Experimental modeling of rifting at craton margins

The deformation pattern resulting from continental rifting often exhibits a large degree of variability in terms of characteristics such as width, symmetry, subsidence, and architecture of faulting that may also vary in short distances along strike of a single rift (e.g., Ziegler and Cloetingh, 2004). Among other controlling parameters, the variability in deformation pattern likely results from the inherited thermal and mechanical structure of the continental lithosphere at the time of rift inception. Previous work has highlighted the localization of continental rifts along weak areas surrounding strong cratonic keels (e.g., Dunbar and Sawyer, 1988; Versfelt and Rosendahl, 1989; Tommasi and Vauchez, 2001; Ziegler and Cloetingh, 2004), where the juxtaposition between an old, cold, and resistant lithosphere and a weaker domain is likely to influence the way rifts develop and propagate, and at a more detailed scale, their architecture and symmetry. The three following examples exemplify these observations.

1. In the Baikal rift system (Fig. 1A), deformation is localized at the southeastern margin of the strong Siberian craton and is characterized by significant along-strike variations in width, symmetry, and subsidence (e.g., Petit and Déverchère, 2006). In the Lake Baikal region, deformation is strongly localized at the craton border and characterized by a prominent asymmetry with the master fault on the craton side; conversely, to the northeast, a more diffuse deformation affects the weaker Sayan-Baikal belt, away from the craton border. Both numerical (Petit et al., 2008) and analog (Chemenda et al., 2002; Corti et al., 2011) models have suggested that the lateral variation in rheology of the lithosphere between the strong craton and the weaker belt together with the presence or absence of a weak suture in between may have strongly controlled the rift characteristics.

2. The West Antarctic Rift System (Fig. 1B), one of the largest continental extension regions on Earth, developed within a heterogeneous weak lithosphere composed of the microplates of West Antarctica at the margin of the East Antarctic craton (e.g., Behrendt et al., 1991). Most of extension in the area was attained during the Cretaceous rifting (e.g., Lawver and Gahagan, 1994), which resulted in a diffuse deformation within the weak lithosphere and localized faulting and strong flank uplift at the margin of the East Antarctic craton, resulting in the uplift of the Transantarctic Mountains, the largest noncontractional mountains in the world (e.g., Busetti et al., 1999). Both analog (Bonini et al., 2007) and numerical (e.g., Huerta and Harry, 2007) models indicate that the heterogeneous pre-rift rheological structure of the lithosphere affected the extension process and the main rift characteristics to a very great extent.

3. The East African Rift south of the Ethiopian Rift Valley bifurcates in two branches that surround the strong Tanzanian craton. In the western branch, rifting is typically localized within preexisting weak domains at the craton margin (e.g., Corti et al., 2007; Morley, 2010), giving rise to elongated, narrow, deep basins (such as that occupied by Lake Tanganyika), the characteristics of which (e.g., subsidence, symmetry) are controlled by the lateral variation in strength at the craton border (e.g., Petit and Ebinger, 2000). Conversely, the eastern branch is characterized by an along-axis transition from a localized deformation at the craton margin in the Kenya rift to a wide deformation zone in the so-called Tanzania divergence. In this region, widening of the rift is associated with the impingement of the southward-propagating rift on a strong lithospheric domain (Masai block) east of the Tanzanian craton (e.g., Nyblade and Brazier, 2002; Ring et al., 2005; LeGall et al., 2008).

These examples suggest that horizontal variations in the rheology of the lithosphere, and not only its vertical layering (e.g., Buck, 1991; Brun, 1999), likely play a major role in controlling the distribution and architecture of the extensional deformation. This influence has been tested in a limited number of previous analogue models (Chemenda et al., 2002; Bonini et al., 2007; Corti et al., 2011); however, these works focused on the analysis of specific natural rift settings (Baikal rift, West Antarctic Rift System), and thus reproduced boundary conditions (e.g., plan-view geometry, thickness, and rheology on crustal and/or mantle layers) specific to the studied natural prototypes. In this paper, we expand and generalize these previous results by presenting analog models of rifting at craton margins. The models are aimed at the systematic analysis of the influence of lateral variations in the pre-rift rheological structure (i.e., differences in rheological layering and strength between a strong craton and an adjacent weaker lithosphere, and presence and characteristics of an intervening weak zone) on different characteristics of continental rifting, such as the width of the deformed zone, basin symmetry, subsidence, and the patterns of lithospheric thinning. The modeling results offer important insights into the development of continental rifts at the margins of cratonic areas and involve general implications for some important aspects of continental rifting, such as the variability in the distribution of deformation and the degree of asymmetry often exhibited by continental rifts and passive margins worldwide (e.g., Ziegler and Cloetingh, 2004).

