We perform both brittle (one-layer) experiments (models BI) and brittle–ductile (two-layer) experiments (models CE) to reproduce a brittle upper crust and a brittle–ductile crust simulating cases of “strong” and “weak” crusts, respectively (Fig. 2). For both types of experiments, the 4 cm thick single or double layers rest on top of a foam layer (8 cm thickness). To simulate the brittle part of the crust, whose behavior is of the Mohr–Coulomb type , we use Fontainebleau quartz sand (NE34, Sibelco, France, D50=210 µm) . This material has a peak friction of 0.74 and an immediate reactivation friction (after 10 s) of 0.64 . Healing of this material increases the reactivation friction to 0.68 after 2.6 h (maximum duration of our experiments) (Table 1). The cohesion (C) of Fontainebleau quartz sand is around 60–70 Pa e.g.,. We sieve the dry sand from a height of ∼15–20 cm, ensuring that its density is 1400 kg m-3 (Table 1). Dry sand exhibits frictional plastic behavior and the geometry of structures that form when deformed does not depend on the applied strain rate. The strength of the sand layer is calculated following : 1σ1-σ3=(K-1)ρgz(1-λ)+S(for compression),2σ1-σ3=(K-1)Kρgz(1-λ)+SK(for extension), with S=2Csin⁡(90+ϕ)1+cos⁡(90+ϕ) and K=1-cos⁡(90+ϕ)1+cos⁡(90+ϕ), where ϕ is the angle of internal friction (Fig. 2).

To simulate the ductile part of the crust, we use PDMS silicone whose density is 965 kg m-3 and viscosity μ is around 3.5×104 Pa s-1 (Table 1). This material has a Newtonian rheology (n=1) for strain rates lower than 10-2 s-1 e.g.,. Strain rates based on the length of the models and applied velocities at model boundaries are in the range of 1.2–4.5×10-5 s-1. The strength of the layer of silicone putty (σ1-σ3) varies with imposed strain rates as 3σ1-σ3=ηϵ˙, where η is the viscosity and ϵ˙ the strain rate. For the values of applied strain rates, differential stress is on the order of 0.4–1.6 Pa. Initial strength envelopes for both types of models and different parts of the models are shown in Fig. 2. The polyurethane foam RG35 used at the base of the model has a Poisson coefficient of 0.12 and allows producing a linearly varying velocity field (no velocity discontinuities) at the base of the deforming pile by compressing it to obtain shortening or letting it decompress to obtain stretching. However, there is a limitation on the amount of applied stretching/shortening. After 20 % of shortening, the foam starts buckling and we therefore limit the amount of applied stretching/shortening under this threshold.

We follow the scaling procedure shown, e.g., in and based on , , and . The scaling parameters are given in Table 2. Stress ratios between the laboratory and nature σ* are calculated as follows: 4σ*=ρ*g*L*, where ρ* represents the density ratio, g* the gravity ratio, and L* the length ratio. Considering that we simulate the upper 15 km of the crust with our 4 cm thick pile of material, it gives σ*=1.33×10-6, i.e., that 1 Pa in the lab corresponds to 0.75 MPa in nature.

For the two-layered models, the density ratio between the brittle and ductile parts of the crust is high (1.45). Increasing the density of the viscous layer would have resulted in a strong increase in its viscosity, which would have required applying speeds too low for the capacities of the engines used. We acknowledge that it leads to buoyancy forces that may trigger gravitational instability of the silicone layer and possible amplification of the folding of the brittle–ductile interface during shortening/stretching. However, given the relatively high viscosity of the silicone layer and the limited amount of deformation and short duration of these models (between 1.6 and 2.6 h), we consider that such a process may remain limited during the experimental time frame.

The strain rate ratio ϵ˙* is obtained as the ratio between the stress ratio σ* and the viscosity ratio η*: 5ϵ˙*=σ*/η*.

We assume a natural viscosity for the crust of 1021 Pa s, within the range of proposed values in varying tectonic contexts (η=1019-1023 Pa s; e.g.,). It gives ϵ˙*=3.8×1010; i.e., imposed strain rates correspond to strain rates of 0.32-1.18×10-15 s-1 in nature.

