distribution of the deformation during the stretching phase of rifting processes, and also during thinning processes (see below).

Mechanism of Thinning and Exhumation

The models demonstrate that rifts are likely to acquire their architectural uniqueness from strength variations inherited from crustal fabrics during the distributed stretching and localized thinning phases of deformation. The exhumation phase does not seem to be affected by inherited structures and appears to be similar in all models. The late phase of extension is always controlled by a detachment fault (synthetic or antithetic to previous structures) that exhumes mantle to the seafloor, leading to breakup. In their review study, Manatschal et al. (2015) suggested that inherited structures do not significantly control the location of breakup. Our numerical study confirms this observation and reaffirms that inherited structures have no control on the structures leading to exhumation. However, crustal heterogeneities strongly control thinning processes and play an important role in the coupling of deformation between the crust and the mantle lithosphere.

It appears that neither the sequential or detachment faulting models described earlier in this paper dominate the thinning process in our models. Localized thinning affecting our models appears to occur via a combination of processes from both models, or via a variation of either model, or even an alternative model. In model 1 (Fig. 5A; Fig. S1 [footnote 1]), anastomosing shear zones in the middle/lower crust and upper mantle initially control the phase of crustal thinning until low-angle crustal detachments thin the remaining crust and exhume the mantle lithosphere, as described in the detachment faulting model (Lavier and Manatschal, 2006; Péron-Pinvidic and Manatschal, 2009). However, in contrast to this model, deformation remains decoupled, and several sequential detachments migrating toward the axis of the rift have to develop to progressively thin the crust and finally couple crustal and mantle deformation. This mechanism has been previously described by Svartman Dias et al. (2015), Brune et al. (2014), and Jammes and Lavier (2016) and results in an extremely wide and asymmetric margin. In the case of model 2 (Fig. 5B; Fig. S2), we observe the formation of two conjugate low-angle detachment faults decoupled along a midcrustal décollement bordering an isolated upper-crustal block, as described in the detachment faulting model. However, as extension continues, the crustal block is gradually thinned by sequential normal faulting, before the development of an antithetic detachment fault exhuming upper mantle to the seafloor. The thinning process appears therefore to occur by a combination of the sequential and detachment fault concepts. In the case of model 4 (Fig. 5D; Fig. S4), the dipping heterogeneities favor coupling between crustal and lithospheric mantle deformation. As a result, a low-angle detachment fault cutting the entire crust initiates in the early phase of deformation and rapidly exhumes the mantle beneath the hyperextended crust. If this thinning process presents similarities with the detachment faulting model, it differs from that model by the absence of a symmetric stretching phase and the early coupling of the deformation. Finally, given vertical crustal fabrics (model 3; Fig. 5C; Fig. S3), crustal thinning processes differ completely from the published models. Indeed, thinning of the crust is due to the formation of high-angle faults in the upper crust and a low-angle detachment fault in the upper mantle, exhuming mantle lithosphere at the base of the crust. The crustal deformation is therefore completely decoupled from the mantle deformation during thinning processes. This mechanism, never before described in literature, could explain geological observations of mantle exhumed beneath hyperextended crust (e.g., Espíritu Santo Basin; Zalán et al., 2011).

Resulting Geometry of the Margin

As observed in nature (Fig. 4), our models demonstrate that, depending on the orientation of crustal fabrics, the final structure of the margin can vary substantially. The widths of the necking and hyperextended domains vary considerably among the models (Figs. S1–S4 [footnote 1]). However, some common trends can be observed. In three models with inherited fabrics (model 2, 3, and 4), the presence of aborted rift basins results in lithospheric-scale boudinage. These basins are formed in the early phase of stretching and present different characteristics depending on the inherited fabrics. Core-complex structures are formed with horizontal and dipping fabrics, whereas horst-and-graben basins are formed with vertical fabrics.

