Contact Stress Distribution and Slip Stability on Experimental Faults

The roughness of natural fault surfaces means that faults make contact at discrete, high-stress bumps and other geometrical highs, producing extreme spatial heterogeneity in normal stress on the sliding interface. This heterogeneity plays a crucial role in nucleating and arresting earthquake rupture. However, heterogeneity resulting from contact of rough surfaces has not been systematically tested in laboratory experiments, which have previously been restricted to nominally flat surfaces. We investigate how the spacing of macroscopic contact regions controls slip stability by shearing cement blocks designed and manufactured to have prescribed contact spacing, as well as replicas of a natural fault surface. Arrays of hemispherical bumps were manufactured with initial spacings of 22-235 mm. Experiments were conducted in direct shear at normal loads ranging from 1 to 10 kN, resulting in local contact stresses of ~20-120 MPa.

Across 27 experiments, sliding ranged from stable creep to unstable stick-slip behavior. Instability is controlled by the minimum spacing between adjacent contact regions (λc), which evolves during wear. Faults remain stable when λc is less than the critical nucleation length Lc predicted by fracture mechanics (tens to ~180 mm for our measured G, Dc, σ, and Δf). When λc exceeds Lc, stick-slip initiates regardless of overall friction coefficient or surface type (regular hemisphere arrays, random bumps, or natural fault replicas). Increased contact normal stress also promotes instability by reducing Lc. These findings are corroborated by a case study of a single experiment, in which λc increased abruptly upon the loss of a few individual contacts, resulting in the immediate transition from stable to unstable sliding that occurred precisely as λc crossed Lc, independent of changes in contact radius.

Our results demonstrate that the spacing of high-stress contact patches may significantly influence slip stability on faults. Because this spacing length scale can be directly observed on real fault surfaces, it provides a physically grounded predictor of where rupture can nucleate or arrest across scales from hand samples to fault segments.