The time ratio t* can be obtained from 6ϵ˙*=1/t*.

It implies that 1 h in the lab corresponds to 4.34 Ma in nature. Given that the duration of our models is at a maximum 2.6 h, we simulate geological processes lasting for ∼11 Ma at a maximum.

The velocity ratio v* is obtained from 7ϵ˙*=v*/L*.

The imposed values for extension/shortening rates between 0 and ∼60 mm h-1 in the lab correspond to 0–5.2 mm yr-1 in nature, typical values for continental rifting e.g.,, subduction- and collision-related (e.g., Andes, ; Alps, ; Zagros, ), or intra-continental collision orogens (e.g., Pyrenees, ; Tian Shan, ).

The dynamic similarity between our experiments and the natural case for the viscous regime is verified by computing the ratio between lithostatic pressure and viscous strength (Ramberg number Rm): 8Rm=(ρgL2)/(ηv), which gives a value of ∼37 considering the maximum deformation velocity for both the model and the natural case (Table 2). For the brittle regime, the dimensionless friction coefficient is similar (0.6–0.7) in the laboratory and in nature. The ratio between gravitational stress and cohesive strength Rs=ρgL/C is also similar with values ranging between 7.85 and 9.16 considering a cohesion in nature of 45–53 MPa (Table 2), which is well within the range of cohesion values measured on natural rocks at 24–110 MPa .

The 60×60×8 cm layer of foam is initially compressed in one direction and is maintained in this state for the rest of the preparation phase. For the two-layer models, we place a pre-cut PDMS layer with dimensions of 50 cm × 50 cm × 2 cm (Fig. 2). Sieving of the sand is performed from a distance of 15–20 cm above the PDMS layer, and the sand is leveled with a rigid plate until the desired thickness is achieved. The brittle part of the crust is made of white quartz sand that is randomly sprinkled on top with black colored sand in order to allow particle detection for digital image correlation. Before deformation, the models cover an area of ca. 46 cm × 46 cm that would represent an area of 172 km × 172 km in nature.

We also include “seeds” in our models to help localize deformation. They may represent weak zones inherited from previous phases of deformation. This weaker zone in our experiments is also compatible with the requirement of a finite-width low-viscosity zone underneath the fault zones possibly caused by grain size reduction, shear-heating, and localized presence of fluids e.g.,. These seeds are linear pieces of silicone that are placed on top of the silicone layer or directly on top of the foam for the brittle models (Fig. 2). They are placed at the center of the models, orthogonal to the extension direction, have a rectangular shape, and extend along the entire length of the model. The dimensions of the seed are 46 cm × 1.5 cm × 1.5 cm for the brittle–ductile models and 46 cm × 1.5 cm × 1 cm for the brittle models (Fig. 2). The strength of the crust is decreased at these locations owing to the reduced thickness of the overlying sand layer (Fig. 2), which in turn may help deformation to localize.

The layer(s) are then deformed by applying a constant velocity boundary condition at the edges of the model through pistons activated by step motors, which allows us to precisely control the stretching rate to shortening rate ratio. Stretching is applied to both edges of the models by letting the foam decompress while shortening is only applied to one side of the models, the other side having a no-motion boundary condition (Fig. 2). We arbitrarily consider the non-moving wall as the north in our experiments. We performed two types of experiments: (i) one-stage experiments with either shortening only or synchronous orthogonal shortening and stretching and (ii) two-stage experiments with a first phase of 5 % stretching and a second phase with either shortening only or synchronous orthogonal shortening and stretching in order to study the possible reactivation/inversion of structures formed during the first stage (Fig. 3). Applied velocities for stretching vary between 0 and 75 mm h-1 and for shortening between 0 and 43 mm h-1 (Fig. 3).

We do not include surface processes in the models (erosion, deposition), especially in between the two stages of deformation, meaning that the created grabens remain unfilled when the second stage of deformation starts. While we acknowledge that redistribution of mass associated with surface processes may impact stress distribution and further deformation e.g.,, we wanted to ensure similar conditions between models that are difficult to achieve when manually intervening during the course of the experiment.

