The nature of the boundary between the upper and the lower mantle and its influence on convection

In the PREM seismic model, the boundary between the upper and the lower mantle is accepted at a depth of 670 km, where seismic velocities and density increase. However, until recently there was an obvious inconsistency in this model. The density increases abruptly, and the velocities, in addition to the jumps, have also the subsequent zones of increased gradient. The discontinuity between the upper and the lower mantle is related to the transition of olivine from the ringwoodite phase into the mixture of perovskite and magnesiowustite. However, in the pyrolyte model, the transition zone of the upper mantle consists not wholly of olivine, but partly of olivine (60%) and partly of garnet (40%). The latest data of the garnet measurement at high pressures show that it also experiences phase transition, being converted into magnesium perovskite with the impurity of calcium perovskite. In contrast to the sharp transition in olivine (within a depth interval of only 5 km), the transition in garnet is spread over the interval of depths of 660–710 km. In the widely used PREM and AK135 models, this additional transition corresponds to the zone of the increased gradient in seismic velocities, while in the density distribution it is included in the sharp transition of ringwoodite. Thus, the mineralogy data indicate the need for correction of the PREM and AK135 seismic models: the density jump at a depth of 660 km should be reduced by approximately a factor of two, and a subjacent layer with the increased density gradient should be added at the depth interval of 660–710 km. The phase transition in olivine hampers the mantle flows, although in garnet it accelerates them. Therefore, with an allowance for the smaller jump in density, the decelerating effect of the subducting plates, caused by the phase transition in olivine, decreases, and, furthermore, the effect of their acceleration, caused by the phase transition in garnet, is added. The decrease in the density jump by almost a factor of two will lead to essential changes in the results of the majority of recent works addressing the assessment of the deceleration of convection at the upper/lower mantle discontinuity on the basis of the PREM model.