Two-dimensional, viscous flow modeling of roll-back subduction : numerical investigation into the role of slab density in subduction dynamics

Abstract

Observations of retreating subduction systems in the Mediterranean region suggest the density of subducting lithosphere is dynamically related to trench retreat rate and upper-plate deformation. Most numerical and analog studies of retreating subduction systems have not explored the effects of lithospheric density variations on subduction processes. This study is a preliminary effort to construct a two-dimensional, viscous flow model of "roll-back" subduction to explicitly examine how slab density influences retreat rate, mantle flow, and slab geometry. For a given lithosphere-mantle density contrast, the model computes the evolution of a viscous, thermal slab using a finite element code for incompressible convection (ConMan). Imposed velocity boundary conditions guide lithospheric material into a uniformly weak "subduction zone" and out into the mantle below, generating stable, asymmetric subduction. Slabs driven faster than the "intrinsic" (dynamically consistent), steady-state retreat rate of the system (vr) are characteristically arcuate, pushed upward from the base of the mantle layer by strong horizontal "return flow" beneath the descending lithosphere. Slabs driven slower than vr are sigmoidal: the slabs steepen at depth, where vertical buoyancy forces overcome lateral viscous forces set up by weak surface velocities. The diagnostic behaviors of slabs driven faster and slower than vr define a set of qualitative criteria (slab geometry, mantle flow patterns) for converging on the consistent, steady-state retreat rate of the system. For slab-mantle density contrast [delta]p = 198 kg/m³ (defined as the density difference between lithosphere at surface of the system and mantle material at the base of the system), vr ~~ 16 mm/yr. The slab is roughly planar, with 500 dip. For [delta]p = 168 kg/m³, vr is slightly slower (14 mm/yr), and steady-state slab geometry is nearly identical (moderately-dipping planar surface). It is found that the angle at which lithospheric material is forced into the mantle does not significantly affect either steady-state retreat rate or slab geometry.