The experiments reproduced continental rifting in an artificial gravity field of ∼18 g by using the large-capacity centrifuge at the Tectonic Modeling Laboratory of the Institute of Geosciences and Earth Resources (National Research Council of Italy) at the Department of Earth Sciences, University of Florence. The experiments simulated extension of the brittle-ductile transition upper part of the continental lithosphere (crust plus upper lithospheric mantle) floating above a low-viscosity material representing the asthenosphere (Fig. 2). The models were built inside a rectangular Plexiglas box (with internal dimensions of 25 × 16 × 7 cm) and confined by two moveable side walls; removal of rectangular blocks (spacers) at the sides of these moving walls allowed vertical thinning and lateral extension of the models in response to the centrifugal forces to fill the empty space (Fig. 2). Sequential removal of spacers during successive runs in the centrifuge allowed controlling the amount and rate of extension. Top-view photos and laser scans of the models were taken after the end of each centrifuge run. After a successful experiment, the models were frozen before taking a number of cross sections to study their internal geometry.

The brittle-ductile model lithosphere was characterized by a lateral variation in rheological layering between a high-strength craton and a weaker mobile belt. Most of the models involved the presence of a weak zone between the two lithospheres (e.g., a weak intervening suture; Fig. 2), although different rheological configurations were tested (see following discussion of Experimental Series). The application of a homogeneous stress field (imposed by the centrifugal body forces) to this laterally varying rheology allowed exploration of the response of this heterogeneous continental lithosphere to extension, thus providing insights into the role of the variation in vertical rheological layering on the evolution and architecture of continental rifting.

A vertical sequence of brittle and ductile materials was used to reproduce the rheological multilayering characteristic of the craton and mobile belt lithospheres (Fig. 2; e.g., Corti et al., 2011). The materials used to reproduce the crustal and mantle layers were the same in both lithospheres, although the thicknesses of these different layers were varied in order to reproduce variations in lithospheric strength; in a limited number of models, weaker materials were used within the mobile belt (see following). The rheology of these lithospheres (e.g., resistance and thickness) was simulated by reference to the vertical rheological layering and strength profiles illustrated in previous works (e.g., see numerical models of Petit et al., 2008).

The brittle upper crust was simulated by using a K-feldspar powder characterized by a linear increase in strength with depth, reproducing the natural brittle behavior (Fig. 2G). The lower crust was modeled with a ductile mixture of silicone (Wacker Silicone Bouncing Putty 29 distributed by CRC France, referred to herein as Wacker BP29) and corundum sand (100:20 by weight). In some models, a mixture of Wacker BP29, corundum sand, and oleic acid (100:20:10 by weight) was used to simulate a weaker lower crust. The strong uppermost lithospheric mantle was simulated with a mixture (100:20% in weight) of plasticine (Pongo Fantasia modeling dough, distributed by FILA) and PDMS (polydimethylsiloxane; silicone SGM36 distributed by Dow Corning). In some models, a weaker version of the uppermost lithospheric mantle was reproduced with a mixture of plasticine (Pongo modeling dough) and PDMS (100:45% in weight). The lower lithospheric mantle was made of a Wacker BP29–corundum sand mixture (100:80% in weight): a mixture of Wacker BP29, corundum sand, and oleic acid (100:80:10% in weight) simulated a weaker lower lithospheric mantle in some models. A weak zone between the craton and the mobile belt was modeled with a Wacker BP29–corundum sand–oleic acid mixture (100:80:15% in weight). The weak zone was placed in the lower crust and in the mantle, whereas in nature the weak zone likely extends through the entire lithosphere; however, this simplified set-up represents a reasonable approximation of the natural process being investigated (see Corti et al., 2011). In some models, lower crustal and mantle layers within the mobile belt were made of weaker silicone-sand mixtures (see Table 1).

The crustal-mantle layers were on a low-viscosity mixture made of a Wacker BP29–corundum sand–oleic acid mixture (100:100:20 by weight), representing the asthenosphere that offers isostatic support to the deforming lithosphere. Details of the viscous materials were summarized in previous work (e.g., Agostini et al., 2009; Corti et al., 2011) and their flow curves are illustrated in Figure 2F; their use allowed obtaining an increase in density with depth and a variable strength reproducing a typical “Christmas tree” strength profile of the continental lithosphere (Fig. 2G).

