Effects of Fault Roughness and Geometry on Friction

Julia Baumgarte, McGill University
Jamie Kirkpatrick, University of Nevada, Reno
,
,
,
,
,
,

Abstract

The friction of rocks has generally been shown to be 0.85 at normal stresses up to 200 MPa and 0.6 at higher normal stresses (Byerlee, 1978). These values come from experiments run on nominally flat pieces of rock (or "faults"). Even on nominally flat surfaces, it has been shown that surface roughness matters – for example, sand-blasted sawcut surfaces exhibit different friction than polished sawcut surfaces (e.g., Biegel et al., 1992; Marone and Cox, 1994). Numerical simulations reinforce that friction should depend on surface roughness (e.g., Fang and Dunham, 2013; Tal et al., 2020), but to date, lab experiments have not systematically tested the effect of fault roughness on friction.
Faults show power-law scaling of roughness (e.g., Candela et al., 2012) and can be quantified by power spectral density (PSD), calculated from the Fourier Transform of many profiles of the fault surface. In using the Fourier Transform, however, information about the conspicuous streaks and grooves on a fault surface (in other words, the spatial distribution of roughness) is lost. Thus, our previous understanding of fault friction is oversimplified in two nested ways: 1) in omitting the presence of fault roughness, and 2) in not incorporating the non-randomness of fault roughness (i.e., roughness organized into streaks and grooves) into its quantification.
In this work, we designed a suite of experiments to test how these two characteristics of natural faults influence friction. We made cast-cement replicas of three fault case studies (smooth, medium, and rough) along with three synthetic surfaces constructed to match the roughness defined by the PSD but with spatially random roughness. Friction increased systematically with the magnitude of roughness. However, for a given magnitude of roughness, natural fault replicas were consistently weaker than their synthetic counterparts with random roughness. Synthetic surfaces exhibited friction values 15–50% higher than those of natural surfaces with equivalent roughness, except in the smooth case at low normal stress, where values converged. This work shows that at low confining pressure, the magnitude of roughness increases the friction coefficient. Furthermore, the difference between natural and synthetic faults emphasizes that the spatial arrangement of roughness has a measurable effect on friction and that natural fault surfaces are systematically weaker than their random-roughness equivalents.

Session

Experiments of all sorts