LAUSR

Polycrystalline samples composed of either tabular or equiaxed forsterite grains +diopside (5 and 20 vol %) were deformed with a grid etched onto the lateral surface. In Part 1 of this study, we identified grain boundary sliding (GBS) and rigid body-like grain rotation during deformation by diffusion creep where samples with tabular forsterite grains were shown to develop low-index plane grain boundaries that result in crystallographic preferred orientation (CPO). Here we examine how grain rotation depends on the sample strain, grain size, phases, grain shapes, and orientations relative to the compression axis and long axes of tabular forsterite grains. Based on these results, we model grain rotation due to GBS that occurs preferentially along low-index plane boundaries. The model reproduces all of the characteristics of grain rotation and together with the observed grain rotation rates in tabular and equiaxed grain samples, we estimate that low-index plane boundaries have a lower viscosity by a factor of ~3 relative to general grain boundaries, which results in the development of CPO during diffusion creep. The observed constant rotation rate of ~0.4 (radian/strain) in equiaxed-grain samples and in tabular-grain samples deformed to a strain of >0.5 is considered to be a minimum and further, a material-independent rotation rate during diffusion creep, indicating grain rotation as a primary microprocess during diffusion creep. We discuss the possible consequences of GBS-induced grain rotation and CPO development in rock microstructure and the seismic properties of the Earth's mantle.