Introduction

Rift systems and divergent continental margins are the result from extension of the continental lithosphere. Understanding the physical parameters that control the development and evolution of faults in rifts and divergent margins is important for their interpretation in nature and for assessing their spatial distribution (e.g., orientation and spacing) and temporal evolution. Structures such as those interpreted from seismic sections across the North Sea graben suggest that thickness and distribution of weak layers (such as salt) and pre-rift faults have played a major role on the subsequent development of graben structures (Stewart et al., 1996; Stewart, 1999). Other factors that are considered to influence the geometry and evolution of rifts include the orientation of the regional stress field (i.e. extension vs. oblique extension), synkinematic sedimentation and strain rate (salt has a time-dependent rheological behaviour).

Extensional analogue models to date have largely been of two kinds: (i) those where a deformable hanging-wall block is translated over a rigid, non-deformable footwall, and (ii) those that involve extension above a basement that undergoes a stretching deformation. Although the first type of model has yielded valuable information on specific extension-related structures, such as roll-over anticlines in the hangingwall of listric normal faults (e.g., McClay et al., 1991), the fact that the footwall is not allowed to undergo deformation is a limitation for modelling rifts and passive margins. Extension experiments of sandpacks above a stretching basement by McClay & Ellis (1987) showed that the geometries of both faults and fault blocks change with progressive deformation, and that extension above a sloping basement resulted in faults dipping in the same direction as the detachment.

We carried out experiments to investigate the influence of brittle-viscous multilayers on faulting in models undergoing a stretching deformation. Models were analysed by X-ray computerised tomography (Hounsfield, 1973) a non-destructive technique that allows a detailed study of the internal geometry and kinematics of analogue models.

Analogue materials and experimental procedure

In normal gravity experiments dry granular materials with low cohesion are commonly considered to be a good analogue for brittle rocks in the upper crust, because they obey the Mohr-Coulomb criterion of failure. Viscous materials on the other hand, are generally used to simulate viscous flow of salt or evaporites in the upper crust or rocks in the lower crust. We selected quartz sand (grain size: 80-200 µm) and corundum sand (grain size: 90-125 µm) as brittle analogue materials. The angles of internal friction for quartz and corundum sand are 35° and 36°, respectively, and these materials can be used to simulate upper crustal rocks which have comparable angles of internal friction at low normal stresses (Byerlee, 1978). We used a Newtonian viscous polymer (polydimethylsiloxane, PDMS) as viscous analogue material. This transparent material has a linear viscosity of 5 x 10⁴ Pa.s at room temperatures and at strain rates below 3 x 10^-3 s^-1 (Weijermars, 1986).

The experimental apparatus comprised a rectangular box, whose wooden base was overlain by an alternation of nine plexiglass bars (each bar is 5 mm wide, 5 cm high and 80 cm long) and eight foam bars (each bar is 3 cm wide, 5 cm high and 80 cm long). Before constructing the stratified brittle-viscous analogue model, the assemblage of plexiglass and foam bars was shortened 6 cm by displacement of a mobile vertical wall. In the shortened state, the width of each foam bar was reduced to ca. 2.25 cm, whereas the width of the plexiglass bars remained unchanged. The brittle-viscous multilayer model was then constructed on top of the shortened assemblage. A 1 cm thick layer of viscous PDMS was placed directly over the assemblage of plexiglass and foam bars. The stratified model was completed by adding brittle and viscous layers. The presence of foam bars alternating with plexiglass bars in combination with the overlying basal viscous layer prevented localisation of deformation near the extending mobile wall. The foam bars “decompressed” during extension and deformation was distributed over the width of the model. Displacement of the vertical mobile wall, driven by a motor, produced extension of the model at a constant velocity of 2.4 cm/h.

Experiment 077 simulated single stage rifting. The length of the brittle-viscous model was 80 cm, and the initial width was 22.5 cm. The undeformed multilayered model consisted of a 1 cm thick basal viscous layer (PDMS) and an upper, 5 mm thick viscous layer embedded in brittle strata made up of corundum and quartz sand (Fig. 1). Total initial height of the model was 3.5 cm. The upper viscous layer was placed adjacent to the extending mobile wall. It had one PDMS-sand boundary parallel to the extension direction, and the opposite boundary oblique to the extension direction (at an angle of 30°). The width of the upper viscous layer was 15 cm, and the length of the layer near the mobile wall was 40 cm. Total extension of the model amounted to 5 cm or 22.2%.

Figure 1. Experimental set-up for single stage rifting experiment 077 schematically illustrated by a plan view and two vertical sections. Two areas (077a and 077b) were analysed by X-ray tomography. Initial grid spacing on surface was 4 cm; grid consisted of coloured sand.

In the multiple rifting experiment 078 (Fig. 2), the initially stratified model consisted of a 1 cm thick basal viscous layer, overlain by 1 cm sand. The length of the model was 80 cm and the initial width was 20.5 cm. During the first stage of rifting the model was extended by 2.5 cm or 12.2%. Conjugate normal faults formed in the brittle sand layers and extended down to the top of the viscous layer (Fig. 2). The experiment was temporarily halted and the grabens were filled with corundum powder. An additional, 5 mm thick layer of corundum was sprinkled on top. A 5 mm thick viscous PDMS layer was now placed on top covering part of the corundum layer, and the PDMS layer itself was overlain by 10 mm of sand. The width of the PDMS layer was 22 cm and it was placed right adjacent to the extending mobile wall. Its shape was very similar to the upper viscous layer in experiment 077. It had one PDMS-sand boundary parallel to the extension direction, and the opposite boundary at an angle of 30° to the extension direction. The second stage of rifting consisted of an additional 5 cm extension. Final width of the model was 28 cm and total extension (with respect to undeformed stage) was 36.6%.

Figure 2. Plan view and vertical sections depict the set-up for multiple rifting experiment 078 before the onset of the second rifting event. Two domains (078a and 078b) were analysed by X-ray tomography. Initial grid spacing on surface was 4 cm.

Deformation of the analogue model took place in the investigation field of a helical X-ray computerized tomographer, and 3-D volumetric raw data were acquired of two parts of the model (see Fig. 1 and 2) at the initial undeformed state and after every 5 mm of progressive extension. The 3-D raw data allowed the computation of contiguous cross-sectional slices perpendicular to the long dimension of the model, i.e. parallel to the extension direction. Computer visualisation software was used to reconstruct the analogue model in three dimensions and to reconstruct horizontal and longitudinal sections. On the basis of 3-D reconstructions, computer animations were created in order to study the spatial evolution of structures at a specific stage, or to study the 4-D evolution (i.e. 3-D evolution with time) of the model.