Earthquakes are among the most destructive of natural hazards known, and lead to considerable loss of life and property world-wide every year. In spite of the large impact that earthquakes have on society, little is known about the mechanisms through which earthquakes nucleate and dynamic ruptures propagate. This hampers earthquake hazard assessments and forecasting attempts, which are at present largely based on statistical inferences rather than physical principles.
Laboratory studies of fault friction may contribute to the accuracy of earthquake hazard assessments by detailed investigation of the mechanisms involved in fault rock deformation. However, the spatial- and temporal scales of a typical laboratory experiment are dwarfed by those of natural faults and earthquakes, and so upscaling of the laboratory results is required. This is often performed through a mathematical framework called rate-and-state friction. This framework is empirical in nature, and so predictions made through the rate-and-state friction framework must be interpreted with great care.
In this thesis, alternative means of upscaling are explored that are based on physical principles and have a micro-mechanical origin, as opposed to being empirical. This is done by conducting laboratory friction experiments, and interpreting the results in terms of micro-scale processes that constitute deformation, such as frictional grain sliding (granular flow) and pressure solution creep. Constitutive relations describing these micro-scale processes are subsequently implemented into numerical models, and the model predictions are tested against laboratory results. Finally, the microphysically based numerical models are used to extrapolate the laboratory observations to natural scales and conditions, and to make predictions pertaining to the natural seismic cycle.
The outcomes of this thesis demonstrate that microphysically based models can explain a wide variety of laboratory and natural observations, and are a suitable means for the extrapolation of laboratory results to nature. This offers new opportunities for future studies of earthquake hazard and risk by providing a physical basis for making long-term predictions.