Relating fault mechanics to fault zone structure

The fault zone of a mature large-displacement fault may be idealized as a nested hierarchical structure consisting of a core of extremely fine grained material surrounded by coarser granulated gouge and breccia which is in turn bordered by fracture-damaged wall rock in which the fracture density decreases with distance to a regional background level. While there are significant variations in the symmetry of this structure, virtually all fault zones have a core of deformed granular rock within which most of the displacement appears to have occurred, often on an extremely narrow prominent slip surface. In many faults, the gouge and breccia layer is missing from one or both sides of the fault zone. For strike-slip faults, this appears to be associated with variations in the lithology of the wall rock. For dip-slip faults it is most likely a consequence of the exhumation of only one wall by the fault motion. In normal faults, the layered structure appears on the footwall while the hanging wall shows almost no damage. For reverse faults it is the hanging wall which contains the layered structure. Mechanisms proposed to explain the formation of these fault zone structures are reviewed with an eye toward whether they shed any light on the earthquake process. Conversely, mechanical models for earthquake nucleation, propagation, and arrest are reviewed to see how fault zone structure might affect these processes. A key parameter in these models is the characteristic displacement D-c required to reduce friction from its static to its dynamic value. In nucleation, D-c determines the size of the smallest earthquake. The observation of magnitude zero earthquakes, implies D-c is on the order of microns, comparable to values measured in the laboratory. For propagation, D-c is closely related to the fracture energy, which is the integrated value of stress times displacement from zero to D-c Dynamic models of large earthquakes find fracture energies on the order of 1 MJ/m(2) and D-c on the order of centimeters. The implication is either that only large earthquakes nucleate on large faults, or that the D-c which controls nucleation is different from that which controls propagation because they reflect different processes. Two such processes that might produce a larger D-c for propagation are the frictional heating of pore fluid and off-fault damage. Finally, models for earthquake arrest based on barriers requiring large values of fracture energy also yield fracture energies on the order of MJ/m(2).