The models were built with a geometric scale ratio of ∼2.5 × 10⁻⁷, such that 1 cm in the experiments corresponded to ∼40 km in nature (Table 2); this allowed modeling ∼40 km of total extension of ∼80-km-thick continental lithospheres. Dynamic and kinematic similarity of gravitational, viscous, and frictional stresses acting on the system (Ramberg, 1981) ensured that the velocity of extension in the models (∼2.5 × 10⁻⁵ m s⁻¹) scaled to natural values of ∼8 mm yr⁻¹. The scaled total resistance of the craton lithosphere was ∼4 10¹³ Nm⁻¹; the strength of the mobile belt was normally ∼40% that of the craton, and dropped to ∼10% in the weak zone.

Two different experimental series were designed to investigate the influence on rifting of lateral variation in rheology between a strong craton and a mobile belt (Fig. 3; Table 1). Series 1 experiments investigated the influence of the strength contrast between the two lithospheric domains (Fig. 3). This was achieved by changing the strength and/or thickness of ductile materials in the mobile belt domain in different models for a fixed craton lithosphere rheology. This experimental series was divided into subseries 1a and 1b, which differed in the thickness of the upper and lower crust in the two domains (Fig. 3). In subseries 1a, the thickness of the upper and lower crust was constant in the craton and the belt, and the strength contrast was varied by changing the strength of ductile materials in the belt. In subseries 1b, the lower crust was always thicker in the mobile belt: the strength contrast in these models was varied by either changing the thickness of the ductile layers (lower crust or uppermost mantle) and/or by using weaker materials in the belt.

Series 2 experiments considered the presence of a weak zone between the two lithospheric domains. Comparison of these experiments (with intervening weakness) with models of series 1 (with no intervening weakness) allowed us to evaluate the influence of the presence of a preexisting weak zone at the craton-belt boundary on the distribution of the extensional deformation. This series was divided into subseries 2a and 2b. Experiments of subseries 2a investigated the role, on rift architecture, of the variable depth of the brittle-ductile crust transition (i.e., variations in the thickness ratio of the upper and lower crust) in the mobile belt, for a constant craton rheology and a flat Moho at both sides of the weak zone (Fig. 3). Another parameter investigated (although less extensively) in this experimental series was the variation in the width of the weak zone (subseries 2b; Fig. 3).

In each experimental series or subseries, apart from the parameter under investigation, the other boundary conditions were kept constant. The experiments were analyzed in terms of the main characteristics of deformation (e.g., symmetry, rift width, depocenter location, fault architecture, subsidence); each experiment was repeated at least twice and, although the models may have differed in small details, the first order characteristics of deformation (e.g., fault architecture and evolution) were always comparable. In total, 27 experiments were performed.

The set-up of series 1 models considered the two lithospheric domains with no intervening weak zone. In this series the lithospheric strength of the belt domain was changed in different models in order to investigate the role of a variable strength contrast at the craton border on the style of the extension-related deformation (see discussion, Experimental Series).

All the models of this experimental series displayed a typical deformation pattern characterized by diffuse faulting within the mobile belt; no significant deformation affected the craton lithosphere in the models (Figs. 4–7; Animation 1). Deformation within the mobile belt was characterized by a large number of normal faults with small vertical throw; this pattern was associated with a limited subsidence of the rift and limited thinning of the extending lithosphere at depth.

The results of both subseries 1a (thickness of the upper and lower crust constant in the craton and the belt) and 1b (thicker lower crust and thinner upper crust in the mobile belt) indicate that the final deformation pattern was strongly controlled by the strength contrast at the craton-belt boundary. When the strength contrast was relatively high (i.e., the mobile belt was characterized by relative low resistance), deformation localized at the craton margin, leading to major normal faulting at the boundary between the two lithospheric domains (see models RCM17 and RCM26; Figs. 4–6). When the strength contrast was maximized by reduction of the thickness of the brittle crust and use of weak materials in the mobile belt (model RCM17, Figs. 4 and 5; Animation 1), a single major fault system formed on the craton side and accommodated the majority of extension; diffuse deformation, with minor normal faults, affected the mobile belt. The resulting deformation pattern was asymmetric, with maximum subsidence of the rift depression and maximum lithospheric thinning at depth located at the craton margin (Fig. 5). The major normal fault system formed in the early stages of extension. For increasing extension, deformation progressively propagated inside the mobile belt, affecting a region that widened with time and extension (Figs. 4 and 7D). When the strength contrast was low (i.e., the mobile belt was characterized by relatively high resistance) deformation was different. No major normal faults formed at the craton-belt boundary, and extension was accommodated by a series of normal faults with comparable, limited vertical throw, diffuse over a large region within the strong mobile belt (see models RCM02 and RCM03 in Fig. 5). Extension gave rise to a wide deformed zone after the early stages of extension; there was no appreciable lateral migration of faulting in these low-strength contrast models, and the width of the area affected by faults was constant throughout the experiments (Fig. 7D). In any case, the width of deformed zone was generally higher than in models with high strength contrast, and subsidence and lithospheric thinning were less pronounced than in the corresponding high-strength contrast model RCM17 (Fig. 5).