Experiments are recorded from the top by a DSLR camera (Nikon D3300) taking pictures every 2 min. Pictures are then automatically analyzed using an image cross-correlation technique, particle image velocimetry (PIV), using the PIVlab software . We pre-process the images with a CLAHE (contrast-limited adaptive histogram equalization) filter with a window size of 64 px to enhance contrast in the pictures and allow particle detection. PIV analyses are made with the direct Fourier transform correlation with multiple passes and deforming windows with interrogation areas of 128, 64, and 32 px and with a step of 50 %. PIV results are then calibrated using spatial scales set on top of the models. We obtain velocity maps with a spatial resolution of 16 px, corresponding to ∼4 mm (∼1.5 km in nature).

Velocity fields obtained from the PIV analyses are then processed with the StrainMap algorithm that allows tracking the cumulative deformation field and as such mapping the distribution of deformation over time. In particular, this algorithm is based on the description of shape changes in terms of Hencky strains. It allows us to discriminate between coexisting strike-slip faults, thrust faults, and normal faults, and their evolution over time and as such efficiently complement inherently subjective visual inspection. Videos and strain analysis of the 12 experiments are available in .

Our models confirm that under plane-strain conditions, i.e., no strain in one of the horizontal directions (models BI01 and BI10 and first stage of models BI08, BI11, CE17, CE18, and CE20), the crust accommodates deformation through structures orthogonal to the direction of transport, i.e., either normal faults when the maximum principal stress (σ1) is vertical or thrust faults when the minimum principal stress (σ3) is vertical. However, when boundary conditions satisfy non-plane-strain conditions (i.e., strain occurs along the three principal axis), deformation is accommodated along a combination of normal faults, thrust faults, and strike-slip faults, whose location and presence/absence appear to depend on the applied boundary conditions, pre-existing heterogeneities, and possibly the amount of accumulated deformation.

Model BI05 (brittle model with Ve/Vs = 1.4) exemplifies the possible coeval activity of N–S conjugate normal faults, E–W thrust faults, and NE–SW and NW–SE conjugate strike-slip faults (Fig. 8). After 10 % of E–W stretching and 7.7 % of N–S shortening, in the center of the model, the type of structure evolves from compressional structures to the south (close to the piston) to extensional structures to the north. The transition from one regime to another results from the permutation of one of the horizontal principal stress axes with the vertical one, controlled by the relative magnitude between the principal stresses as described by b=(σ2-σ3)/(σ1-σ3). The Mohr–Coulomb analysis for a frictional material like sand also implies that the differential stress necessary for material failure is maximal for thrust faults (Fig. 8). This could explain why thrust faults are only found close to the southern moving piston where the maximum horizontal stress is applied (zone 1). In this area, σ1 is horizontal and N–S oriented, while the minimum stress (σ3) is vertical. Going further north from the piston (zone 2), σ1 decreases, which implies a lower σ3 at failure and a vertical intermediate principal stress (σ2), favoring the development of strike-slip faults. Further north (zone 3), the normal faults that were created during the first stage of extension are reactivated as normal faults, thus indicating that σ1 is vertical. A progressive increase in the southern wedge thickness accompanying N–S shortening in the absence of erosion would imply an increase in the vertical stress. As a consequence, thrust faults would propagate toward the north (as evidenced between 4.2 % and 7.7 % of shortening for model BI05; Figs. S1c and 4c) and could possibly reach places where strike-slip faults and normal faults were previously active. Therefore, the redistribution of stress resulting from deformation could drive temporal variations in the tectonic regime at a specific location.

