Rupture termination controlled by tuned stress heterogeneity on a 6-m-long laboratory rock fault

Understanding where and how rupture terminates on a fault is crucial because it controls earthquake magnitude and associated damage. But the in situ stress state, which is one of the key parameters governing rupture dynamics, is not directly measurable on natural faults. Laboratory experiments therefore provide an essential approach for investigating rupture termination (e.g., Bayart et al., 2016, 2018; Ke et al., 2018). Previous studies showed that termination can often be interpreted using an energy balance at the rupture tip within linear elastic fracture mechanics (LEFM), while also suggesting that incorporating additional processes such as long-tailed weakening or rate-dependent friction may improve the description (e.g., Paglialunga et al., 2022; Brener and Bouchbinder, 2021). In order to deepen our understanding of rupture dynamics, including how they terminate, it should be efficient to conduct a systematic investigation that controls the stress state and resulting rupture behavior in the laboratory. From this point of view, we have started a large-scale rock friction experiment. In our experiments, two metagabbro specimens are stacked vertically within the experimental frame. The contacting nominal area is 6.0 m long by 0.5 m wide. Six hydraulic jacks apply normal load to the upper block, and a single hydraulic jack applies shear load to the lower block, which is supported on low-friction rollers. Strain gauge arrays along the fault measure local shear stress every 130 mm at a sampling rate of 1 MHz. In experiment GB01-051, we first imposed 5 mm of shear displacement under a macroscopic normal stress of 2.8 MPa, generating repeated stick-slip events that nucleated at either the leading or trailing edge. We then gradually reduced the normal load on one of the normal jacks on the trailing-edge side while maintaining the shear load. This procedure produced clear nucleation near the unloaded jack followed by a full rupture across the entire fault. After restoring the loads to near-critical conditions, we repeated the procedure at the leading-edge side to generate fault ruptures. In a subsequent trailing-edge attempt, however, rupture terminated approximately halfway along the fault, despite a similar macroscopic stress level. Local stress measurements indicate that previous ruptures reduced the shear stress on the leading-edge side, lowering the available energy release rate for propagation and promoting termination. These results demonstrate that rupture initiation and termination can be manipulated through the evolving stress heterogeneity. We also estimated a lower bound on fracture energy from the measured stress drop using LEFM. Accounting for uncertainty in the termination location, the inferred value ranges from 0.032 to 0.29 J/m², consistent with prior experiments on the same rock type (Xu et al., 2019). Ongoing work will further quantify how controlled stress heterogeneity governs rupture termination.