A suite of true triaxial tests were performed on Castlegate sandstone to assess the influence of the intermediate principal stress on mechanical response and failure. Five independent deviatoric stress states were employed, for which the intermediate principal stress ranged from equal to minimum compression (axisymmetric compression) to maximum compression (axisymmetric extension). For each deviatoric stress state, five constant mean stress tests were conducted, covering mean stresses ranging from brittle to ductile failure. At low mean stresses, shear bands formed, and the peak stress required to induce failure decreased with increasing intermediate principal stress. Thus, failure at low mean stresses depends on the third invariant of deviatoric stress. Shear bands formed under all deviatoric stress states and over a wide range of mean stresses. The band angle (defined as the angle between the band normal and the direction of maximum compression) decreased with increasing mean stress. There was no clear trend in band angle with respect to intermediate principal stress; however, a small trend would be obscured by data scatter due to specimen variability. At higher mean stresses, no localization was observed. The upper bound mean stress at which shear localization occurred increased with increasing intermediate principal stress. Therefore, the mean stress that demarcates the brittle‐ductile transition depends on the third invariant of deviatoric stress.

For assessing energy‐related activities in the subsurface, it is important to investigate the impact of the spatial variability and anisotropy on the geomechanical behavior of shale. The Brazilian test, an indirect tensile‐splitting method, is performed in this work, and the evolution of strain field is obtained using digital image correlation. Experimental results show the significant impact of local heterogeneity and lamination on the crack pattern characteristics. For numerical simulations, a phase field method is used to simulate the brittle fracture behavior under various Brazilian test conditions. In this study, shale is assumed to consist of two constituents including the stiff and soft layers to which the same toughness but different elastic moduli are assigned. Microstructural heterogeneity is simplified to represent mesoscale (e.g., millimeter scale) features such as layer orientation, thickness, volume fraction, and defects. The effect of these structural attributes on the onset, propagation, and coalescence of cracks is explored. The simulation results show that spatial heterogeneity and material anisotropy highly affect crack patterns and effective fracture toughness, and the elastic contrast of two constituents significantly alters the effective toughness. However, the complex crack patterns observed in the experiments cannot completely be accounted for by either an isotropic or transversely isotropic effective medium approach. This implies that cracks developed in the layered system may coalesce in complicated ways depending on the local heterogeneity, and the interaction mechanisms between the cracks using two‐constituent systems may explain the wide range of effective toughness of shale reported in the literature.

Despite the wide application of hydraulic fracturing in enhanced geothermal system (EGS) development and unconventional hydrocarbon production, our understanding of the many factors affecting the propagation of hydraulic fractures has relied on circumstantial evidence. This is partly due to the scarcity of direct observations in the subsurface. The EGS Collab project attempts to address this issue in the context of EGS research by performing intermediate-scale (∼10 m) hydraulic stimulation experiments in a thoroughly characterized and heavily instrumented underground testbed. This paper analyzes the data collected from the first suite of hydraulic fracturing tests in this testbed, consisting of seven stimulation episodes. High-quality microseismic data delineate five planar features very clearly. Combining fracture-wellbore intersection observations from distributed temperature sensing (DTS) and visual observations from an open-hole well, as well as prior in situ stress measurements, we conclude with high certainty that the four larger planes were hydraulic fractures. The growth of the hydraulic fractures was temporarily halted by a prominent, open natural fracture in the testbed but they eventually crossed and slightly reoriented under continued stimulation. Mineral-filled (i.e. healed) natural fractures, though prevalent in the testbed, did not have perceivable effects on hydraulic fracture propagation. The high quality, mutually corroborating datasets allowed conclusions to be drawn with high confidence and attests to the advantage of intermediate-scale experiments in subsurface research.