For brittle models, the coexistence of the three types of structures (after 10 % of E–W stretching) starts for Ve/Vs ratios > 1.4 (Fig. 4). For lower Ve/Vs, σ1 is horizontal and N–S oriented and maximal close to the piston applying shortening, favoring the development of thrust sequences in the southern part of the models. Close to the western and eastern pistons applying stretching, in the case where Ve/Vs≠0, the vertical stress becomes the maximum stress and N–S normal faults develop. For values of Ve/Vs≥2, the E–W horizontal stress is low enough in the entire model to maintain as the minimum stress σ3, which in turn implies that even if the N–S horizontal stress is the maximum stress σ1, the crust cannot fail along E–W thrust faults but fails along strike-slip faults. However, we cannot preclude that thrust faults could also develop at later stages for these models with high Ve/Vs ratios, as experimental limitations prevent us from imposing large amounts of N–S shortening. In model BI07 for instance, it is at a maximum of 3.6 %, which may be insufficient to locate deformation along an E–W thrust fault.

For brittle–ductile models, over the range of tested Ve/Vs, we do not observe the coexistence of the three types of structures within the same model (Fig. 6). We only observe E–W thrust faults associated with conjugate strike-slip faults for models with low Ve/Vs and no initial E–W extensional stage. This implies that the stress conditions required to develop N–S normal faults (E–W horizontal stress is σ3 and vertical stress is σ1) are never encountered in the model. The first condition is met as testified by the presence of strike-slip faults. Instead, the second condition is not met, which could partly be explained by the limited thickness of the brittle portion of the crust, which is half that of the brittle-only models. For models with an initial E–W extensional stage, we observe at the final stage of deformation a combination of N–S normal faults and conjugate strike-slip faults and no clear E–W thrust faults, even when Ve/Vs is low (model CE18). For the latter, shortening is active in some areas of the model (Fig. 6c and point 3 in Fig. 7b), but it does not localize along discrete structures. For weak crusts, crustal thinning associated with an initial stage of extension therefore appears to control the future development of thrusts during the subsequent stage of combined shortening/stretching. However, the amount of applied deformation remains limited with only 5 % of horizontal E–W stretching (and ∼10 % of horizontal N–S shortening), and we cannot preclude that further shortening would eventually result in thrust development.

In the brittle-only models, a majority of the structures are consistent with the Coulomb fracture criterion considering either a N–S-oriented horizontal σ1 or an E–W-oriented horizontal σ3. In particular, thrust faults are E–W oriented, normal faults are N–S oriented, and strike-slip faults are generally oriented at ∼30∘ from a N–S σ1. However, in the northern part of the models, strike-slip faults do not obey this criterion and exhibit larger angles with respect to a N–S horizontal σ1, with values around 60–70∘ (e.g., model BI05, Figs. 4c and 9a). These anomalously oriented strike-slip faults could result from a combination of factors. As they nucleate from the edge of the model, we cannot exclude that they result from some unwanted boundary effects associated with the high friction wall–sand interface. As a result of the applied stretching and boundary effects, the maximum principal stress σ1 may have rotated from a N–S direction toward a NE–SW direction in the eastern part of the model and NW–SE direction in the western part of the model, possibly explaining why these strike-slip faults do not lie at 30∘ with respect to the imposed N–S shortening. However, one can also notice that not all of these strike-slip faults have the same exact orientation (Fig. 4d and e), some of them being directed toward the northward termination of the normal faults bounding the central graben. Their orientation therefore could also be controlled by the graben structure that forms in the center of the model above the crustal seed.

Interestingly, in brittle models with an initial stage of extension (e.g., model BI08, Figs. 4g and 9b), it is not only the northern strike-slip fault orientation that departs from the orientation expected from the Coulomb fracture criterion with a N–S shortening but also that of some of the strike-slip faults that develop above the wedge in the southern part of the model. While we cannot preclude some boundary effects here too, these faults with a larger than expected angle with respect to σ1 also connect with the normal faults bounding the central graben formed during the initial stage of stretching. In comparison, model BI05, which shares the same stretching / shortening ratio and almost the same amount of total stretching (∼10 %), does not show any anomalous strike-slip faults in the southern part of the model (Fig. 9a). Pre-existing structures may also exert a control on the geometry of subsequent structures even for areas close to where shortening is applied.

