We are going to construct a new global dynamic model of the mantle and to describe internal dynamic processes that strongly influence the Earth's surface. A principal improvement compared to previous studies should be a comprehensive model of the transition zone (TZ) and determination of its effect on the global mantle convection, dynamic geoid and plate velocities. Another principal step-forward will be a 3D viscosity model of the mantle, the parameters of which will be tuned to provide best fit to the observed fields (geoid, dynamic topography, plate velocities).
Joint inversion of the observed geoid and tomography models of seismic velocities has been commonly used to constrain mantle properties and convection flow. However, most of the employed tomography models have been obtained without considering the effect of the mantle transition zone (TZ). We use the new-generation tomography model of Gu et al. (2003), in which seismic velocity perturbations are estimated together with depth variations of the main TZ discontinuities.
In the inversion of this model with the observed geoid the velocity-to-density scaling factor and density jump at these discontinuities are determined simultaneously. By this, the mantle flow across TZ is defined self-consistently: the undulations of the TZ boundaries suppress or accelerate mantle currents depending on the determined density contrast. For the 410-km discontinuity we obtain the scaling factor and density jump, which are close to mineral physics predictions. Therefore, we conclude that these effects are decoupled in the tomography model. In contrast, the calculated density jump at the 660 discontinuity is approximately 4 times less than the PREM value. We suggest that this is the effect of multiple phase transitions within a depth range of 640-720 km.
Under normal thermal conditions, the post-spinel phase transformation is relatively sharp (~5 km) and clearly visible in seismic models. On the other hand, the transition of majorite garnet to perovskite is much broader (up to ~ 50 km) and therefore hardly detectable. Due to the different sign of the Clapeyron slope, the total gravity effect is drastically decreased. To fit the obtained results, the required value of the Clapeyron slope for the majorite garnet to perovskite transformation should be equal to about +1.7 МPa/К. In the cold zones the same effect might be produced by the transition from ilmenite to perovskite at 610-640 km depth, which is in agreement with multiple reflections revealed in regional seismic studies near the bottom of TZ (Deuss et al., 2005). The estimated amplitude of the mantle flow across TZ is about ± 20 mm/y, which corresponds to the whole-mantle convection scheme.
One of the principal features of the model is the low-viscosity zone right above the 660 discontinuity, so called “notch”, which is required to provide best fit between the observed and calculated geoid. Before, this feature was typically recognized by postglacial rebound studies (e.g. Milne et al., 1998). In fact, it serves like a barrier between the upper and lower mantle, which significantly reduces the dynamic topography induced by lower mantle heterogeneity and, as a result, affects the dynamic geoid.
The calculated geoid better fits to the observed one than the obtained without considering the TZ effect.