T13D-01 13:40h
Dramatic Strain Weakening in Wet and Dry Olivine Aggregates
Evidence of strain localization in olivine aggregates has been observed both in obducted slices of upper mantle and in high stress experiments (Post 1977). Most experimental studies of olivine aggregates have been done in gas apparatus at low P ($<$300 MPa) and thus at low flow stress, where deformation occurs by climb-accommodated dislocation creep (CADC) with little if any strain weakening. However, experimental studies of quartz and feldspar at high P ($>$1 GPa) have shown the existence of a low T, high stress dislocation creep regime where climb is limited and bulging recrystallization leads to extreme strain weakening and localization. Several recent papers have suggested that for olivine aggregates deforming by dislocation creep at high P ($>$1 GPa) at both high ($>$300 MPa) and low ($<$300 MPa) flow stresses, the addition of water causes a switch in the dominant slip system and thus the lattice preferred orientation (LPO), with important consequences for interpreting seismic anisotropy (e.g. Jung and Karato 2001). To explore high stress dislocation creep in olivine at both wet and dry conditions, we have deformed synthetic aggregates in general shear, using a modified molten salt assembly in a Griggs apparatus at P=1.65 GPa, T=1100$\deg$C, and shear strain rates of 10$^{-4}$/s and 5x10$^{-5}$/s. Aggregates were hot pressed from San Carlos olivine powders (10-25 $\mu$m) at 1.65 GPa for 12 hours. Some samples were dried for 48 hours at 900$\deg$C in a CO/CO$_{2}$ atmosphere, whereas for others 0.6 $\mu$l of water were added prior to weld sealing the sample capsule. Results: {\bf (1)} Wet samples undergo significantly more strain weakening (75-50%) before reaching steady state flow stresses than dry samples (40-25%) deformed at the same conditions. TEM microstructures indicate that the dominant deformation mechanism at the peak stress (400-800 MPa) in all samples is recrystallization-accommodated dislocation creep (RADC). As the samples strain weaken, dislocation climb becomes easier; in most samples RADC remains the principal deformation mechanism, although in those that weaken to $<$200 MPa there is a switch to CADC. {\bf (2)} The recrystallized grain size - flow stress data follow one piezometer relation for all conditions, regardless of deformation mechanism or water content. This result contrasts with the results of Jung and Karato (2001) who found that the recrystallized grain sizes increased with high water contents, and with results for quartz and feldspar which show a change in the slope of the recrystallized grain size/stress relationship for RADC. {\bf (3)} Sample LPOs are consistent with slip on (010)[0{\it kl}] (Type D) at high stresses ($>$300 MPa) and on (010)[100] (Type A) at low stresses, even though the amount of water added to the samples is greater than was found by Jung and Karato (2001) to cause a switch to (010)[001] and (100)[001], respectively. These results indicate that strain localization in olivine aggregates should occur at high stresses where deformation occurs by RADC, and that localization will be greater in wet aggregates.
