Discrete element modeling of triggered slip in faults with granular gouge: application to dynamic earthquake triggering

Recent seismological observations based on new, more sensitive instrumentation show that seismic waves radiated from large earthquakes can trigger other earthquakes globally. This phenomenon is called dynamic earthquake triggering and is well-documented for over 30 of the largest earthquakes worldwide. Granular materials are at the core of mature earthquake faults and play a key role in fault triggering by exhibiting a rich nonlinear response to external perturbations. The stick-slip dynamics in sheared granular layers is analogous to the seismic cycle for earthquake fault systems. In this research effort, we characterize the macroscopic scale statistics and the grain-scale mechanisms of triggered slip in sheared granular layers. We model the granular fault gouge using three dimensional discrete element method simulations. The modeled granular system is put into stick-slip dynamics by applying a conning pressure and a shear load. The dynamic triggering is simulated by perturbing the spontaneous stick-slip dynamics using an external vibration applied to the boundary of the layer. The influences of the triggering consist in a frictional weakening during the vibration interval, a clock advance of the next expected large slip event and long term effects in the form of suppression and recovery of the energy released from the granular layer. Our study suggests that above a critical amplitude, vibration causes a significant clock advance of large slip events. We link this clock advance to a major decline in the slipping contact ratio as well as a decrease in shear modulus and weakening of the granular gouge layer. We also observe that shear vibration is less effective in perturbing the stick-slip dynamics of the granular layer. Our study suggests that in order to have an effective triggering, the input vibration must also explore the granular layer at length scales about or less than the average grain size. The energy suppression and the subsequent recovery and increased activity period in our model explain the abundance of observations that support the hypothesis of the delayed dynamically triggered earthquakes, as well as provide clues for improving the methods for identifying triggered earthquakes. The results of our simulations also provide a physical basis for the methods suggesting time intervals of increased hazard following the occurrence of major earthquakes. (author)