[1] Si and Mg self-diffusion coefficients were measured simultaneously in single crystals of MgSiO 3 perovskite under lower mantle conditions. There is little difference in Si volume diffusivity measured directly using single crystals (this study) and those retrieved from experiments with polycrystals (earlier studies). This agreement between studies establishes the reliability of Si diffusion coefficients measured in perovskite. Within the uncertainties of our measurements, no anisotropy in the diffusion of either Si or Mg could be resolved. À11 m 2 /s for Mg, DH = 308 kJ/mol for Si, and 305 kJ/mol for Mg. Mg diffusivity in MgSiO 3 perovskite is distinctly lower than those measured in olivine, wadsleyite, and ringwoodite. We find that Mg has very similar diffusivity to Si in perovskite. As a consequence, the rheological properties of the lower mantle may be controlled by the coupled motion of Si and Mg. A point defect-based model is discussed that may account for the diffusion behavior of Si and Mg in MgSiO 3 perovskite. Our data indicate that, within realistic ranges of temperature, grain size, and state of stress, both diffusion creep as well as dislocation creep may be observed in the lower mantle.

Homogeneous single crystals of synthetic monticellite with the composition Ca 0.88 Mg 1.12 SiO 4 (Mtc I) were annealed in a piston-cylinder apparatus at temperatures between 1000 and 1200• C , pressures of 1.0-1.4 GPa, for run durations from 10 min to 24 h and applying bulk water contents ranging from 0.0 to 0.5 wt% of the total charge. At these conditions, Mtc I breaks down to a fine-grained, symplectic intergrowth. Thereby, two types of symplectites are produced: a first symplectite type (Sy I) is represented by an aggregate of rod-shaped forsterite immersed in a matrix of monticellite with end-member composition (Mtc II), and a second symplectite type (Sy II) takes the form of a lamellar merwinite-forsterite intergrowth. Both symplectites may form simultaneously, where the formation of Sy I is favoured by the presence of water. Sy I is metastable with respect to Sy II and is successively replaced by the latter. For both symplectite types, the characteristic spacing of the symplectite phases is independent of run duration and is only weeakly influenced by the water content, but it is strongly temperature dependent. It varies from about 400 nm at 1000• C to 1200 nm at 1100• C in Sy I, and from 300 nm at 1000 • C to 700 nm at 1200• C in Sy II. A thermodynamic analysis reveals that the temperature dependence of the characteristic spacing of the symplectite phases is due to a relatively high activation energy for chemical segregation by diffusion within the reaction front as compared to the activation energy for interface reactions at the reaction front. The temperature dependence of the characteristic lamellar spacing and the temperature-time dependence of overall reaction progress have potential for applications in geo-thermometry and geo-speedometry.