High-Pressure Creep of Serpentine, Interseismic Deformation, and Initiation of Subduction

See allHide authors and affiliations

Science  21 Dec 2007:
Vol. 318, Issue 5858, pp. 1910-1913
DOI: 10.1126/science.1148494


The supposed low viscosity of serpentine may strongly influence subduction-zone dynamics at all time scales, but until now its role could not be quantified because measurements relevant to intermediate-depth settings were lacking. Deformation experiments on the serpentine antigorite at high pressures and temperatures (1 to 4 gigapascals, 200° to 500°C) showed that the viscosity of serpentine is much lower than that of the major mantle-forming minerals. Regardless of the temperature, low-viscosity serpentinized mantle at the slab surface can localize deformation, impede stress buildup, and limit the downdip propagation of large earthquakes at subduction zones. Antigorite enables viscous relaxation with characteristic times comparable to those of long-term postseismic deformations after large earthquakes and slow earthquakes. Antigorite viscosity is sufficiently low to make serpentinized faults in the oceanic lithosphere a site for subduction initiation.

Subduction zones, in which slabs of oceanic lithosphere sink into the mantle, are active zones where frequent large earthquakes cause considerable human and material damage. Such events are triggered by stress buildup or strain localization, the understanding of which relies on identifying the materials involved and their rheology. On top of slabs of many subduction zones, a layer with low seismic velocity and high Poisson ratio (>0.29) is interpreted as extensively serpentinized mantle material (1, 2), and may accommodate most of the deformation at the slab/mantle wedge interface. Serpentinites form by peridotite hydration either during hydrothermal alteration of the oceanic lithosphere before subduction or by percolation of the fluids released by mineral dehydration within the downgoing slab through the overlying mantle wedge (3). The high-pressure variety of serpentine, antigorite, can remain stable down to ∼180 km depth in cold subduction zones (4). Serpentinites are highly deformed as compared to other exhumed materials in paleosubduction zones (5), which points to their crucial mechanical role. The expected low strength or viscosity of serpentinite has strong seismic implications because it may govern stress buildup and downdip relaxation over the slab surface, which are critical parameters for earthquake triggering and for the downdip extent of major ruptures (6). So far, only viscous relaxation of the anhydrous mantle has been considered a potential trigger of major earthquakes, such as the Tonankai 1944, Nankaido 1946 (7), and Alaska 1964 events (8). Serpentinites also have global geodynamic importance on the time scale of mantle convection because a serpentinite layer may decouple the mantle wedge from the downgoing slab (9). Its presence therefore is a defining condition of the plate tectonic regime on Earth.

The limitations of apparatus have restricted previous high-temperature deformation experiments on serpentinites to pressures below 0.7 GPa (5, 911). Below the antigorite dehydration temperature (600°C), such low confining pressures favor brittle behavior, with deformation being governed by frictional forces, whereas different deformation mechanisms are to be expected at higher pressures (11), as suggested by numerous defects allowing for intracrystalline creep commonly observed in antigorite (12). In the absence of high-pressure data, quantifying the role of serpentinite at long and short time scales in subduction zones has remained beyond reach. We performed in situ measurements (13) of antigorite flow stress using the recently developed deformation-DIA (D-DIA) apparatus coupled with synchrotron x-ray analysis (14) under conditions of low constant strain rates (∼10–4 to 10–6 s–1) and pressure and temperature (P-T) of 1 and 4 GPa and 200° to 500°C, respectively; that is, over most of the antigorite stability field (4, 15). We obtained a stress-strain curve for 14 sets of experimental conditions (tables S1 and S2). Strain values ϵ(t) were measured on synchrotron x-ray radiographs, and differential stress σ was measured from elastic lattice strains on angle-dispersive x-ray diffraction patterns (13, 16). The stress value taken or extrapolated at 15% axial strain was used arbitrarily as a measure of the ultimate flow stress (table S3). Because sample observation shows features consistent with intracrystalline deformation (13), flow stress values were fitted to power-law equations (Table 1), in which the stress exponent depends on the dominant deformation mechanism (dislocation creep, diffusion, etc.), and to an exponential law appropriate for low-temperature creep processes [the Peierls mechanism (13)]. The best fit to the present data, at 1 and 4 GPa, was obtained with a single power-law equation that yielded an activation volume of 3.2 ± 0.7 cm3 mol–1, activation energy of 8.9 ± 5.4 kJ mol–1, and a stress exponent of 3.8 ± 0.8 (Table 1), consistent with deformation by dislocation creep. The decrease of the stress exponent with increasing pressure when fitting data at each pressure independently (13) is consistent with the activation of intracrystalline deformation mechanisms at the expense of frictional grain boundary sliding at low confining pressure.