T13D-02 13:55h
The Effect of Large Melt Fraction on the Deformation Behavior of Peridotite: Implications for the Viscosity of Io$'$s Mantle and the Rheologically Critical Melt Fraction
To date, laboratory studies of the rheological properties of partially molten mantle rocks have reached melt fractions of $\phi$ $<$ 0.15, a value much smaller than thought to be appropriate for the asthenosphere of Io where the degree of partial melting may be as large as 40%. Therefore, we have performed a series of high-temperature, triaxial compressive creep experiments on dry, synthetic peridotites in a gas-medium apparatus at a confining pressure of 300 MPa and temperatures from 1500 to 1553 K in order to determine the influence of large amounts of melt (0.15 $<$ $\phi$ $<$ 0.30) on the rheological behavior of partially molten rocks in both the diffusion and dislocation creep regimes. After hot-pressing San Carlos olivine (10 to 50 $\mu$m) plus MORB ($\sim$8 $\mu$m), the melt is homogeneously distributed between grain-size melt pockets at triple junctions and smaller pockets at two-, three- and four-grain junctions. Stress vs strain rate data from samples in the diffusion creep regime (stress exponent $\ital{n}$ = 1) reveal a drop in rock viscosity of several order of magnitude between $\phi$ = 0.25 and $\phi$ = 0.30, indicative of a rheologically critical melt fraction (RCMF). The combined results from experiments in the diffusion and dislocation creep regimes indicate that the flow behavior in both creep regimes is well described by the published flow laws [e.g. $\ital{Hirth & Kohlstedt}$, 2003] with strain rate $\dot{\epsilon}$ $\propto$ exp($\alpha\phi$) and $\alpha$ = 25 for diffusion creep or $\alpha$ = 30 for dislocation creep. By comparing these flow laws to published values of viscosity in models of whole-mantle or thin-asthenosphere convection on Io, we can place constraints on the likely grain size, melt fraction, and differential stress. For convection limited to a thin (10 to 100 km) asthenosphere, the published range for viscosity of $10^{8-12}$ Pa s requires grain sizes on the order of 10 to 100 $\mu$m in diffusion creep, or a differential stress of 5 to 120 MPa in dislocation creep with $\phi$ $\geq$ 0.25 over the temperature range 1500 to 1800 K. In contrast, for convection extending through the whole mantle, the published value for viscosity of $10^{17}$ Pa s necessitates a grain size $\ital{d}$ $\simeq$ 1 to 10 mm in diffusion creep, or differential stresses of $10^{-2}$ $<$ $\sigma$ $<$ 1 MPa in dislocation creep over the same temperature range and melt fraction. The extremely high differential stresses or the fine grain sizes required for asthenosphere convection are unlikely in a rock with the high melt fraction expected in Io$'$s mantle. A comparison of these values for $\ital{d}$, $\dot{\epsilon}$ and $\sigma$ to those for Earth$'$s upper mantle with $\ital{d}$ = 1 mm, $\dot{\epsilon}$ = $10^{-12}$ $s^{-1}$, and $\sigma$ = 0.1 MPa suggests that whole mantle convection is most likely for Io regardless of creep regime.
T13D-03 INVITED 14:10h
Pressure Sensitivity of Olivine Slip Systems: Evidence From High-Pressure Deformation Experiments
The mineral olivine dominates the composition of the Earth's upper mantle and hence controls its mechanical behavior and physical properties such as viscosity and seismic anisotropy. Until recently our understanding of olivine rheology relied on experimental work carried out at high temperature and moderate pressure and on extensive data on naturally deformed mantle rocks outcropping at the Earth's surface. Pressure extrapolation was thus necessary to infer mantle rheology. We have carried out a series of high-pressure deformation experiments on olivine polycrystals at 1300-1400°C and mantle pressures - ranging from 2 to 11 GPa - using the multianvil apparatus and the deformation-DIA (D-DIA). Microstructural investigations by transmission electron microscopy and crystal preferred orientations determined by electron back-scattered diffraction show that [001] slip becomes predominant at high pressure. Further experiments were carried out on single crystals in a D-DIA coupled with x-ray synchrotron radiation to determine in situ the rheological laws corresponding to the activation of [100](010) and [001](010) slips. It is shown that the inversion that is stress sensitive takes place between 3 and 7 GPa in the experiments. Ab-initio calculations of generalized stacking faults in olivine shows that [100] glide induces more dilation than [001] glide, providing a first tentative explanation for this differential pressure sensitivity. [001] slip produces crystal preferred orientations with an extremely low seismic anisotropy. Transition from [100] to [001] slip in olivine may therefore explain the fast decrease in P and S waves anisotropy in the upper-mantle below 250 km .
