Progressive rock failure under different loading conditions – sound and vision

Catastrophic failure is the end result of a progression of damage towards brittle failure on a variety of system scales in the Earth. However, the factors controlling this evolution, and the relationship between deformation and the resulting earthquake hazard, are not well constrained.  In particular, induced seismicity is a growing cause of concern in the engineering required for the net-zero carbon transition, including subsurface storage of carbon and geothermal energy production, and mining for critical metals. Here we address the question of how to optimize operational controls to minimize induced micro-seismicity in a ‘scale-model’ laboratory experiment where we can simultaneously image the underlying damage using acoustic emissions (sound) and x-rays (vision). We confirm that using continuous servo-control to maintain a constant acoustic emission event rate slows down deformation compared to standard constant strain rate loading, and demonstrate that it also suppresses micro-seismic events of all sizes, including extreme events, and reduces the proportion of seismic to total strain. We develop a new model to explain these observations, based on the observed evolution of microstructural damage and the fracture mechanics of subcritical crack growth.  The model is validated with high precision (r~99%) by comparison with the independently-observed stress history and acoustic emission statistics.

Qualitative inspection of comparable grey-scale x-ray volumes between the two experiments (peak stress and post-failure after unloading) showed that at peak stress microcrack damage accumulated initially, in both samples and in the same area of each sample, as localised pore collapse, pore-emanating and Hertzian tensile intra- and trans-granular cracks and pore-emanating shear and tensile inter-granular cracks. These features were mostly similar in length and aperture, although the sample loaded only under constant strain rate showed a few longer and more open cracks. Strain localisation was apparent at the same stage in both samples, but there was some evidence of earlier en-echelon microcrack localisation in the sample loaded under a constant strain rate. Post-failure, microcracks were longer and more open in the sample loaded under a constant strain rate than in the sample loaded under a constant AE event rate. The visible proportion of damaged rock was greater, with a broader shear zone around two to three grains wide (compared with <1-2 grains) and a greater degree of cataclasis throughout. Off-fault microcracking was limited, but there were some trans- and inter-granular microcracks that extended up to four to five grains long in both samples. These were more common in the sample loaded under constant strain rate and tended to be more open. Finally, branching of the fault zone appeared to be more pronounced in the sample loaded under a constant strain rate.

Our results explain the effectiveness of seismic event rate control on seismic hazard mitigation in mining settings, and imply that it may be more effective in managing the risk from induced seismicity in a pre-emptive way than the commonly-applied ‘traffic light’ system, which is based on reacting after the fact to extreme events.