Table 1.

Preferred fits to power law Embedded Image and exponential law Embedded Image. Standard errors (1σ) are in parentheses at the right of each parameter. A and Ap are material constants, Ea and Ep are activation energies, V* is activation volume, n is a stress exponent, and τ is Peierls stress.

View this table:

The ductile deformation of antigorite observed above 1 GPa complements observations from previous triaxial experiments, showing brittle behavior of serpentinite below 0.7 GPa and a transition toward a distributed semi-brittle deformation up to 1 GPa (5, 11). If controlled by antigorite, the transition from brittle to ductile creep at the slab interface with either the crust or the mantle wedge should depend mainly on depth, while the thermal structure of the subduction zone exerts only minor effects. In the case of high porosity (microcracks) with reduced effective confining pressure, the transition depth may depend indirectly on temperature through the amount of water released by mineral dehydration.

In order to depict further the potential role of antigorite rheology on wedge dynamics, we calculated a strength profile at a constant strain rate along a slab surface (Fig. 1), assuming a P-T profile in a moderately hot subduction zone and considering two extreme cases. In the first model, we assumed a 300-m-thick serpentinite layer formed by hydration above the subducting slab and sheared by 10 cm year–1, corresponding to a strain rate of 10–10 s–1. Such conditions correspond to those of a subduction zone characterized by strong mechanical coupling or fast postseismic deformation due to sparse serpentinization. In the second case, a 10-km-thick serpentinite layer sheared by 1 mm year–1 deforms at a strain rate of 10–14 s–1. Such conditions hold for a subduction zone with a layer of extensively serpentinized mantle decoupled from a slowly downgoing slab. These two end-member models indicate that, regardless of strain rate and subduction-zone setting, antigorite is the only mineral among the major phases in the subducting lithosphere and mantle wedge that is capable of yielding by creep at geophysically relevant strain rates and temperatures below 600°C (Fig. 1). The exception is fine-grained olivine, which may become weaker than antigorite above 2.5 GPa (Fig. 1). Shear instabilities may, therefore, be reconsidered as a possible mechanism for intermediate-depth seismicity, which may either be related to antigorite dehydration producing very fine-grained olivine or occur within fine-grained partly serpentinized peridotites (17).

Fig. 1.

Strength of antigorite (atg) and other major silicates along a slab surface. Strengths are calculated from deformation laws at a strain rate of 10–10 s–1 (thick curves) and 10–14 s–1 (thin curves) along a slab surface P-T profile [profile number i50 from (28)] (upper graph). The low-temperature dislocation glide Peierls mechanism (PM) (29), is the most realistic deformation mechanism for olivine (ol) at these pressure, temperature, and strain-rate conditions. Because the Peierls mechanism cannot be extrapolated to low strain rates, strength can be calculated from the creep law only at 10–14 s–1. At 10–10 s–1, the transition from frictional behavior according to Byerlee's law (thick black line), and antigorite deforming plastically by the Peierls mechanism, occurs between 0.7 and 1 GPa. Antigorite strength remains at least one order of magnitude lower than that of other major mantle and crust minerals regardless of the deformation mechanism and the strain rate, except for very fine-grained (1 μm) olivine deforming by diffusion creep (dif, dotted curve), the strength of which becomes inferior to that of antigorite before its dehydration at strain rates below ∼10–12 s–1. Millimeter-sized olivine deforming by diffusion creep yields much higher strength and plots off the graph. The Peierls mechanism is from (29) and the wet olivine dislocation creep and diffusion creep from (30). di, wet diopside; an, anorthite, dislocation creep laws (31).