In general, comparison of the different models of both subseries indicates that a decrease in the resistance of the mobile belt (i.e., an increase in the craton and belt strength contrast) led to (1) an increase in rift asymmetry (i.e., dominance of a major fault system on the craton side); (2) an overall decrease in the width of deformed zone; (3) an increase in rift subsidence and lithospheric thinning; and (4) an increase in the number of extension-related faults (Supplemental Fig. 1¹). In summary, an increase in strength contrast induces a more localized, asymmetric deformation with large boundary faults on the craton side accommodating a prominent basin subsidence and lithospheric thinning.

The typical evolution of series 2 models is exemplified by deformation of models RCM01 and RCM04 (illustrated in Fig. 8; Animations 2 and 3). These experiments displayed a marked difference with respect to the previous series 1 models in terms of distribution and characteristics of deformation. Whereas in the previous models deformation diffusely affected the mobile belt, in series 2 models extension was localized in correspondence to the weak zone, and did not migrate in the craton or in the belt during the duration of the experiment (Fig. 8; Animations 2 and 3). Extension gave rise to conjugate normal faults, which accommodated subsidence of a rift valley localized above the weak zone. Deformation remained localized within the weak zone throughout the experiment: progressive extension led to continued slip on major boundary faults, and to widening and deepening of the rift depression, with no detectable deformation within the strong craton or the weaker belt (Fig. 7D; Animations 2 and 3). At depth, the localized upper crustal faulting corresponded to thinning of the ductile narrow weak zone, with focused asthenospheric upwelling beneath the rift depression. From a mechanical perspective, this deformation pattern corresponded to a necking instability in the extending lithosphere.

The experiments of subseries 2a investigated the role of the variation in depth of the brittle-ductile crust transition on rift architecture (i.e., variations in the thickness ratio of the upper and lower crust) in the mobile belt, for a constant craton rheology and a flat Moho at both sides of the intervening weakness. The step in the brittle-ductile transition passing from the craton to the mobile belt was progressively varied in different experiments, from 0 mm (no brittle and ductile crust thickness variation at both sides of the weakness) to 6 mm (thickness of the brittle and ductile crust of 8 and 2 mm and 2 and 8 mm in the craton and the mobile belt, respectively; see Fig. 9; Table 1). With no brittle and ductile crust thickness variation at the sides of the weakness (model RCM04), the localized deformation gave rise to a rather symmetric narrow rift valley, with extension taken up by major boundary faults at the margins of the weak zone with comparable slip (Figs. 8, 9, and 10). The major fault on the craton side was characterized by a mean vertical throw of ∼3.2 mm (55% of total slip on both fault systems), whereas the opposite major fault system accommodated a vertical displacement of ∼2.8 mm. However, model RCM01 (with a 3 mm step in depth of the brittle-ductile crust transition at the sides of the weakness) was characterized by an asymmetric basin, with a major boundary fault on the craton side and a system of minor normal faults on the belt side. In this case, the major normal fault system on the craton side was characterized by a mean vertical throw of ∼3.9 mm (67% of total slip on both fault systems), whereas the faults on the belt side accommodated only a vertical displacement of ∼1.9 mm. This basin asymmetry also corresponded at depth to a less prominent lithospheric thinning and upwelling of the asthenosphere (Fig. 8). The rift asymmetry was further enhanced by increasing the brittle-ductile transition step (Fig. 9): the amount of slip of the fault system on the craton side increased to 74% and 87% of total slip on both fault systems for models RCM07 (brittle-ductile transition step of 5 mm) and RCM25 (brittle-ductile transition step of 6 mm), respectively (Figs. 9 and 10). The depression of model RCM25 was strongly asymmetric, with a gentle monocline and only minor faults on the belt side. The transition from a fairly symmetric rift valley (brittle-ductile step of 0 mm) to a largely asymmetric extension (brittle-ductile transition step of 5–6 mm) also corresponded to a widening of the deformed region, from the ∼20 mm of model RCM04 to the ∼40 mm of model RCM25 (Figs. 9 and 10).

In summary, as the step in the brittle-ductile transition from the craton to the mobile belt increased, the deformation became wider, was accommodated by a larger number of faults, and became more asymmetric (Supplemental Fig. 2²); the rift depression progressively attained the shape of a half-graben with major boundary faults at the craton side. Basin subsidence initially increased, passing from a roughly symmetric to an asymmetric extension. However, when the basin became largely asymmetric (with lack of faults and presence of a single monocline on the belt side), subsidence decreased (Supplemental Fig. 2 [see ^{footnote 2}]).

