Inferring damage state and evolution with increasing stress using direct and coda wave velocity measurements in faulted and intact granite samples

A better understanding of damage accumulation before dynamic failure events in geological material is essential to improve seismic hazard assessment. Previous research has demonstrated the sensitivity of seismic velocities to variations in crack geometry, with established evidence indicating that initial crack closure induces rapid changes in velocity. Our study extends these findings by investigating velocity changes by applying coda wave interferometry (CWI). We use an array of 16 piezoceramic transducers to send and record ultrasonic pulses and to determine changes in seismic velocity on intact and faulted Westerly granite samples. Velocity changes are determined from CWI and direct phase arrivals. This study consists of three sets of experiments designed to characterize variations in seismic velocity under various initial and boundary conditions. The first set of experiments tracks velocity changes during hydrostatic compression from 2 and 191 MPa in intact Westerly granite samples. The second set of experiments focuses on saw-cut samples with different roughness and examines the effects of confining pressure increase from 2 to 120 MPa. The dynamic formation of a fracture and the preceding damage accumulation is the focus of the third type of experiment, during which we fractured an initially intact rock sample by increasing the differential stress up to 780 MPa while keeping the sample confined at 75 MPa. The tests show that: (i) The velocity change for rough saw cut samples suggests that the changes in bulk material properties have a more pronounced influence than fault surface apertures or roughness. (ii) Seismic velocities demonstrate higher sensitivity to damage accumulation under increasing differential stress than macroscopic measurements. Axial stress measured by an external load cell deviates from linearity around two-third through the experiment at a stress level of 290 MPa higher than during the initial drop in seismic velocities. (iii) Direct waves exhibit strong anisotropy with increasing differential stress and accumulating damage before rock fracture. Coda waves, on the other hand, effectively average over elastic wave propagation for both fast and slow directions, and the resulting velocity estimates show little evidence for anisotropy. The results demonstrate the sensitivity of seismic velocity to damage evolution at various boundary conditions and progressive microcrack generation with long lead times before dynamic fracture.