Modeling 3D dynamic rupture and arrest of spontaneous fluid-induced microearthquake

Understanding the dynamics of microearthquakes is a timely challenge to solve current paradoxes in earthquake mechanics, such as the stress drop and fracture energy scaling with seismic moment. Dynamic modelling of microearthquakes induced by fluid injection is also relevant for studying rupture propagation following a stimulated nucleation. The ERC-Synergy project FEAR (Fault Activation and Earthquake Ruptures) in the Bedretto Underground Laboratory (Swiss Alps) at approximately 1500m depth offers a unique opportunity to investigate fluid-induced micro-events on broadband seismic arrays. In this study, we leverage this opportunity to perform dynamic ruptures caused by fluid injection on a target pre-existing fault (50m x 50m), generating a Mw 1 seismic event. We conduct fully dynamic rupture simulations coupled with seismic wave propagation in 3D using a linear slip-weakening constitutive law, implemented on the supercomputer Leonardo (CINECA) with a multi-GPU distributed system. Stress field and fault geometry are constrained by in-situ characterization, allowing us to minimize the a priori imposed parameters. We investigate the dynamics of rupture propagation and its arrest for a target Mw < 1 induced earthquake with spatially heterogeneous stress drops caused by pore pressure changes and different constitutive parameters (i.e., critical slip-weakening distance, Dc, dynamic friction). We explore different homogenous conditions of frictional parameters, and we show that the spontaneous arrest of a propagating rupture following a dynamic instability is possible in the modeled stress regime by assuming a high fault strength parameter S, that is high ratio between strength excess and dynamic stress drop characterizing the fault before injection. The arrest of rupture propagation in our modeled induced earthquakes depends on the heterogeneity of dynamic parameters caused by the spatially variable effective normal stress, which controls the on-fault spatial increment of fracture energy Gc. Furthermore, in faults with high S values (i.e., low rupturing potential), we find that even minor variations in Dc (from0.45 to 0.6 mm) have a substantial effect on the rupture propagation and on the ultimate size of the earthquakes. Our results show that modest variations of dynamic stress drop determine the rupture mode, distinguishing self-arresting from run-away ruptures. Studying dynamic interactions (stress transfer) among slipping points on the rupturing fault provides insights on the dynamic load and shear stress evolution at the crack tip. The inferred spatial dimension of the cohesive zone in our crack models is roughly ~0.3-0.4m, with a maximum slip of ~0.6cm. Finally, analyzing the radiated synthetic waveforms, we examine the differences in the high-frequency content of simulated waveforms between self-arresting and run-away earthquakes and provide an estimation of the source parameters obtained through the spectral inversion. This estimation is then compared with source parameters of the dynamic forward models. Our results suggest that several features inferred for accelerating dynamic ruptures differ from those observed during rupture deceleration in a self-arresting earthquake caused by the spatial gradients of normal stress and pore-pressure. These results related to rupture arrest integrate those obtained with spatial variations of the initial stress, highlighting the role of the heterogeneities of stress drop and Gc.