These deformation experiments provide an upper bound for serpentinite viscosity, because naturally occurring localizations would induce high strain rates and lower the effective viscosity. The values we calculated for effective serpentinite viscosity, ∼4.1019 Pa·s for a strain rate of 10–13 s–1 (13), are of the same orders of magnitude as those used in current numerical models (18). Serpentine viscosity as determined by us does not vary much with temperature, which precludes substantial shear heating in a low constant strain-rate system. Our flow law predicts that strain rate, hence viscosity as well, depends nonlinearly on stress. This would enhance positive feedbacks between strain and stress variations, as compared to models using linear stress dependence such as Newtonian rheology (18).

Seismologists define three zones downdip along the slab: seismic, transitional (locked during interseismic time), and aseismic. The factors controlling the downdip limit of the seismogenic and locked transitional zones will also govern the downdip propagation of megathrust ruptures, such as the event of 26 December 2004 in Java (19). Because serpentinite has a low viscosity, with little pressure and temperature dependence above 1 GPa, the depth at which nonseismogenic creep is possible is governed exclusively by the extent of the serpentinite layer in the subduction zone. This is consistent with observations in Japan, where shallow depths (a maximum of 30 km) of seismogenic zones are associated with Poisson ratios higher than 0.29 (20), a strong indication of serpentinization, whereas deeper (50 to 70 km) downdip limits coincide with no indices of serpentinization (21). In Sumatra, where no evidence of serpentinization is found, the downdip limit occurs even deeper in the mantle (22).

Because of its low viscosity, serpentine can relax stress at rates comparable to those of postseismic and slow seismic deformations. Using a modified Maxwell body with a nonlinear viscous behavior, subject to a permanent deformation ϵ0 producing an initial stress σ0 = ϵ0E, where E is the Young modulus (in pascals), the characteristic relaxation time τc required to relax half of the initial stress σ0 is Math(1) (23), where Ea and V* are the activation energy and volume, respectively; A and n are material parameters; R is the gas constant; P is the effective confining pressure; and T is the temperature. At temperatures of 200° to 500°C relevant to a slab surface, the relaxation times for antigorite are at least 10 orders of magnitude shorter than those for olivine (Fig. 2). For subduction-zone flow stress estimates up to ∼100 MPa (18), antigorite relaxation times compare well with characteristic times of co- or postseismic surface deformations such as those measured by geodetic measurements for slow slip events, episodic tremor and slip, silent earthquakes, afterslips, and viscous relaxation (Fig. 2). Viscous relaxation of serpentinite therefore accounts for slow-slip events and for slow earthquakes occurring over periods of a few days to 1 year and which follow a scaling law different from that for regular earthquakes (24). These results also suggest that the importance of viscoelastic relaxation processes for triggering large earthquakes in subduction zones over interseismic periods of several years (25, 26) should be reassessed, taking into account the low viscosity of serpentinites measured here. Thus, the triggering of future earthquakes, such as the Tokai event expected in Japan, may depend on serpentinite viscous relaxation (7).

Fig. 2.

Maxwell relaxation time for antigorite rheology and post- and slow seismic deformation time scales. τ1/2 is the time calculated (Eq. 1) to relax half of an initial stress σ0 imposed at t = 0 (for instance, by the displacement field of an earthquake occurring close to a serpentinite body), using the antigorite power-law equation at 2 GPa (table S4) and E = 89 GPa (20). Stress above 20 MPa will be half-relaxed in less than 1 year by antigorite at 500°C (∼40 MPa at 200°C), whereas stress relaxation is not possible in wet olivine deforming by dislocation creep (30) at these temperatures. Olivine diffusion creep (30), not reported here, leads to a constant characteristic relaxation time of ∼107 years. Relaxation of high natural stress by antigorite flow is compatible with the time scales of postseismic deformation after large earthquakes (7, 8, 25, 26) and slowslip events or silent earthquakes (24). ETS, episodic tremor and slip; SSE, slow slip events.

Together with a strong stress dependence, the low viscosity of antigorite at P-T conditions where other minerals have viscosities orders of magnitude higher confirms that serpentinite is an ideal candidate for strain localization within subduction zones. Moreover, antigorite-bearing serpentinites formed deeply in oceanic transform faults and passive margins may constitute decisive weak zones in the oceanic lithosphere, because their viscosity is lower than the critical value of 1020 Pa·s required to initiate subduction (27). The present data quantitatively relate viscous deformation of serpentinites to interseismic deformation and slow earthquakes. They should help improve numerical modeling of seismicity and convection in subduction zones.

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Tables S1 to S3


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article