For models with a brittle–ductile crust, strike-slip faults are expected to range in between 30 and 45∘ following the Coulomb criterion and the slip-line theory e.g.,. Most of the strike-slip faults that crosscut the models do indeed show strikes compatible with the overall state of stress, i.e., N30–45 for left-lateral strike-slip faults and N135–150 for right-lateral strike-slip faults. However, in models with an initial stage of stretching (e.g., model CE20, Figs. 6d and 9c), strike-slip faults also develop with anomalous orientations. In particular, at model corners, strike-slip faults bend with angles that become larger with respect to the N–S shortening direction, possibly indicative of some boundary effects. More interestingly, some strike-slip faults are oriented almost parallel to the shortening direction in areas that were previously affected by normal faulting (Figs. 6 and 9c). This clearly indicates that inherited structures, here in the form of previously developed N–S normal faults, can control the subsequent location and geometry of structures.

Overall, our models show that whatever the strength of the crust, its past deformation history, and the relative ratio of stretching rate over shortening rate, both normal faults and thrust faults remain with similar orientations, i.e., N–S and E–W, respectively. Instead, strike-slip faults exhibit a wider range of possible orientations with respect to the shortening (or stretching) direction. As such they could give us insights into the tectonic context in which they formed. In particular, strike-slip faults with angles of up to ∼65∘ with respect to a N–S shortening direction are found for strong crusts. They particularly form far away from the shortening location when the crust has not been previously deformed, and their anomalous orientation may be indicative of a local perturbation of the stress field owing to a possible combination of boundary effect, dominance of stretching in this area, and coeval formation of a graben in the center of the domain. High-angle strike-slip faults can also be observed closer to the area where shortening is applied if the crust has been previously extended (and thus weakened). In this case, strike-slip faults branch to the former location of normal faults that bounded the central graben. For weak crusts, strike-slip faults can even be parallel to the shortening direction if they reactivate former extensional structures.

Under comparable boundary conditions, the strength of the crust plays a fundamental role in controlling the location of deformation and the types of structures that accommodate deformation. Models without initial stretching phase and similar Ve/Vs (e.g., models BI09 and CE16, Ve/Vs=0.9) exhibit significantly different patterns of deformation (Fig. 10a and b). A strong crust (thick brittle part) favors the development of in-sequence E–W thrust faults verging to the north to accommodate shortening, while stretching is accommodated through N–S normal faults at the western and eastern boundaries of the model (Fig. 4b). Instead, when the entire model is made of a weak crust, shortening results in a doubly vergent system of conjugate E–W thrust faults and in large-scale conjugate strike-slip faults that also accommodate the stretching. The lack of normal faults indicate that the horizontal stress remains high enough in the entire model to prevent the vertical stress to become σ1. This can readily be explained by the decrease by a factor of 2 of the thickness of the brittle crust in the weak models. Interestingly, in the strong model (BI09), there is also a variation in the thickness of the brittle part of the crust owing to the presence of the viscous seed in the center of the model. However, it does not result in significant lateral variations in deformation style, due to the small variation in brittle thickness (decrease by 25 %) and/or to the fact that the area of relatively weak crust only represents a small fraction of the entire model (∼1.2 % of the crustal volume in model BI09 instead of 51.2 % in model CE16). A more systematic change in the width and thickness of the weak part of the crust would be required to isolate the main controlling parameter.

Interestingly, model CE17, which shares the same boundary conditions as model CE16 but with an initial stage of E–W stretching, exhibits a significantly different distribution of deformation during the second stage of deformation. This initial stage of stretching locally modifies the crustal strength by creating zones of thinned ductile and brittle portions of the crust accommodating the extension (Fig. 10c). These areas of even weaker crust concentrate deformation during the second stage through normal faulting or N–S strike-slip faulting that were both absent in model CE16. In turn, there are no thrust faults accommodating the shortening.

Experimental results again highlight the importance of past tectonic history and associated inherited structures and changes in crustal strength on the distribution of deformation and the types of structures that form during subsequent phases of deformation (Fig. 11). As such, the association of specific types of faults cannot be used as an indicator of the relative rate of shortening and stretching, unless previous history of deformation is properly constrained.

