Rubin Home

Department of Geosciences 
319 Guyot Hall 
Princeton University 
Princeton, NJ 08544

Phone: (609) 258-1506
E-Mail:

Episodic Aseismic Slip in Subduction Zones

One of the most exciting developments in seismology in the last decade has been the discovery of “episodic slow slip events” and associated “deep non-volcanic tremor” in subduction zones worldwide.  The episodic slip manifests itself as the reversal in displacement of geodetic stations above the subducting slab.  The example on the left below is taken from Vancouver Island, where the generally eastward displacement of station ALB is interrupted every 14 months or so by a several-week period of lesser westward drift.  Inversion of displacement data from many stations shows that the slow events can extend for 100 kilometers in the dip direction and 300 kilometers along strike, with extremely small slips of a few cm. The events occur near the transition from the shallow locked portion of the subduction interface (the velocity-weakening region that slips in large earthquakes) to the steadily-creeping velocity-strengthening region below.  Slip rates are only a couple orders of magnitude larger than the plate convergence rate, and propagation rates are of order 10 km/day, far below elastic wave speeds.  Intriguingly, these slip events (in Cascadia, Japan, and elsewhere) are accompanied by an extended-duration, low amplitude “tremor” signal.  The tremor example on the right below comes from the Cholame section of the San Andreas fault, discovered after the subduction tremor but demonstrating that this “new” phenomenon is not restricted to those settings.

Over the last year I have been applying what we have learned about earthquake nucleation to the interpretation of episodic slow slip events.  For faults in elastic media, where the nucleation zone is free to choose its own length, there are multiple length scales relevant to nucleation.  Slow slip events can arise if the fault is large enough to nucleate an event but too small for that event to reach dynamic instability.  The disparity between these length scales, and hence the range of fault sizes capable of hosting slow events, increases as the fault becomes closer to velocity-neutral, and can be quite large for the “aging” evolution law for fault friction.  However, we have shown that this capability of the aging law flies in the face of existing lab friction data.  For the “slip” evolution law, which has fallen out of favor recently but does a much better job of matching the relevant lab data, the range of fault lengths (from the shallow locked region down to the velocity-strengthening region) capable of hosting slow events is too small to credibly explain why such events are so common worldwide.

These difficulties with the standard laws has led me to explore other friction laws, including (with Paul Segall at Stanford University) the possibility that the slow events are stabilized by dilatancy of fault gouge (the inelastic increase in porosity with increasing slip speed), followed by pore fluid diffusion back into the fault zone.  Preliminary results suggest that for reasonable parameters, this mechanism can give rise to slow events that appear similar to those observed over an exceedingly large range of fault lengths (Figures 2–3).

Figure 2. Snapshots of sip speed, normalized by the plate velocity V_pl, during 2 simulated slow slip events.  The fault is locked for x/L_b<0, velocity-weakening between x/L_b=0 and the vertical dashed line, velocity-strengthening to the right of that line, and forced to slip at the plate rate far to the right.  Time progresses from the red curve to the blue to the black.  The thick horizontal black bars show the minimum size of the velocity-weakening region that allows slow events to occur, for the adopted value of the rate-and state friction parameter a/b (a/b<1 is velocity-weakening; a/b=1 is velocity-neutral, and a/b>1 is velocity-strengthening).  Top panel: A slip law simulation with the largest stable velocity-weakening region for a/b=0.95.  Thus for the slip law the range of fault lengths capable of hosting slow slip events is less than a factor of 2 even when the fault is this close to velocity-neutral.  Bottom panel:  A slip law simulation where the fault gouge undergoes inelastic dilation and pore pressure reduction as the velocity increases, with a simplified (“membrane diffusion”) model of pore pressure recovery. Even though a/b is farther from velocity-neutral, slow slip events are stable over a very larger range of fault lengths (note the size of the horizontal black bar).

Figure 3. Log slip speed (red curve) and slip (blue) at the center of the velocity-weakening region for the simulation from which the bottom panel in Figure 2 is extracted.  For plausible parameters, events with slips of roughly 1 cm occur roughly yearly, placing this simulation in a range that is consistent with observations.

We have also learned from these models that the low ratios of propagation speed to slip speed and low ratio of slip to length are all consistent with very large pore fluid pressures (close to lithostatic, or effective normal stresses of only a few MPa) in the source region.  This is qualitatively consistent with independent seismic evidence, and also with expected dehydration reactions in the subducting plate.  It is also consistent with an increased role for inelastic dilation of the gouge and pore pressure reduction.

Currently I am trying to apply these same concepts to understanding the mysterious tremor that seems to always accompany slow slip.  This tremor is thought to represent faster, more localized slip on the same fault, and it should be interpretable in terms of the same constitutive laws.

Related Publications:

Rubin, A.M., Episodic slow slip events and rate-and-state friction, submitted to J. Geophys. Res., 2008.