Variations of the width of the weak zone (subseries 2b) directly controlled the width of the deformation zone, in particular, increasing the volume of the weak material resulted in a wider rift characterized by a less prominent subsidence of the rift floor and more limited lithospheric thinning at depth (Fig. 11; Supplemental Fig. 3³).

Studies of continental rift zones worldwide indicate that these structures often display significant variability in terms of characteristics such as width, symmetry, subsidence, and lithospheric thinning (e.g., Ziegler and Cloetingh, 2004). All these characteristics may vary not only in different rifts, but may also display significant changes within a single rift; for example, the cross-sectional structure of rift basins may vary from fairly symmetric to asymmetric and, in the asymmetric case, basin polarity may alternate along strike (e.g., Rosendahl, 1987). Because continental extension affects a prestructured, highly anisotropic lithosphere, horizontal variations in the rheology of the lithosphere have a strong influence on the rifting process, both at local and regional scales (e.g., Versfelt and Rosendahl, 1989; Tommasi and Vauchez, 2001). In particular, rifts normally develop within preexisting weak zones and tend to avoid stronger regions such as old cratons (e.g., Dunbar and Sawyer, 1988; Versfelt and Rosendahl, 1989; Tommasi and Vauchez, 2001; Ziegler and Cloetingh, 2004). When rifts develop at the margins of strong cratonic regions, the juxtaposition between an old, cold, and resistant lithosphere and a weaker domain likely has a strong effect on the distribution and architecture (e.g., symmetry) of the extensional deformation. This observation is supported by previous lithospheric-scale analog modeling works reproducing significant lateral variations in lithospheric strength (e.g., Chemenda et al., 2002; Bonini et al., 2007; Corti et al., 2011) as well as by our experiments, which allow us to generalize the role of the strength contrast between a strong craton and a weaker mobile belt and the influence of an intervening weakness (such as a suture zone) on the architecture, evolution, and distribution of deformation during continental extension.

Although other parameters (e.g., presence and importance of magmatic processes, variable rift orientation and kinematics, thermal state of the lithosphere) may contribute to the observed variability in the distribution of deformation during rifting at craton margins, our experiments suggest that the presence or absence of weak zones between the craton and the belt is an important parameter in controlling these rift characteristics (e.g., Corti et al., 2011; Corti, 2012). These weak zones may correspond to ancient major lithospheric sutures (as suggested for the Baikal rift; Petit and Déverchère, 2006) or to thermally and/or rheologically anomalous regions influenced by processes such as small-scale convection in correspondence to major cratonic areas or channeling and ponding of plume-related deep hot material in correspondence to preexisting lithospheric heterogeneities, as suggested for the East African Rift (e.g., Ebinger and Sleep, 1998; King and Ritsema, 2000).

Regardless of the strength contrast between the craton and the belt, the presence of a weak zone interposed between the strong craton and the mobile belt strongly localizes deformation. The weak zone determines a strong local decrease in lithospheric strength, allowing a necking of the extending continental lithosphere (Fig. 12; Mulugeta and Ghebreab, 2001). Deformation does not migrate outside the weak region throughout the extension process, resulting in a narrow, deep rift valley where a few large boundary faults accommodate basin subsidence, and the lithosphere is significantly thinned at depth (Figs. 7 and 12). The effect is maximized when both the craton and the mobile belt are characterized by a four-layer lithosphere. In this case, the interruption of the lateral continuity of the strong lithospheric mantle by the weakness results in a significant localized thinning of the extending lithosphere (e.g., Callot et al., 2001, 2002; Chemenda et al., 2002; Corti et al., 2011), in a fashion similar to the rupture of the brittle lithospheric mantle in correspondence to a lateral velocity discontinuity observed in previous normal-gravity modeling works (e.g., Brun and Beslier, 1996; Michon and Merle, 2003). In these conditions, the width of the rift valley is directly related to the width of the weak zone (plus the thickness of the brittle layer; Allemand and Brun, 1991), whereas subsidence (and lithospheric thinning at depth) is inversely related to this parameter, as observed in previous models (e.g., Corti et al., 2007; Corti, 2012).

