Constraining the rheology of the lower mantle with the global trend in slab sinking

The rheology of Earth’s lower mantle plays a crucial role in shaping mantle convection and consequently in planetary evolution but is still under debate1. Propositions differ on the rheology of ferropericlase, whether it is stronger2 or weaker than bridgmanite1,3,4, the majority phase, but especially also on the mechanism by which bridgmanite deforms, either by pure diffusion creep1,5-7 or by pure dislocation climb2,8-10, or possibly in combination. Here we put the pure climb creep rheology to practice in mantle convection experiments with a novel focus on the role of atomic self-diffusion which is pivotal for the effectiveness in mantle convection of both diffusion creep and pure climb creep. From flow models that achieve a close fit to the inferred trend in slab sinking11, we introduce new constraints on the coefficient of self-diffusion. From this we show that pure climb creep prevails over diffusion creep in high-stress regions where the lower mantle deforms most strongly, i.e., ambient to sinking slabs and rising plumes, at last providing more clarity on the rheology of the lower mantle. An immediate implication is that the supposed ancient bridgmanite-enrichment below ~1000 km12,13 can only have survived in low-stress regions remote from sinking slabs and rising plumes and is subordinate for the style of mantle convection. The stress-dependence of pure climb creep leads to a dynamic viscosity field, a propensity for localization of flow in high-stress regions and fast plume ascent. Flow speeds are 1-2 cm/a ambient to sinking slabs and rising plumes and sub-cm/a flow elsewhere. In all, this predicts a much different dynamic role of the lower mantle than modelled so far in the investigating of Earth and rocky planets alike.