TSK 11 Göttingen 2006 Schmatz et al.
 Figure 1: Sandbox-apparatus. Water-
 saturated setup of layered sand-clay exper-
 iment. Stage: Immediately after deforma-
 tion.
 Experimental study of the
 evolution of fault gouge in
 layered sand–clay sequences
 Poster
 Joyce Schmatz1 Peter Vrolijk2 Janos
 L. Urai1 Steffen Giese3 Martin
 Ziegler3 Wouter van der Zee4
 This study focuses on clay smear pro-
 cesses during fault gouge evolution in
 sand-clay sequences at depths up to
 2 km. A clay-rich fault gouge can dra-
 matically lower the fault’s permeability,
 and prediction of this process is there-
 fore relevant in groundwater modelling
 and hydrocarbon geology (Fulljames et
 al. 1997, Yielding et al 1997, van der
 Zee et al. 2003, 2005).
 We constructed an ‘underwater’ sand-
 1 Geologie-Endogene Dynamik, RWTH
 Aachen, Germany 2 ExxonMobil Up-
 stream Research Co., Houston, TX, USA
 3 Geotechnik im Bauwesen, RWTH Aachen,
 Germany 4 GeoMechanics International,
 Mainz, Germany
 box to deform layered sand-clay models
 of 20 × 40 × 20 cm above a 70°-dipping
 rigid basement fault (Fig. 1). The
 experiments are run completely water-
 saturated to allow deformation of wet
 clay and cohesionless sand. The base-
 ment fault moves at 20 to 120mmh−1
 to a maximum offset of 60mm. We use
 quartz sand with grain size between 0.1
 to 0.4mm and an illite-rich clay with a
 water content between 28 and 55 wt.%.
 Water content of the clay is used to con-
 trol its shear strength and state of con-
 solidation. We defined three strength
 classes: soft clay (up to 55 wt% wa-
 ter), normal clay (around 40wt% wa-
 ter) and firm clay (less than 30wt% wa-
 ter). Soft and firm clay represent under-
 consolidated and overconsolidated end-
 members, respectively. The normal clay
 is in equilibrium with the overburden
 load. In a number of experiments 0.25–
 10 wt% ‘Portland Cement’ was added
 to the clay to make it strong and brit-
 tle. Material properties were carefully
 characterized by a series of geotechnical
 measurements.
 Our sandbox experiments are governed
 by two important boundary conditions.
 The basement is a well-defined plate
 acting as a rigid guide compared to the
 properties of the sediment. The inter-
 face on top of the model material is
 either water or aluminum plates pre-
 cut along the kinematically ideal plane.
 These two points of discontinuity in the
 boundary velocity field initiate a de-
 formation band curving away from the
 kinematically preferred plane. On the
 other hand, the boundary plates force
 the two deformation bands to join into
 a continuous zone. Results of ten series
 of experiments will be presented. We
 systematically studied the effect of clay
 strength, thickness, number and posi-
 1
Schmatz et al. TSK 11 Göttingen 2006
 tion of the clay layers and thickness of
 the cover sand.
 Around 300 high resolution digital im-
 ages for each experiment were processed
 into time-lapse movies and analyzed us-
 ing PIV (Particle Image Velocimetry).
 PIV calculates the displacement field
 (Fig. 3) at the scale of individual sand
 grains, and allows calculation of all com-
 ponents of the incremental strain field.
 The main effects of the different param-
 eters are as follows:
 1. absence of the top pre-cut plate
 makes the fault zone much wider,
 and renders the formation of a con-
 tinuous clay gouge more difficult.
 2. in models containing soft clay and a
 pre-cut top plate, a continuous clay
 smear is formed irrespective of the
 details of geometry, and deforma-
 tion approaches simple shear with
 mechanical mixing of sand and clay
 (Fig. 2)
 3. lateral clay injection is a rare pro-
 cess in our experiments, even for
 soft clay;
 4. stiff clay behaves in a brittle fash-
 ion, fault motion is associated with
 rotation of rigid blocks and no
 continuous clay gouge is formed
 (Fig. 2);
 5. for the same amount of clay in the
 sequence the presence of many thin
 layers prefers the formation of a
 continuous clay gouge.
 High resolution PIV shows details of the
 initial phase of deformation, when elas-
 tic strain is overprinted by plastic strain
 and localization (Fig. 3). This shows
 that the structure of the later defor-
 mation bands is already present in the
 models after the first increment of de-
 formation.
 The validity of statistical methods to
 predict fault seal by a clay gouge (SGR
 — Shale Gouge Ratio, CSP — Clay
 Smear Potential) was examined criti-
 cally. Our results clearly show that
 these methods can be unreliable if the
 effects of clay properties and geometry
 are not included in the analysis.
 Figure 2: Image sections of experimental results. Images 1–3 show three stages (16, 24
 and 50mm offset) of deformation of sequence with two cemented clay layers. Images 4–6
 show the same stages for setup with four soft clay layers.
 2
TSK 11 Göttingen 2006 Schmatz et al.
 References
 Fulljames JR, Zijerveld LJJ, et al. (1997) Fault
 seal processes: systematic analysis of fault
 seals over geological and production time
 scales, In Moeller-Pedersen, & Koester AG,
 Hydrocarbon Seals, NPF special publication
 7: 51-59
 Horsfield, WT (1977) An experimental ap-
 proach to basement-controlled faulting. Ge-
 ologie en Mijnbouw 56(4): 363–370
 Van der Zee W, Urai JL & Richard PD (2003)
 Lateral clay injection into normal faults.
 GeoArabia 8(3): 501–522
 Van der Zee, W. & Urai, JL (2005) Processes
 of normal fault evolution in a siliciclastic se-
 quence: a case study from Miri, Sarawak,
 Malaysia. Journal of Structural Geology
 27(12): 2281–2300
 Yielding G, Freeman B, et al. (1997) Quanti-
 tative Fault Seal Prediction. AAPG Bulletin
 81(6): 897–917
 Figure 3: PIV-results for a water-saturated sand experiment. Vector field points to elastic
 behaviour in the initial phase followed by plastic strain with localization of a fault zone
 (last image).
 3