Segregation of subducted oceanic crust in the convecting mantle

Subducted oceanic crust, transformed into dense mineral assemblages at high pressure, may gravitationally segregate at the bottom of the convecting mantle, for example, the D″ layer. Here it could be stored for a long enough time to develop an “enriched” isotopic signature, before it is recycled in mantle plumes and hence control the geochemical character of hot-spot basalts. We study both the geodynamical and geochemical aspects of this hypothesis in two-dimensional numerical convection models, in which plate motion is imposed by a velocity boundary condition. About 250,000 tracer particles are used to identify the basalt fraction in the oceanic crustmantle system. High-, average-, and low-tracer densities indicate basalt or eclogite, peridotite, and harzburgite, respectively. The tracers are negatively buoyant to account for the density differences between the rock types. At surface divergence zones, crust formation is simulated by extracting tracers and transferring them into a thin layer at the surface. The tracers carry a certain amount of the relevant nuclides of the U-Pb and Sm-Nd systems, which are fractionated between the basalt tracers and a second species of residue tracers during crust formation. Using reasonable parameter values, we find that of the order of 1/6 of the subducted crust accumulates in pools at the bottom, which reside underneath thermal plumes. After 3.6 Ga, the [207Pb]/[204Pb], [206Pb]/[204Pb], and [143Nd]/[ 144Nd] ratios in various parts of the model cover the observed HIMU-MORB range. A systematic study of the influence of control parameters on the results indicates that the amount of segregation and the diversity of isotope ratios (1) increases strongly with Rρ, the ratio of chemical to thermal buoyancy; (2) decreases moderately with the Rayleigh number Ra, where Ra = 106 is the highest value that we employed; (3) increases strongly with the degree of temperature dependence of the viscosity; and (4) is not very sensitive to the partitioning between internal and bottom heating. The largest uncertainty in applying our model results to the Earth lies in the lack of accurate density data for conditions at the core-mantle boundary. We conclude that, if our density estimates are correct, segregation and reentrainment of subducted crust is of fundamental importance for the dynamics and chemistry of mantle plumes.