Nanostructures as an indicator for deformation dynamics

Grain-size reduction with amorphization or melting as its extreme forms plays a crucial role in fault-zones dynamics, e.g., the nucleation or arrest of earthquakes. Previous experiments have been mostly conducted on powdered samples and structural investigations of experimentally generated fault-gouge material provide contrasting results when it comes to the initiation of melting during fault slip. In the present study, we deformed four intact Westerly granite samples, to decipher whether there is a correlation between failure mode, i.e., controlled, self-stabilised, or dynamic, and grain-size reduction within the developing fault gouge. Controlled failure took place over several hours, self-stabilised failure occurred within a few seconds and dynamic failure lasted less than a second. To test the influence of aqueous fluids on the grain-size evolution within fault gouges, two runs were performed on samples, dynamically failing either in the presence or absence of pore fluids. All samples were deformed at the same effective pressure of 40 MPa and displacements along the newly created faults were with 1.2 to 2.0 mm in a similar range. We investigated the microstructures of each sample using a scanning electron microscope (SEM) and cut two focused-ion beam (FIB) sections per sample from selected areas, located within the fault gouges, to analyse their nanostructures using a transmission electron microscope (TEM). At low magnification at the SEM, no striking differences between the different fault gouges are visible. Features resembling cooling cracks become apparent at the highest magnification at the SEM and are only found in the samples that failed dynamically. Major differences between the samples are only obvious when comparing their nanostructures using TEM imaging. In the high-resolution TEM images as well as with the aid of selected area electron diffraction (SAED), we observe a clear correlation between failure mode or rupture speed and grain-size reduction, with an increase in amorphous material as rupture speed increases. Regardless of the availability of fluids, the samples that underwent dynamic failure reveal similar nanostructures. Both exhibit flow structures created by amorphous material. We believe that latter is the result of melting as we find numerous structural and chemical evidence for melting, e.g., euhedral magnetite crystals of a few tens of nanometer with adjacent depletion halos. Such indicators for melting are absent in samples that failed in a controlled or self-stabilised manner, highlighting the importance of rupture speed on fault gouge melting.