Microphysical Model Predictions of Fault Restrengthening Under Room‐Humidity and Hydrothermal Conditions: From Logarithmic to Power‐Law Healing

Abstract
The maximum fault strength and rate of interseismic fault strengthening (“healing”) are of great interest to earthquake hazard assessment studies, as they directly relate to event magnitude and recurrence time. Previous laboratory studies have revealed two distinct frictional healing behaviors, referred to as Dieterich‐type and non‐Dieterich‐type healing. These are characterized by, respectively, log‐linear and power‐law increase in the strength change with time. To date, there is no physical explanation for the frictional behavior of fault gouges that unifies these seemingly inconsistent observations. Using a microphysical friction model previously developed for granular fault gouges, we investigate fault strengthening analytically and numerically under boundary conditions corresponding to laboratory slide‐hold‐slide tests. We find that both types of healing can be explained by considering the difference in grain‐contact creep rheology at short‐ and long time‐scales. Under hydrothermal conditions favorable for pressure solution creep, healing exhibits a power‐law evolution with hold time, with an exponent of ~1/3, and an “apparent” cutoff time (α) of hundreds of seconds. Under room‐humidity conditions, where grain‐contact deformation exhibits only a weak strain‐rate dependence, the predicted healing also exhibits a power‐law dependence on hold time but it can be approximated by a log‐linear relation with α of a few seconds. We derive analytical expressions for frictional healing parameters (i.e. healing rate, cutoff time, and maximum healing), of which the predictions are consistent with numerical implementation of the model. Finally, we apply the microphysical model to small fault patches on a natural carbonate fault and interpret the restrengthening during seismic cycles.
Funding Information
  • Agence Nationale de la Recherche (ANR‐15‐IDEX‐01)
  • European Research Council (335915)