Self-healing slip pulses driven by thermal decomposition: Towards identifying dynamic weakening mechanisms in seismic observations

Seismological observations indicate that earthquake ruptures commonly propagate as self-healing slip pulses, with slip duration at any location on the fault being much shorter than the total event duration [Heaton 1990]. Theoretical work has linked these slip pulses to low values of the background driving stress on the fault [Zheng and Rice 1998]. Recent experiments [Han et al. 2007;Brantut et al. 2008] have shown that fault materials may thermally decompose during shear. These endothermic reactions release pore fluid, leading to an increase in pore pressure and a decrease in temperature [Sulem and Famin 2009]. An Arrhenius kinetic controls the reaction rate, and dynamic weakening only occurs when the temperature reaches a critical temperature triggering the reaction. This abrupt change is in sharp contrast with thermal pressurization where the pore pressure increases smoothly with slip. Previous theoretical studies of thermal decomposition have focused on simple mechanical systems with imposed slip rates [Sulem and Famin 2009], or coupling to a spring-slider model [Brantut et al. 2011]. We present the first solutions to couple thermal decomposition with dynamic rupture, extending the model in Garagash [2012] to solve for self-healing slip pulses. For a range of driving stresses there are two possible slip pulses, compared with a single solution for thermal pressurization alone. One solution corresponds to small slip and a low temperature rise that precludes the reaction; the other is a larger slip solution with weakening due to thermal pressurization at the rupture tip, and weakening due to thermal decomposition in the middle of the pulse. A dramatic drop in fault strength accompanies the onset of the reaction, leading to peak slip rates coinciding with the onset of the reaction. For thermal pressurization alone the maximum strain rate always occurs at the rupture tip, and depends sensitively on the driving stress. Thermal decomposition is identified by slower rupture speeds, longer slip duration and more dramatic strength drops. The peak slip rates occur away from the rupture tip, and are insensitive to changes in the driving stress. For deeper events the ambient temperature is higher, causing the reaction to initiate earlier, and the peak slip rate to move towards the rupture tip. Often the total slip in a pulse is linked to a critical slip required to activate the reaction, suggesting a decrease in slip with depth. Our results could also be linked to observed variations in fault zone mineralogy, with different reactions activated on different faults. Since the peak slip rate is achieved at the onset of reaction, maximum pore pressure generation by thermal pressurization coincides with the maximum generation by thermal decomposition, leading to pore pressures exceeding the normal compressive stress on the fault. One possible mechanism to cap the pore pressure is to allow the permeability to increase with the pore pressure [Wibberley and Shimamoto 2003], enhancing healing by hydraulic diffusion at the trailing edge of the pulse. This leads to slip pulses with shorter slip durations and higher rupture velocities.