Previous attempts to deduce the stress distribution in the bending lithosphere near a consuming plate margin have relied on the observed bathymetry and an assumed constitutive relation for lithospheric behaviour, e g . perfectly elastic, viscous/perfectly plastic, or elastic perfectly plastic. From the point of view of rock mechanics, each of these approximations fails to describe one or more of several basic phenomena, including brittle failure of rock, temperature dependence of elasticity, and temperature and/or strain rate dependence of ductile behaviour. In order to formulate a more realistic constitutive relation, a limiting yield strength curve, which is primarily a function of temperature, is constructed from data from brittle failure and ductile flow experiments. The moments which can be supported by plates with this constitutive behaviour are compared to the moments calculated from bathymetric profiles. The comparison indicates that moments required by the bathymetric data are consistent with moments supported by plates with experimentally determined constitutive laws as extrapolated to geologically reasonable temperatures and strain rates. The stresses developed in such models are required to reach values greater than 100 MPat in the depth range 25-45 km. Geotherms necessary for strength curves consistent with moments calculated from the bathymetric data match those derived from heat flow data for the Aleutian, Bonin, Mariana and Tonga trenches. Of the trenches studied, only the geotherm inferred from the Kuril trench data is significantly different, perhaps implying that the Kuril plate is weaker than the others. The strength curves show that as a first approximation it is better to assume that bending moment is independent of curvature of the plate than to assume that bending moment and curvature are linearly related.

The variation of hardness with temperature was measured for olivine on a number of crystal faces by the Vickers diamond pyramid technique (up to 800øC) and by a mutual indentation technique (for temperatures up to 1500øC). A comparative review of hardness data and compressive creep measurements obtained under large confining pressures confirms the hypothesis of Rice [1971] that single-crystal hardness measurements, corrected for elastic effects, can be correlated to the fully ductile yielding of a polycrystal by intragranular dislocation mechanisms, including dislocation climb and glide. The computed differential yield stresses, o (in gigapascals), which empirically correspond to a bility problems attendant with high temperatures • = • [1 + in (• tan •)] (4) tion scheme designed to reduce chemical incompati-• is the angle between the surface of the cone and the test surface. As the parameter E/o in-Hardness of Olivine becomes much smaller than the strain accommodated by ductile flow; in the limit of vanishingly small elastic strain the constraint factor

We summarize the progress made in providing experimental verification for the deformation map of polycrystalline olivine published by Stocker & Ashby in 1973 ( Rev. Geophys . 11, 391). Porosity-free polycrystalline deformation data, applicable to the mantle, were found to be obtainable only from high-pressure deformation studies. Combination of the results of such studies with hardness measurements and single crystal deformation studies on olivine provides narrow constraints on the flow of olivine resulting from dislocation mechanisms from room temperature to the melting point along a band of experimentally accessible strain rates. A good fit is obtained combining a Dorn law above 2 kbar differential stress, e/s -1 =5.7x10 11 ex{-128kcal/mol/RT (1-cr 1 -cr 2 /85000) 2 }, with a power law below 2 kbar, e = 70(cr 1 —cr 3 ) exp{— 122(kcal/mol)/RT}, where stress is measured in bars ( l bar = 10 5 Pa). Indirect data on a mechanism phenomenologically resembling the Goble creep regime are now available from two sources. The observed strain rates are only slightly faster than those predicted by Stocker & Ashby (1973). The ' wet’ data, previously believed to show hydrolytic weakening, are found to fall within this Coble field. The asthenosphere is still expected to deform by the dislocation mechanism summarized by the two formulae given above, but higher stress deformation within the lithosphere is almost certainly dominated by this Coble creep regime once dynamic recrystallization sets in.

A total of 41 selectively oriented single crystals of olivine (F092) were deformed under uniaxial stresses of 100-1800 bars in the temperature range 1150 0 -1600 0 C. Under uniform stress no strain inhomogeneities were observed. For all orientations both strain rate and dislocation structure stabilized after 1-2% strain and remained stable to the highest strains achieved (40%). For all orientations, creep could be represented by a power law of the form E = A0 3.6 where A varied with orientation by a factor of 50. Shape change and crystallographic rotation data for some orientations could only be accounted for by a substantial dislocation climb contribution to the strain rate. All specimens were decorated to show the dislocation structure.

Gem quality single crystals of dry olivine were plastically deformed in compression at differential stresses between 50 and 1500 bars, with zero confining pressure, and at temperatures between 1428° and 1650°C. When these new data are combined with existing creep data for dry olivine‐bearing rock, a flow law of the form ε˙ = ƒ(σ) exp (−Q/RT) is obtained, whereƒ(σ) is an empirical function that is not simply proportional to σn for constant n and Q is equal to 125 ± 5 kcal/mol. The flow law covers the stress range 50–10,000 bars and the temperature range 1100°–1700°C and is consistent with all published data to 1 decade in strain rate.