Many modeling studies have previously investigated deformation associated with the inversion of extensional half-grabens subject to subsequent contraction e.g.,and references therein. In general, the applied directions of stretching and successive contraction are parallel (in 2D numerical or analog models) or oblique but at low angles. However, in nature this may not always be the case. In Patagonia for instance, Cretaceous extensional structures formed in the San Jorge basin as a response to N–S stretching. This area has subsequently undergone E–W Cenozoic contraction, parallel to the direction of the normal faults. We therefore more specifically explored the conditions leading to the reactivation of extensional structures under fault-parallel contraction (and possible fault-orthogonal extension).

From our experimental dataset, there is no clear evidence that normal faults lying parallel to the contraction direction are inverted as contractional structures during the second stage of deformation (Fig. 11). While these areas are favorable to further contraction owing to their thinned crust and associated low vertical stress, the angular relationship between pre-existing normal faults and the direction of contraction (∼0∘) prevents the normal faults from being inverted. We further show that if stretching orthogonal to the contraction direction is inhibited, there is no reactivation of pre-existing normal faults as normal faults or strike-slip faults (model BI01). Instead, allowing orthogonal stretching always results in the reactivation of normal faults. However, different regimes are found depending on the boundary conditions and crustal strength. In models with low Ve/Vs and a weak crust (model CE18), extensional structures are preferentially reactivated as strike-slip faults, participating in the accommodation of the shortening that dominates. As Ve/Vs increases and becomes closer to 1, pre-existing normal faults are reactivated as either normal faults or strike-slip faults (models CE17 and CE20). Both types of faults therefore develop and are juxtaposed as almost parallel structures without involving changes in boundary conditions. Finally, when stretching rate dominates over shortening rate, pre-existing normal faults are reactivated as normal faults (models BI08 and BI11).

While our experimental study was not initially designed to reproduce any specific natural case, the observed deformation patterns resemble those documented in certain areas where deformation is accommodated by different combinations of normal, thrust, and strike-slip faults. In general, these complex structural frameworks take place during continental plate convergence, indentation, and lateral escape of lithospheric blocks associated with continental collision (Fig. 12). From our models, we observe that at least two different deformation regimes coexist when Ve/Vs=0.7-0.9 and the three regimes operate simultaneously mostly when Ve/Vs is ≥1.4 (Figs. 4 and 6). The latter could indicate a minimum value for this ratio for effective lateral escape to occur. In nature, the tectonic escape of crustal blocks takes place along conjugate strike-slip faults, commonly referred to as indent-linked strike-slip faults or V-shaped conjugate strike-slip faults e.g.,and references therein that take place along with variable degrees of crustal extension within the escaping crustal blocks. Typical V-shaped conjugate faults have been documented in the eastern Anatolia region e.g.,Fig. 12a, the eastern Alps e.g.,Fig. 12b, the Tibetan Plateau e.g.,Fig. 12c, and the central Asian intraplate region e.g.,Fig. 12d. Previous mantle-scale brittle–ductile analog experiments successfully reproduced the formation of V-shaped conjugate faults during continental indentation e.g.,. However, a direct comparison with our models is difficult because these previous models were aimed at simulating larger domains (several hundreds of kilometers) and the lateral escape of material was not kinematically controlled. Instead, 3D lithospheric-scale numerical models by imposing coeval orthogonal extension and shortening at the edges of the models also show that deformation is accommodated along strike-slip faults with a degree of localization that increases with a decrease in the viscous strength of the lower crust. Results obtained in our models at the crustal scale could therefore be possibly upscaled at the lithospheric scale.

