Considering velocity-dependent cohesion in fault sliding stability

In laboratory studies, the tendency for earthquake nucleation on major plate-boundary fault zones is typically evaluated by measuring the change in shear strength upon controlled step changes in driving velocity.  In such laboratory shear experiments simulating fault sliding, it is typical to neglect the cohesion and assume that all the measured shear strength is due to frictional resistance.  Therefore, the results of such experiments are evaluated in terms of a friction coefficient calculated as the ratio of the shear strength to effective normal stress, where the “velocity-dependent friction” is quantified by the parameter a-b.  However, previous work has shown that the sliding cohesion is not necessarily negligible, especially for water-saturated samples rich in clay-minerals.  This comes from recent experiments have shown that, using a single-direct type shear geometry, the cohesion can be directly measured as a peak shear strength with zero applied normal load (zero effective normal stress).  The technique can also be used for samples that have accumulated slip, thus providing a measure of the cohesion that exists during sliding or “sliding cohesion”.

Here, I use measurements of sliding cohesion to test the assumption that velocity-dependent fault strength is completely controlled by friction, and determine if cohesion plays a significant role.  For these tests I use water-saturated powdered illite-rich Rochester shale, a material that consistently exhibits velocity-strengthening behavior.  The velocity-dependent strength is first obtained with a series of standard 3-fold step increases in driving velocity in the range 0.1-100 μm/s under 10 MPa effective normal stress.  The sliding cohesion is then measured in a series of experiments in which the samples were sheared for 5 mm under 10 MPa effective stress, the normal stress subsequently removed, and then sheared at each of the velocities used in the velocity-step test.  From these tests, the velocity-dependent cohesion is calculated by substituting the cohesion for shear strength and calculating an equivalent “a-b” value that can be subtracted from the standard a-b value measured from the velocity step tests. 

Preliminary results show that velocity-dependent cohesion is of the same order as a-b values, and accounts for up to about a third of the measured a-b values.  The percentage of strength change related to cohesion rather than friction decreases as a function of increasing driving velocity.  Although cohesion is a significant proportion of the velocity-dependent strength changes, removing the velocity-dependence of cohesion is insufficient to cause negative a-b values.  However, this result can also be affected by the choice parameters used in the modeling technique that extracts the a-b values and must therefore be evaluated carefully.  The magnitude of velocity-dependent cohesion suggests that it may represent a signification proportion of velocity-dependent sliding strength.