Fault core structure affects fault slip during fluid injection: insights from laboratory friction experiments

Natural faults when subjected to stimulation by fluid injection may result in slip acceleration because pore pressure (Pf) increases in the rock volumes inside and surrounding the fault zone leading to reduction of effective normal stress (σn’). Slip mode ranges from aseismic creep to seismic ruptures defining a spectrum of fault-slip behavior. Fault stimulation experiments will be conducted in the Bedretto Underground Laboratory for Geosciences and Geoenergies (BULGG, Switzerland) to understand fault reactivation processes on a target well-identified fault zone, fully instrumented to monitor deformation and seismicity during both fluid injection and fault reactivation. This is envisioned in the ERC-Synergy FEAR (Fault Activation and Earthquake Rupture) project. In BULGG, the target fault zone has both a sub-centimetric fault core containing fault gouge and granite asperities in contact and other fractures in the surrounding rock volume.  Therefore, it becomes important to define the frictional properties and slip mode of both gouges and bare rock surfaces taking advantage of a laboratory controlled experimental environment. Fault stimulation by increasing Pf was simulated in laboratory following an injection protocol suitable for the BULGG fluid stimulation. Experiments were performed on the target fault gouge and on bare rock surfaces made of nearby Rotondo Granite. We employed a rotary shear apparatus (SHIVA) allowing the fluid injection under a controlled shear stress. First, we imposed the stresses measured at depth in the underground laboratory, halved due to apparatus limitations: 7.5 MPa σn’, 7.5 MPa confining pressure and 2.5 MPa Pf. Second, we imposed a slip rate of 10-5 m/s for 0.01 m to have a reference texture. Third, we applied a shear stress so that an equivalent slip tendency of 0.35 (equal to the one measured in the target fault) is achieved (ca. 2.7 MPa) keeping it constant. We then increased stepwise the pore fluid pressure by 0.1 MPa every 150 s. After fault slip initiation, the maximum allowed slip velocity was 0.1 m/s. Between each of the experimental stages, permeability and transmissivity were measured with the gradient or Pf oscillations methods. We show that reactivation occurs at lower Pf in bare rock surfaces (4.7 MPa) with respect to MC fault gouge (6.5 MPa), suggesting that the effective coefficient of friction, the ratio of shear stress and σn’, is larger in gouge (0.58) than in bare rock surfaces (0.49). Moreover, upon the application of last Pf step, reactivation is slower in fault gouge (150 s delay) with respect to bare rock surfaces (10 s delay), consistently with the lower hydraulic transmissivity measured for target fault gouge with respect to bare rock surfaces (i.e., 10-19 vs 10-17 m3). Our experiments also show that creep and dilatancy precede reactivation in fault gouge, whereas reactivation is sudden and not preceded by dilatancy in bare rock surfaces. We suggest that well-oriented and smooth bare rock surfaces might be easily reactivated similarly to what observed for fault gouge during fluid stimulation. Our data and observations will contribute to shed light on the mechanics of faults and induced earthquakes by fluid stimulation experiments.