T13D-04 INVITED 14:25h
Seismological Applications of Laboratory Measurements of Dispersion and Attenuation in Upper-Mantle Materials
Progress in the use of seismic-frequency experimental techniques for the measurement of dispersion and attenuation in melt-free and melt-bearing olivine polycrystals will be reviewed. The generally mild frequency and grain-size sensitivities of the observed attenuation are suggestive of grain-boundary relaxation - particularly grain-boundary sliding for which there is not yet an adequate microphysical theory. Under these circumstances, our approach is to model the high-temperature viscoelastic rheology with an empirically successful creep function. We have recently identified an optimal creep function of the generalised Burgers type that simultaneously represents our shear modulus and dissipation data obtained at oscillation periods of 1-1000 s and temperatures of 1000-1300 C for a suite of four genuinely melt-free olivine polycrystals ranging in mean grain size from 3 to 165 micron. Application of this laboratory-based seismic-frequency model to the conditions of teleseismic wave propagation in the upper mantle requires only modest extrapolation in grain size, temperature and pressure - the latter through an activation volume. The calculated shear wave speeds and attenuation for melt-free olivine for plausible upper-mantle grain sizes and geotherms reproduce many of the first order features of the seismic structure of the oceanic upper mantle. A zone of low wave speed and high attenuation is predicted that becomes progressively less pronounced and deeper with increasing lithospheric age in accord with seismological observations. Similar calculations suggest that distinctive upper-mantle seismic structures for contrasting tectonic provinces within continental regions are attributable mainly to systematic differences in the depth at which the conductive geotherm intersects a common adiabat. The continental geotherms are calculated with literature values for surface heat flow and heat production; comparison with seismological models confirms observations from Canada and South Africa (Jaupart and Mareschal, Lithos, 1999) that surface heat-flow variations are due to crustal heat production with low and near constant heat-flow across the Moho.
T13D-05 INVITED 14:40h
Evolution of plate tectonics, secular cooling of the Earth, and the global water cycle
Plate tectonics is a surface manifestation of solid-state convection in the Earth's mantle. How plate tectonics has evolved through time is thus intimately connected to how the Earth has been cooled by mantle convection. The energetics of plate tecotnics is not simply that of thermal convection, because it is most likely affected by chemical differentiation processes associated with plate tectonics. Combined with the thermal history of the Earth, a recently-developed heat flow scaling law for plate tectonics indicates a gradual increase in the speed of global plate motion since the late Archean. This notion of more sluggish plate tectonics in the past appears to be consistent with the geological record of the Wilson cycle, and also has a potential to resolve several nagging issues in global geodynamics and geochemistry. In this presentation, I will also discuss an intriguing connection between the thermal evolution of the Earth and the history of ocean volume. The new evolution model of plate tectonics may provide a robust estimate on the long-term net water flux from the hydrosphere to the solid Earth.
T13D-06 14:55h
Effects of a Low Viscosity Zone on Plate Tectonics and Mantle Convection in the Terrestrial Planets: A Physical Theory and Numerical Experiments
A variety of geophysical evidence suggests that there may be a relatively thin zone of low viscosity at the base of the Earth's lithosphere, at least in oceanic and active continental regions. Numerical experiments on mantle convection and, more recently, on the self-consistent generation of plates show that the effects of a low viscosity zone (LVZ) are probably very important. An LVZ strongly increases the horizontal lengthscale of convection, and Earth's LVZ may be a key ingredient to the occurrence of plate tectonics. We have developed a simple boundary layer theory that yields considerable physical insight into these effects, and have tested the theory using 2-D numerical simulations. The theory successfully predicts how the preferred horizontal length scale of convection is controlled by the thickness and viscosity contrast of the LVZ (i.e., how Nusselt number varies with length scale.) Surprisingly, the degree of influence an LVZ exerts on the wavelength of convection actually increases with decreasing LVZ thickness, provided that the viscosity contrast is sufficiently strong according to a straightforward criterion that is also derived. An extended version of the theory further predicts that the effects of an LVZ are particularly pronounced if the LVZ is "submerged" beneath a strong lithosphere, a result initially discovered in 2-D numerical simulations and consistent with the apparently profound influence of an LVZ in promoting plate tectonics.