In our models, whether brittle or brittle–ductile, we also observe the development of conjugate strike-slip faults that are accompanied by different degrees of normal faulting during compression, which are reminiscent of those structural systems seen in nature (BI05-08, Fig. 4c and g; CE16–20; Fig. 6). Our results are partly similar to those by (model M-3), in which crustal-scale brittle–ductile analog models were subjected to simultaneous shortening and orthogonal extension at Ve/Vs=1 imposed by one piston in each direction. Despite the slightly different boundary conditions (stretching being allowed only along one boundary), both approaches show crustal escape through a V-shaped strike-slip system. Instead, models by that used a similar setup than with extension applied to one wall during compression but with a brittle crust, do not show V-shaped conjugate strike-slip faults. Interestingly, in these experiments, the authors used a low Ve/Vs value of 0.2. The absence of strike-slip faults is therefore consistent with our minimum value (Ve/Vs>0.9) for effective lateral crustal escape. Strike-slip fault systems active during compression have also been previously addressed by brittle analog models . However, in these studies, strike-slip fault strike and activity were constrained by the chosen setup that included confined walls and prescribed fault direction along a mobile pre-existing discontinuity. Because of these constraints, the experiments did not develop typical V-shaped strike-slip systems.

In our models, we did not include the effect of compression obliquity, which has been acknowledged as an important factor driving simultaneous conjugate strike-slip systems and orthogonal extension during indentation . Also, we note that experiments including an overall weaker crust, simulated by adding a silicon layer as an analog for ductile lower-crust materials, are more prone to have well-developed conjugate strike-slip systems that crosscut the entire model (CE16–20; Fig. 6). This is compatible with previous findings from analog models by that explored the role of orogen parallel flow and rheological stratification on vertical and lateral development of structures in hot orogens. These authors simulated a hot orogenic crust by including a weak lower crust with a variable buoyancy and a lithospheric mantle that jointly allowed orogen parallel flow during compression. In these experiments, the development of conjugate strike-slip faults within the orogen appeared to be a common feature in models with buoyant ductile lower crust, whereas in models with a non-buoyant lower crust, conjugate shear zones were generally absent, except in the outer regions of the extruding orogen.

In general, our results are compatible with natural examples analyzed in Fig. 12a–d, where Cenozoic conjugate strike-slip systems formed on board thick and hot and hence weak orogenic crusts of the Anatolian and Tibetan plateaus, Central Asian intraplate orogenic system, and the Alpine orogen. Therefore, a weak crust is an important factor contributing to lateral tectonic escape in those settings by allowing the formation of fault-bounded brittle upper-crust blocks translated laterally by strike-slip systems and constrictional ductile flow in the lower crust .

In models that include an initial stage of extension, and under the applied boundary conditions, we observe variable degrees of extensional to strike-slip reactivation of the previous extensional basins during basin-parallel shortening (BI08-11 and CE17–20, Figs. 4g, h, 6b, c, d). These results differ from previous basin inversion analog models by , , and , which also applied a basin-parallel shortening after a first stage of orthogonal extension. In these models, despite the low angle between basin strike and shortening direction, thrust faults formed parallel or at a low angle with the shortening direction, ultimately producing basin inversion. This difference could result from the boundary conditions applied in our models BI08-11 and CE17–20, where shortening is accompanied by coeval orthogonal extension, which could explain why basin inversion does not occur in our models. However, this may not be the only controlling factor given that in model BI01, which does not have stretching applied during shortening, there is no evidence of shortening-parallel thrusting during the second stage of deformation (Fig. 4f). This implies that internal parameters are also crucial in allowing basin inversion during orthogonal shortening. In particular, the presence of a larger weak zone at the base of the model, a thinner rift basin crust, and inclusion of surface processes may yield a weaker crustal area leading to an easier localization of deformation during compression. also showed that folding/thrusting parallel to the shortening direction can be achieved at the transition between crustal blocks with different thicknesses/strengths in models with isostatic compensation, showing that anisotropy and isostasy are two additional key ingredients that may favor basin inversion under parallel shortening. Besides these contrasting results, our models are compatible with natural cases of synorogenic foreland rifting–transtensional reactivations that are complex processes of basin reactivation during regional compression . These cases have been well documented in the Baikal region in Central Asia and in central Patagonia (Fig. 12d and e). In both cases, former extensional basins orthogonal to neighboring plate margins were subsequently reactivated in transtension by a distal basin-parallel compressional stress field (Fig. 12d and e).