Fault roughness as a driver of rupture complexity in fluid-induced microearthquakes

Subsurface fluid-injection can induce microearthquakes whose rupture complexity and arrest mechanisms remain poorly understood, limiting our ability to assess the seismic hazard associated with geo-energy applications. We investigate how fault roughness modulates the response of fluid-pressurized faults using an ensemble of 1395 3D dynamic rupture simulations on self-similar rough faults, varying the maximum roughness wavelength, roughness amplitude, and dynamic friction. We show that rupture propagation and arrest are governed by a competition between the large-scale heterogeneity imposed by the pore-pressure gradient and the local shear and normal stress heterogeneity generated by variations in fault geometry. Our results demonstrate that dynamic rupture complexity emerges already at the scale of Mw~1 fluid-induced microearthquakes, with fault roughness promoting acceleration and deceleration phases and, in some cases, pulse-like rupture behavior. The maximum roughness wavelength controls rupture complexity by determining the effectiveness of asperities and barriers, thereby constraining rupture growth and maximum attainable magnitude. Simulated magnitudes are inherently bimodal, with fault roughness and friction dictate the critical transition from failed nucleations to self-arresting and runaway ruptures. Furthermore, smoother faults, characterized by lower roughness amplitudes, progressively build up higher loading conditions at the rupture front, favoring coherent ruptures exhibiting predominantly crack-like propagation. Fault roughness affects stress redistribution and produces spatial variations in the local energy balance driving rupture front propagation, evidenced by spatial variations in cohesive zone size and rupture velocity. These findings indicate that reliable estimates of the maximum size of induced earthquakes must account for initial stress conditions as well as fault geometric complexity.