In the absence of an intervening weak zone, the extensional deformation is more distributed and the lithospheric thinning more homogeneous (Figs. 7 and 12). From a mechanical perspective, the lithosphere is affected by distributed instabilities, expressed at surface by numerous normal faults and basins extending over wider regions, and the necking instability in correspondence to a single, well-developed rift valley is no longer present. In these conditions the craton and belt strength contrast controls the width of the rift and its evolution in time. When the contrast in strength between the belt and the craton is minimal, no major faults with significant vertical throw are able to form at the boundary between the two lithospheres, and deformation affects a wide region inside the strong mobile belt after the initial stages of extension. Single basins are generally characterized by minor subsidence and the deformation is, on a large scale, symmetric. Conversely, for high strength contrasts, deformation is localized at the craton-belt boundary at the beginning of extension, with development of a major fault system on the craton side; faulting is then able to propagate inside the weak belt, widening with time and progressive extension. This latter observation supports previous inferences from both analog (e.g., Benes and Davy, 1996; Brun, 1999) and numerical (e.g., Buck, 1991) modeling results. In both cases (low and strong strength contrast) deformation is accommodated by a higher number of normal faults, with less prominent vertical displacement and limited lithospheric thinning at depth (e.g., Benes and Davy, 1996). However, significant faulting at the craton border associated with a more pronounced, asymmetric subsidence is to be expected for a very high strength contrast related to the presence of a very weak belt (see following).

Overall, the entire data set of models highlights that the width of the deformed zone is negatively correlated with basin subsidence, and positively correlated with the number of faults (Supplemental Fig. 4⁴). In other words, if the rift width increases, the subsidence (and the lithospheric thinning at depth) decreases, and the number of normal faults increases. This behavior is related to the amount of vertical versus horizontal component of displacement accommodated by the extensional structures (Supplemental Fig. 4 [see ^{footnote 4}]). In wide rift zones, the imposed horizontal displacement is taken up by numerous faults, which can only accommodate a limited vertical displacement, and thus the result is limited subsidence. Conversely, in narrow rift zones the same horizontal displacement is taken up by a few normal faults, which can accommodate large vertical throws, and the result is greater subsidence.

The relations between the width of the deformed zone and basin subsidence and/or lithospheric thinning have obvious implications for the rifting process, influencing fundamental aspects such as its duration and volcanicity. In general, for a constant extension rate, the strong localized deformation in narrow regions is expected to lead to a rapid lithospheric necking and break-up; this process can be facilitated by the expected strong decompression melting and magma uprising throughout the lithosphere that may eventually lead to magma-assisted rifting of the strongly intruded and weakened plate (e.g., Buck, 2004). Note that in these conditions, as a function of the varying proportions between mechanical extension (plate stretching and faulting) and magma intrusion, both crustal thinning and subsidence of the surface may be significantly reduced because of the voluminous injection of magma into the lithosphere, as suggested for the northernmost sector of the East African Rift (Afar, Ethiopia; Keir et al., 2012) or the Baikal rift (Thybo and Nielsen, 2009). Conversely, the more uniform deformation and thinning of the lithosphere in wide zones are expected to inhibit lithospheric necking and major rising of the asthenosphere, thus preventing a rapid progression to the final plate separation and limiting the potential for decompression melting, even in areas where large magnitudes of extension are reached (e.g., as suggested for the West Antarctic Rift System; see following discussion). In very weak mobile belts, surface volcanism may be further inhibited by the dominance of ductile deformation over faulting, a process that reduces the likelihood of magmas migrating upward and reaching upper crustal levels by hindering the development of structures able to connect the surface to potential magma source areas, such as the viscous lower crust and lithospheric mantle.

Our models suggest that lateral variations in the pre-rift rheology are able to effectively control the symmetry of rift basins, supporting the results of previous modeling work (e.g., Corti, 2012). The models suggest that the most convenient rheological configuration leading to a strongly asymmetric basin involves the presence of weakness between the craton and the belt and a significant decrease of the brittle-ductile transition depth in the belt (Fig. 12). In these conditions, the strong reduction in the thickness of the brittle layer on the belt side prevents the development of major normal faults with significant vertical displacement (e.g., Ebinger et al., 1999). As a result, almost the entire deformation is accommodated on a major fault system formed on the craton side, where the basin depocenter is localized, and the margin on the belt side is characterized by a gentle monocline dipping toward the rift valley. Conversely, when the vertical layering of the crust and the depth of the brittle-ductile transition are similar in the belt and the craton, deformation is substantially symmetric, with comparable amounts of deformation on both rift margins and the axial depocenter.

A similar scenario, with a more symmetric deformation and the absence of a prominent rift valley, is observed with no intervening weakness between the craton and the belt. In these conditions, rather symmetric, diffuse deformation is observed for low strength contrasts. A more asymmetric deformation occurs for high strength contrasts, where there is localization of deformation and a major fault system develops at the craton margin (Fig. 12). However, the degree of asymmetry is not comparable to that reached with the presence of weakness and with lateral variations in brittle and ductile thickness at its sides.

