A seismic tomography study [Fukao and Obayashi, 2012] depicted slab images in the circum-Pacific regions, which showed that, although slabs under and near Japan and under the middle part of Southern America stagnate above the 660-km discontinuity, the majority of other slabs stagnate between 660 and 1000 km depths. Another tomographic study [French and Romanowicz, 2015] depicted plume images in the whole Earth, which showed that, vertically low-velocity columns are seen below hotspots in the deep mantle, they become tomographycally invisible or unclear above a depth of 1000 km. These studies imply that mantle viscosity largely changes above a depth of 1000 km. Evaluation of 1-dimensional viscosity distribution by Rudolf et al.  suggested mantle viscosity increases from 660 to 1000 km depth.
Such viscosity increase inferred from geophysical studies should be explained based on mineral physics. Geochemically concluded volumetrical dominance of bridgmanite, absence of first-order phase transitions, absence of seismic anisotropy in the majority of the lower mantle, and very low strain rate of mantle convection lead to the idea that possible viscosity increase in the mid-mantle should be cause by decrease in strain rate of bridgmanite diffusion creep due to change in defect chemistry.
For these reasons, we investigate chemistry, element diffusivity and grain growth rate of bridgmanite under mid-mantle conditions.
In order to investigate defect chemistry, high-quality samples equilibrated under well controlled chemical environments. For this purpose, use of large volume presses is mandatory, but pressure conditions that can be generated by large-volume presses are limited, usually to 26 GPa. However, we have developed experimental technology to practically generate pressures higher than 27 GPa by large-volume presses [Ishii et al., 2016; 2017]. We investigate chemistry and transport properties of bridgmanite under mid-mantle conditions using this technology.