T13D-07 15:10h
A Monte Carlo Inversion for Mantle Viscosity From Post-glacial Rebound and the Influence of 3D Variations
Observations of glacial isostatic adjustment (GIA) of the earth's surface can provide important constraints on mantle viscosity structure. In this study, we investigate how well GIA observations are able to constrain the spherically symmetric (1D) viscosity structure of the earth. We generate synthetic PGR data by calculating the response of an earth model with realistic 3D viscosity. The viscosity model is constructed starting from seismic tomography models. Computation of the earth's response includes realistic glacial loading, gravitationally self-consistent ocean loading via the sea level equation, and the effects of polar wander. The computation is performed with the spherical finite element code CitcomSVE [Zhong et al., 2002]. We also develop a fast spectral technique to solve for GIA, including the sea level equation and polar wander effects, for earth models with 1D (spherically symmetric) viscosity structures. Following a Monte Carlo algorithm, the responses of thousands of 1D viscosity models are computed. Their resulting PGR observables are compared to synthetic measurements generated from the realistic 3D earth (solved by the finite element code). The Monte Carlo method attempts to minimize a measure of misfit which includes the following PGR observables: relative sea level (RSL) change at various locations in North America, $\dot{J_{2}}$, polar wander, and the rate of change of higher order gravity Stokes coefficients (anticipated GRACE data). We find that as we attempt to invert for more than just a few parameters in the 1D model (for example, the viscosity in a few layers), there occur many 1D models of low misfit to the 3D data, and that these models may differ widely between each other. We also find that including GRACE data in the inversion improves the radial resolution of the inversion.
T13D-08 INVITED 15:25h
Combined Study of the 1992 Landers, 1999 Hector Mine, and 2002 Denali Postseismic Deformations: In Search of a Common Lithospheric Rheology
Large earthquakes represent rock deformation experiments in which sudden stress changes trigger observable postseismic surface deformations that can be used to infer rheological properties of the lithosphere. By using geodetic surface measurements as constraints on finite element models, such experiments seek to understand the relative contributions of postseismic viscous flow, poroelastic rebound, and afterslip, the prime candidate processes of postseismic relaxation. The major challenge in such studies is that of non-uniqueness. All three mechanisms can induce similar surface deformations in certain regions making it difficult to distinguish which mechanism is truly dominant. This is especially true if the quantity and quality of observable constraints is limited, as is often the case even for events with reasonable GPS coverage. The effort of finding a unique postseismic solution is also complicated by the fact that formally optimized afterslip models, which lack physical constraints, can usually be found to explain most horizontal GPS data sets. The key toward uniqueness in postseismic solutions lies in improving the number of constraints available for a study and in recognizing key diagnostic features in the spatial or temporal patterns of these data. GPS and InSAR data sets can be combined, but often one of these is not available or of poor quality. One can also take advantage of earthquake sequences, such as the 1992 Landers/1999 Hector Mine sequence, in that a single rheologic model must be able to explain postseismic surface deformations following both events, as their close proximity implies that they share the same lithosphere. However, geodetically observed earthquake sequences are rare. We investigate a new approach to postseismic studies, that of considering multiple, unrelated earthquakes in a single parameter study. We hypothesize that the lithosphere is not as heterogeneous as generally assumed. For example, it is thought that lower crustal flow is dominant in some regions after an earthquake, deeper mantle flow in others. We theorize that such findings are potentially the result of analysis bias, as many postseismic analyses do not consider all potential depths of flow or mechanisms. We suspect that in a robust study one might find that most postseismic responses to earthquakes are controlled by the same mechanism (or combination of mechanisms) operating at consistent depths with similar stress dependence, though varying according to local heat flow, crustal thickness, and lithology. We begin to test this hypothesis by attempting to explain postseismic deformation associated with the 1992 Landers, 1999 Hector Mine, and 2002 Denali earthquakes with similar mechanical models, for example a combination of powerlaw flow in the deeper lithospheric mantle (50 to 100 km deep) and shallow poroelastic rebound. Finding a consistent mechanical model for all three earthquakes may indicate that the lithosphere response relies on the same mechanisms and occurs in a similar mode for different events subject to different tectonic domains. Such a finding would spur further testing by adding additional earthquakes to the study. On the other hand, failure to find a consistent mechanical model to explain all three postseismic events will serve to highlight gross heterogeneities in lithospheric rheology.