Upper-plate and lower-plate margins can be identified following the orientation of the last detachment fault, leading to mantle lithosphere exhumation. We observe that the lower-plate margin corresponds to the left-side margin in models 1, 3, and 4 and the right-side margin in model 2 (Fig. 5). In all our models, we show that the hyperextended domain is wider in lower-plate margins than in upper-plate margins. Boudinage of the lithosphere renders the width of the necking domain difficult to estimate precisely; however, we can observe that, in general, the necking domain is narrower and sharper in upper-plate margins than in lower-plate margins (Fig. 5). These results are in agreement with the definition of “upper-plate” and “lower-plate” margins (Reston, 2009; Péron-Pinvidic et al., 2017), and the observed structures of the Labrador–western Greenland (Fig. 4A) and northern Newfoundland–Iberia margins (Fig. 4B). Uncertainty concerning the positions of the upper- and lower-plate margins remains, however, for the southern Newfoundland–Iberia conjugate margin (Fig. 4C). Described as an upper-plate margin in the literature (Péron-Pinvidic and Manatschal, 2009; Reston, 2009; Sutra et al., 2013), the southern Newfoundland margin presents a wider hyperextended domain than the upper-plate margin model (115 km vs. 49 km; Fig. 4). This inconsistency could be explained by the deformation processes modeled in model 2. In this model, the detachment fault that forms during the exhumation phase is antithetic to the main structures controlling the thinning phase. Determination of the upper-plate versus lower-plate margin is consequently possible only during the latest phase of deformation and does not involve the structures formed during the thinning phase. The late development of out-of-sequence detachments, as described by Gillard et al. (2015, 2016), could therefore explain why distinguishing between upper- and lower-plate margins for the southern Newfoundland–Iberia conjugate margin is not straightforward, as described by Reston (2009) and Péron-Pinvidic et al. (2017).

The numerical experiments presented in this paper demonstrate that crustal fabrics resulting from compositional variations strongly influence the mechanism of deformation, not only in the early phase of rifting, but also during the thinning process. However, crustal fabrics do not significantly control the locations of crustal breakup and mantle exhumation, which appear to be always associated with the development of a detachment fault exhuming mantle to the seafloor. In addition, we observe that the structures developed in our models share many features with those observed in natural magma-poor rifts.

Our models show the following features: (1) Inherited crustal fabrics can explain the distribution of deformation and deformation processes observed during the stretching phase, where vertical fabrics favor the formation of horst-and-graben structures, whereas horizontal and dipping fabrics favor the formation of core complexes. (2) Depending on the crustal fabrics, thinning processes differ and entail mechanisms involving (i) a combination of the detachment faulting and sequential normal faulting models, (ii) a modified version of the detachment faulting model (with early or late coupling between the crust and the mantle), or (iii) an alternative model in which crustal and mantle processes remain decoupled and lead to the formation of high-angle faults in the upper crust and a low-angle detachment fault in the upper mantle. (3) Strength variations inherited from crustal fabrics control the thinning processes and consequently the resulting geometry of the margin. (4) Finally, this study demonstrates that inherited structures do not significantly control the location of the breakup and mechanism of the exhumation phase: The late development of out-of-sequence detachments in some models explains why distinguishing between upper-plate and lower-plate margins is not a straightforward process.

In summary, our study demonstrates that while models do account for micromechanical constitutive behaviors, the distribution of compositional heterogeneities at the outcrop scale inherited from previous phases of tectonic deformation determines the behavior of the bulk lithosphere in rifting settings. Thus, we propose that when determining the rheological behavior of the lithosphere, more attention should be given to the distribution of mineral phases at the mesoscale than has been given so far in the field of rock rheology.

We thank the reviewer D.L. Harry and the Science Editor Raymond M. Russo for their constructive and positive comments that significantly improved the paper. We also thank Harm van Avendonk for helpful comments and fruitful discussions. This study was supported by a National Science Foundation–GeoPRISMS grant for the project “Effect of contrasting structural and compositional inheritances on the development of rifting margins.” Computations were made using the program FLAC, and data can be provided upon request by sending an e-mail to

1Supplemental File. Presents parameters and evolution of models discussed in the paper. For each model, plastic strain (brittle deformation), material, and strain rate are plotted for different phases of deformation. Please visit or access the full-text article on to view the Supplemental File.
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.