Fault heterogeneities control self-arrested vs. run-away induced ruptures: implications for deep enhanced geothermal systems

For traffic light systems, empirical volume hypotheses imply that the hazard of induced seismicity can be managed prescriptively by simply maintaining the net injection volume below a certain threshold value. The limitations of the volume hypothesis are shown by some recent, complex triggered events (e.g. 2017 Pohang M_w 5.5) during deep enhanced geothermal systems, which were able to grow into a run-away rupture larger than predicted magnitudes (e.g. M_w 3.7).

Here we investigate the role played by in situ heterogeneities of stress conditions (e.g. fault healing) and fault rheology (e.g. off-fault damage) on controlling rupture behaviour. We performed slide-hold-slide triaxial rupture tests (hold time 1­-3 10³ s) on biomaterial samples (Westerly granite and Carrara marble) at confining pressure Pc = 5 - 50 MPa. The experimental setup allows failure to initiate on a saw cut portion of the granite, and successively propagate through an initially intact portion of the sample (either marble or granite). Linear dynamic strain gauges and piezoelectric sensors were positioned along the fault plane to acquire strain and acoustic data in the low (Hz) and high frequency (MHz) domains.

Our results show that increasing the amount of fault healing (by increasing the hold duration) induces a transition from stable to unstable failure, which produces stick-slip events with large slip and stress drop, and shorter rise time. A condition which will favour the grow of induced events into large, runaway ruptures. On the other hand, the brittle-ductile transition (Pc > 30 MPa) in the Carrara marble induces a larger amount of off-fault damage that causes the switch from fast stick-slip to slow-slip behaviour in the composite samples. A condition which will increase dissipated fracture energy and favour self-arrested ruptures.

Overall, our results show that the size of threshold magnitude between self-arrested and run-away induced earthquakes is also controlled by key rupture parameters, such as initial stress and dissipated fracture energy. Our findings highlight the limitations of approaches that rely exclusively on production parameters, such as injected/extracted volumes of fluids, and neglect the role played by fault heterogeneities, such as initial stress state, geometry and rheology.