Experiments have shown that the deformation of rocks under high confining pressures cannot be described by terms based on ordinary experience. The writers believe there are only three fundamental macroscopic processes at work: extension fracturing, faulting, and uniform flow. The first involves separation across a plane of no shear normal to the direction of least principal stress and includes tensile fracture as a special case. Faulting involves a shearing displacement and occurs with or without loss of cohesion. It includes shear fracturing as a special case and may occur along a plane inclined at from a few degrees to 45° to the direction of maximum principal stress. From the standpoint of the stress-strain curve, faulting without loss of cohesion is indistinguishable from flow. (The three flow mechanisms—cataclasis, gliding, and recrystallization—cannot always be distinguished either.) Examples of experimental boudins are presented to illustrate these categories of deformation.
The field evidence is that earthquakes are accompanied by shearing displacements and are therefore due to faulting in the general sense. Since stored elastic strain energy is released, there must be at least a momentary and local loss of cohesion. A crack then propagates at close to the speed of sound. For deep-focus earthquakes (down to 700 km) certainly, and most probably even for shallow disturbances (a few tens of kilometers), ordinary Coulomb fracture is impossible. The internal friction of dry rocks under tens or hundreds of thousands of bars pressure would demand impossibly high shearing stresses of many kilobars. The most reasonable mechanism of energy release at great depth is a phase change, and the most probable phase change is melting.
Calculations suggest that, once a crack or a flaw exists, there is ample elastic energy to propagate the crack by shear melting even if the stress difference is only a few tens of bars. It is, of course, inconceivable that an open crack could exist at depth, so that the most baffling problem is the nature of a flaw of the Griffith type under these conditions. Although it is little more than speculation, we suggest that the flaw may be a pocket of already molten rock or of its more volatile constituents. This idea receives some support from the intimate association of earthquake epicenters and zones of volcanic activity.