As briefly discussed in the Introduction, three natural cases exemplify the influence of the strength contrast between an old, cold and resistant lithosphere and a weaker domain on the way rifts develop and propagate, and on a more detailed scale, their architecture and symmetry. These examples are illustrated in Figure 12 and described in the following.

The Baikal rift system is a long-lived (∼30 m.y.) area of continental extension composed of several late Cenozoic sedimentary basins extending for ∼1500 km in a northeast-southwest direction and displaying significant along-strike variation in width, symmetry, and subsidence (Fig. 1A; e.g., Logatchev and Florensov, 1978). Whereas narrow, elongated, asymmetric deep basins bounded by a single, southeast-dipping normal fault characterize the central and southwestern portions of the system (Lake Baikal area; e.g., Hutchinson et al., 1992; van der Beek, 1997; Petit and Déverchère, 2006), the northernmost basins are shorter and spread over a larger area (e.g., Logatchev and Florensov, 1978). The deepest (thickness of the sedimentary sequence to ∼9 km), narrowest depressions developed at the southern termination of the Archean Siberian craton, in correspondence to a major lithospheric suture bounding and separating it from the Paleozoic Sayan-Baikal fold belt (e.g., Petit and Déverchère, 2006). Conversely, the more distributed, en echelon northernmost basins developed farther from this suture, within the mobile belt (Fig. 1A). Our experiments indicate a strong rheological control on the distribution and architecture of extensional deformation between the two different lithospheric domains, supporting inferences from previous analog (Chemenda et al., 2002; Corti et al., 2011) and numerical (Petit et al., 2008) modeling. The modeling results suggest that the narrow, deep depressions hosting Lake Baikal are the result of extensional deformation localized within a nearly vertical weak suture separating the cratonic keel from the mobile belt (Fig. 12). A marked decrease of the brittle-ductile transition in the crust from the craton to the belt is responsible for the prominent asymmetry of the Baikal basins, with a master fault on the craton side and a monocline with no significant faulting on the belt side (e.g., Corti et al., 2011). In these conditions, the strong strain localization within the weak zone may lead to significant decompression melting and magma intrusion into the crust, which may compensate the expected Moho uplift, thus complicating the relations between bulk extension and crustal thinning (Thybo and Nielsen, 2009).

The more distributed deformation in the northeastern Baikal rift system suggests a decrease in the influence of the suture at the craton margin, which may result from variations in characteristics such as weak zone rheology, cross-sectional geometry (e.g., inclined versus vertical weakness; e.g., Petit and Déverchère, 2006), or width. Variations in the rheology and strength of the Sayan belt may also contribute to the observed variations in the style of extensional deformation, with local reductions in strength along the mobile belt favoring a delocalization of deformation (Petit et al., 2008).

Rifting in the West Antarctic Rift System (Fig. 1B) developed within a heterogeneous, weak lithosphere made of the microplates of West Antarctica at the margin of the East Antarctic craton, mostly during an important Cretaceous phase of orthogonal rifting (e.g., Behrendt et al., 1991). Extension resulted in a Basin and Range–like topography with main sedimentary basins separated by large basement highs within the weak lithosphere and localized faulting and strong flank uplift at the craton margin, resulting in the uplift of the Transantarctic Mountains.

In agreement with previous results (Bonini et al., 2007), our models suggest that rifting in West Antarctica has been largely controlled by a sharp contrast in rheology between the strong craton and a weak mobile belt, with no intervening major weakness. In these conditions, extensional strain was mostly accommodated at the boundary between the two lithospheric domains, with development of a major fault system responsible for the uplift of the Transantarctic Mountains (e.g., Bonini et al., 2007). The craton was essentially unaffected by faulting; extension was localized within a very weak lithosphere (with no strong upper lithospheric mantle), resulting in distributed deformation over areas (>1000 km) much wider than the lithospheric thickness. Distributed normal faulting was punctuated by large-strain core complex–like areas (e.g., Siddoway et al., 2004), similar to the wide rifting–style geometry of the Basin and Range of the western United States (Fig. 12). The resulting nearly uniform thinning of the weak lithosphere inhibited strong vertical thinning (necking) and asthenospheric upwelling, which in turn prevented the progression of the rifting process to continental break-up, despite the large magnitude of extension associated with the Cretaceous rifting (∼300 km; Trey et al., 1999; Cande et al., 2000), and inhibited significant decompression melting and volcanism in the area (Bonini et al., 2007).

