DUCTILE CREEP, COMPACTION, AND RATE AND STATE-DEPENDENT FRICTION WITHIN MAJOR FAULT ZONES

The shear traction on major strike-slip faults during earthquakes is much lower than that expected on a frictionally sliding surface in equilibrium with hydrostatic pressure. The low shear traction is explained if the fluid pressure at the time of the earthquake is much greater than hydrostatic pressure. Ductile creep within mostly sealed fault zones compacts the matrix and thus increases fluid pressure between earthquakes. Frictional dilatancy during earthquakes decreases fluid pressure below hydrostatic, and over the earthquake cycle, the fault zone is in long-term equilibrium with the country rock. This ductile mechanism is formally unified with rate and state theory for time-dependent friction when the difference between a critical porosity where the rock loses all strength and the actual porosity of cracks is used as a state variable. This choice is justified by percolation theory of mostly broken lattices. Time-dependent behavior associated with changes in normal traction in the laboratory is explained by the formalism. Instability (earthquakes) sometimes occurs in the numerical experiments. However, fairly small amounts of frictional dilatancy during initial frictional creep decrease fluid pressure and preclude unstable sliding. Two coupled mechanisms for producing dilatancy on faults once an instability is well underway are evident. (1) Expansion of pore fluids associated with frictional heating increases fluid pressure offsetting the effects of increased pore volume during earthquakes. There is some tendency for pore volume increase to balance fluid expansion so that fluid pressure stays relatively constant. (2) Production of isolated voids that do not immediately decrease fluid pressure throughout the fault zone during earthquakes can occur to the extent that the fault zone is not significantly strengthened. Although the extent of both processes is constrained by energy considerations, the variation of fluid pressure during earthquakes is not yet well enough understood to predict stress drop from observable material properties.