What controls the onset of stable fault creep at the bottom of the seismogenic zone?

Abstract:

The termination of seismicity at the bottom of the upper continental
crust is usually attributed to the onset of plasticity in quartz at
temperature of about 300 deg. C, and a transition from the
velocity-weakening to velocity-strengthening behavior of typical
crustal rocks such as granite. Laboratory measurements of the rate
parameter (a-b) of rate-state friction using a triaxial apparatus show
a transition from negative (a-b) (velocity-weakening behavior,
conditionally unstable slip) to positive (a-b) (velocity-strengthening
behavior, stable creep) for Westerly granite at temperatures above
300-350 deg. C (Stesky, 1978; Blanpied et al., 1991). Assuming a
reasonable geotherm, this transition is expected to occur at depth of
12-15 km, in good agreement with the depth distribution of earthquakes
in California. The temperature dependence of the (a-b) parameter is
believed to control the thickness of the seismogenic layer, as
earthquakes would not be able to nucleate under velocity-strengthening
conditions. However, Mitchell et al. (2013) reported
velocity-weakening behavior of Westerly granite at temperatures up to
450 deg. C in laboratory tests using a heated direct shear apparatus
and solid samples. Here we present results from recent experiments
done on thick gouge at both dry and hydrated conditions at
temperatures up to 600 deg. C and normal stresses up to 40 MPa. We
find that the rate dependence parameter (a-b) progressively decreases
with temperature over the entire temperature range, leading to more
unstable slip at higher temperature. This tendency is enhanced in the
presence of water. The new experimental data suggest that friction of
common crustal rocks can be velocity-weakening over a wider depth
range than previously believed, in particular under dry conditions or
low water fugacity. This may help explain deep crustal seismicity in
areas of active extension (e.g., Baikal and East Africa\cite rift
zones) and convergence (e.g., Himalayas and Andes), as well as
non-volcanic tremor associated with deep roots of active faults (e.g.,
San Andreas fault in central California).