Variation in bridgmanite grain size accounts for the mid-mantle viscosity jump.

A viscosity jump of one to two orders of magnitude in the lower mantle of Earth at 800-1,200-km depth is inferred from geoid inversions and slab-subducting speeds. This jump is known as the mid-mantle viscosity jump1,2. The mid-mantle viscosity jump is a key component of lower-mantle dynamics and evolution because it decelerates slab subduction3, accelerates plume ascent4 and inhibits chemical mixing5. However, because phase transitions of the main lower-mantle minerals do not occur at this depth, the origin of the viscosity jump remains unknown. Here we show that bridgmanite-enriched rocks in the deep lower mantle have a grain size that is more than one order of magnitude larger and a viscosity that is at least one order of magnitude higher than those of the overlying pyrolitic rocks. This contrast is sufficient to explain the mid-mantle viscosity jump1,2. The rapid growth in bridgmanite-enriched rocks at the early stage of the history of Earth and the resulting high viscosity account for their preservation against mantle convection5-7. The high Mg:Si ratio of the upper mantle relative to chondrites8, the anomalous 142Nd:144Nd, 182W:184W and 3He:4He isotopic ratios in hot-spot magmas9,10, the plume deflection4 and slab stagnation in the mid-mantle3 as well as the sparse observations of seismic anisotropy11,12 can be explained by the long-term preservation of bridgmanite-enriched rocks in the deep lower mantle as promoted by their fast grain growth.

Basaltic reservoirs in the Earth's mantle transition zone.

The formation and preservation of compositional heterogeneities inside the Earth affect mantle convection patterns globally and control the long-term evolution of geochemical reservoirs. However, the distribution, nature, and size of reservoirs in the Earth's mantle are poorly constrained. Here, we invert measurements of travel times and amplitudes of seismic waves interacting with mineralogical phase transitions at 400-700-km depth to obtain global probabilistic maps of temperature and bulk composition. We find large basalt-rich pools (up to 60% basalt fraction) surrounding the Pacific Ocean, which we relate to the segregation of oceanic crust from slabs that have been subducted since the Mesozoic. Segregation of oceanic crust from initially cold and stiff slabs may be facilitated by the presence of a weak hydrated layer in the slab or by weakening upon mineralogical transition due to grain-size reduction.

Global observations of reflectors in the mid-mantle with implications for mantle structure and dynamics.

Seismic tomography indicates that flow is commonly deflected in the mid-mantle. However, without a candidate mineral phase change, causative mechanisms remain controversial. Deflection of flow has been linked to radial changes in viscosity and/or composition, but a lack of global observations precludes comprehensive tests by seismically detectable features. Here we perform a systematic global-scale interrogation of mid-mantle seismic reflectors with lateral size 500-2000 km and depths 800-1300 km. Reflectors are detected globally with variable depth, lateral extent and seismic polarity and identify three distinct seismic domains in the mid-mantle. Near-absence of reflectors in seismically fast regions may relate to dominantly subvertical heterogeneous slab material or small impedance contrasts. Seismically slow thermochemical piles beneath the Pacific generate numerous reflections. Large reflectors at multiple depths within neutral regions possibly signify a compositional or textural transition, potentially linked to long-term slab stagnation. This variety of reflector properties indicates widespread compositional heterogeneity at mid-mantle depths.

Compositional mantle layering revealed by slab stagnation at ~1000-km depth.

Improved constraints on lower-mantle composition are fundamental to understand the accretion, differentiation, and thermochemical evolution of our planet. Cosmochemical arguments indicate that lower-mantle rocks may be enriched in Si relative to upper-mantle pyrolite, whereas seismic tomography images suggest whole-mantle convection and hence appear to imply efficient mantle mixing. This study reconciles cosmochemical and geophysical constraints using the stagnation of some slab segments at ~1000-km depth as the key observation. Through numerical modeling of subduction, we show that lower-mantle enrichment in intrinsically dense basaltic lithologies can render slabs neutrally buoyant in the uppermost lower mantle. Slab stagnation (at depths of ~660 and ~1000 km) and unimpeded slab sinking to great depths can coexist if the basalt fraction is ~8% higher in the lower mantle than in the upper mantle, equivalent to a lower-mantle Mg/Si of ~1.18. Global-scale geodynamic models demonstrate that such a moderate compositional gradient across the mantle can persist can in the presence of whole-mantle convection.