The East African Rift System south of the Ethiopian Rift Valley is strongly controlled by the presence of the Tanzanian craton. The extensional deformation tends to avoid the strong cratonic lithosphere and the rift system splits into two branches surrounding it. The western branch displays a localized deformation, and extension in this area has been strongly controlled by significant lateral variations in strength at the craton border (e.g., Petit and Ebinger, 2000). Similarly to the Lake Baikal rift, strong strain localization along a weak zone at the western margin of the Tanzania craton (e.g., Petit and Ebinger, 2000; Morley, 2010) likely controlled the development and architecture of rifting, where a limited bulk extension has led to the development of narrow, elongated, and deep depressions (such as that occupied by Lake Tanganyika). As suggested by the experiments, variations in crustal structure (e.g., variations in thickness of brittle and ductile layers) at the sides of the weak suture may have made strong contributions to the along-axis change in rift (a)symmetry typically described for this rift (e.g., Morley, 1988).

Conversely, the eastern branch of the East African Rift is characterized by an along-axis transition from localized deformation at the craton margin in the Kenya rift to a wider deformation zone in the so-called Tanzania divergence, where the rift splays into different seismically active arms, with active extension distributed across zone >300 km wide (e.g., Ring et al., 2005). This transition is associated with the impingement of the southward-propagating rift with a strong lithospheric domain (Masai block) east of the Tanzanian craton (e.g., Nyblade and Brazier, 2002; Ring et al., 2005; LeGall et al., 2008). Thus, the variations in rift characteristics can be attributed to a decrease in the strength contrast when a weak suture is substituted by a strong terrain at the eastern side of the craton (Fig. 12). As supported by the models, this leads to a transition from a well-developed rift valley at the craton border to a series of subparallel basins developing within the strong lithosphere. Consistent with our modeling approach, subsidence of the individual basins is more limited (<3.5 km; Ebinger et al., 1997) in the Tanzania divergence than in the narrow, deep basins of strongly localized deformation (as much as 6–7 km in the narrow basins of the western branch; e.g., Morley, 1988); basin depths decrease further to the south, where the width of the area affected by faulting further increases (Ebinger et al., 1997).

Lithospheric-scale centrifuge models simulated extension between an old, cold, and resistant cratonic lithosphere and an adjacent weaker mobile belt. The experiments focused on different characteristics of continental rifting, such as the width of the deformed zone, basin symmetry and subsidence, and lithospheric thinning, and the dependence on the craton and belt strength contrast, as well as the presence or absence of a weak zone between the two lithospheric domains. The experimental results suggest the following main conclusions.

The presence of a weak zone between the craton and the mobile belt strongly localizes deformation, leading to the development of narrow, deep rift valleys corresponding at depth to a marked lithospheric thinning. Faulting does not migrate outside the weak region throughout the extension process, and the deformation corresponds to a necking of the extending continental lithosphere. This behavior is independent of the craton and belt strength contrast; however, depending on the pre-rift rheology (in particular on the presence of a significant decrease of the brittle-ductile transition depth in the belt domain), the resulting rift can be largely asymmetric, with a single, major border fault system on the craton side.

When the weak zone is absent, deformation is typically more distributed and lithospheric thinning more homogeneous. Mechanically, the lithosphere is not affected by a necking instability in correspondence to a single, well-developed rift valley, but it is affected by more distributed instabilities expressed at the surface as numerous normal faults and basins extending over wider regions. In these conditions the strength contrast between the craton and the belt controls the width of the rift and its evolution in time. When the strength contrast between the craton and the belt is minimal, no major faults with significant vertical throw are able to form at the boundary between the two lithospheres, and deformation affects a wide region inside the strong mobile belt from the initial stages of extension. Conversely, for high strength contrasts, more asymmetric deformation is localized at the craton-belt boundary at the beginning of extension, with development of a major fault system on the craton side; minor faulting is then able to propagate inside the weak belt, with the deformed zone widening with time and progressive extension.

In general, the models indicate that increasing the rift width results in the subsidence (and the lithospheric thinning at depth) decreasing, and the number of normal faults increasing. Extrapolating this observation to the duration of the rifting process, the models suggest that the strong localized thinning of the lithosphere in narrow regions is expected to lead to a rapid lithospheric necking and break-up, whereas the more uniform deformation and thinning of the lithosphere in wide zones is expected to result in a slower evolution and retarded final plate separation.

Comparison of model results with the Baikal rift system, the West Antarctic Rift System, and the East African Rift System supports the conclusion that the juxtaposition between an old, cold, and resistant lithosphere and a weaker domain has a strong influence on the way rifts develop and propagate. Thus, horizontal variations in the rheology of the lithosphere, and not only its vertical layering, play a major role in controlling the distribution and architecture of the extensional deformation.

We thank reviewers Julia Autin and Laurent Michon and associate editor Francesco Mazzarini for detailed constructive comments that helped to significantly improve the paper. We also thank Federico Sani for comments.