Frictional behavior and constitutive modeling of simulated fault gouge

This paper presents an investigation of the factional properties and stability of frictional sliding for simulated fault gouge. In these experiments we sheared gouge layers (quartz sand) under saturated drained conditions and at constant normal stress (50–190 MPa) between either rough steel surfaces or Westerly granite surfaces in a triaxial apparatus. Surface roughness (60 to 320 grit) and gouge layer thickness (0–4.0 mm) were varied in the experiments with granite samples. Porosity ϕ was monitored continuously during shear. Our measurements indicate that granular gouge exhibits strain hardening and net compaction for shear strains γ less than 0.5–1.0. For γ > 0.5–1.0, sliding occurs at approximately constant shear stress and net compaction from one load/unload cycle to the next ceases. Dilatancy occurs at 1/3 to 1/2 the shear stress required for sliding and d2ϕ/dγ2 becomes negative at about the peak stress in a given loading cycle, indicating the onset of shear localization. Oblique shear bands appear in the layers at γ = 1.3–1.5. Experiments with an initial gouge layer exhibit velocity strengthening (the coefficient of friction increases with slip velocity), and initially bare granite surfaces exhibit velocity weakening. The magnitude of velocity strengthening varies inversely with normal stress and directly with gouge thickness and surface roughness. In the gouge experiments the dilatancy rate dϕ/dγ also varies with slip rate. Using a simple energy balance to relate volume change and frictional resistance, we find quantitative agreement between the measured change in dilatancy rate and friction following changes in slip rate. This indicates that velocity strengthening within granular gouge is the result of dilatancy. The slip rate dependence of dϕ/dγ increases with gouge thickness and surface roughness, in agreement with the friction data. Our data therefore suggest that slip within unconsolidated granular material, such as some natural fault gouges, is inherently stable. The results thus provide an explanation for (1) the tendency of gouge accumulation to stabilize slip in laboratory samples, and (2) the tendency for aseismic slip within shallow (< 3–5 km) unconsolidated fault gouge and within unconsolidated sediments such as shallow alluvium and accretionary prisms.