Weakening mechanisms efficiency during seismic slip.

Abstract:

While sliding at seismic slip-rates of ca. 1 m/s, a natural fault undergoes an abrupt decrease of its strength called dynamic weakening. Processes related to fast heating and weakening of the asperities, efficient crystal-plastic yielding, frictional melting or thermal pressurization of fluid saturated faults have been invoked to explain pronounced velocity weakening during initial slip-rate acceleration. However, few weakening mechanisms can explain the experimental evidence of fault strength recovery during the slip-rate deceleration at the end of seismic slip.

Here we present a compilation of ca. 100 experiments performed in the rotary shear apparatus SHIVA, at varying effective normal stress (σ’_n=5-30 MPa), slip-rates (V=0.1-6.5 m/s) and fluid pressures (from room-humidity condition, P_f=0, to P_f=10 MPa) on Carrara marble, microgabbro and Westerly granite. We analyze the efficiency of each weakening mechanism (flash-heating, flash-melting, thermal pressurization, diffusion creep) using a 2D Finite Difference Method model with a norm-based optimization of the parameters. The model reproduces the experimental evolution of the shear stress (τ) with slip which includes: 1) initial phase during slip acceleration when the apparent friction μ=τ/σ’_n decays from the static friction coefficient (~0.6-0.85) to its dynamic value (~0.1), 2) intermediate phase when the sample is sliding at a constant slip-rate and μ remains constant and low (∼0.1), 3) final phase during slip deceleration when the experimental fault recovers part of the initial strength.

Our results show that during the initial phase, flash heating is the most efficient weakening mechanism independently of rock lithology and experimental conditions. During the intermediate phase, for Carrara Marble experimental faults, diffusion creep is the most efficient weakening mechanism, even at large P_f pressures (up to 10 MPa). Instead, for microgabbro and Westerly granite, a layer of melt (bulk melting) lubricates the slipping zone. The contribution of thermal pressurization is negligible at all the investigated experimental conditions and for all rock types. Lastly, the healing of the fault during the deceleration stage can be well represented by the diffusion creep model in Carrara marble faults and by cooling of the rock melt in granitoid faults.