Mantle convection with strong subduction zones

Because mantle viscosity is temperature-dependent, cold subducting lithosph ere should be strong, which implies that the rapid, localized deformation a ssociated with subduction should strongly resist plate motions. Due to comp utational constraints. the deformation of a subducting plate cannot be accu rately resolved in mantle-scale convection models, so its effect on convect ion is difficult to investigate. We have developed a new method for impleme nting subduction that parametrizes the deformation of the oceanic lithosphe re within a small region of a finite element grid. By imposing: velocity bo undary conditions in the vicinity of the subduction zone, we enforce a geom etry for subduction, producing a slab with a realistic thermal structure. T o make the model dynamically consistent, we specify a rate for subduction t hat balances the energy budget for convection, which includes an expression for the energy needed to deform the oceanic lithosphere as it subducts. Th is expression is determined here from a local model of bending for a strong viscous lithosphere. By implementing subduction in this way, we have demon strated convection with plates and slabs that resemble those observed on Ea rth, but in which up to 40 per cent of the mantle's total convective resist ance is associated with deformation occurring within the subduction zone. T his additional resistance slows plate velocities by nearly a factor of two compared to models with a weak slab. For sufficiently strong lithosphere, t he bending deformation slows surface plates sufficiently that they no longe r actively participate in global-scale convection, which occurs instead ben eath a 'sluggish lid'. By introducing a low-viscosity asthenosphere beneath the oceanic plate, we demonstrate that small-scale convection at the base of oceanic lithosphere may limit plate thickness, and thus the resistance t o bending, and cause plate velocities to depend on the strength of the bend ing lithosphere rather than on the viscosity of the underlying mantle. For a cooling Earth, the effective lithosphere viscosity should be nearly const ant, but the mantle viscosity should increase with time. Thus, subduction-r esisted convection should produce nearly constant plate velocities and heat flow over time? which has implications for the Earth's thermal evolution. We estimate that this style of convection should apply if the effective vis cosity of the bending lithosphere is greater than about 10(23) Pa s, but on ly if some mechanism, such as small-scale convection, prevents the bending resistance from stopping plates altogether. Such a mechanism could be funda mental to plate tectonics and Earth's thermal history.