Reevaluating plate driving forces from 3-D models of subduction

Subducting lithospheric slabs mechanically attached to tectonic plates provide the main driving force for surface plate motion. Numerical models historically simulate slab dynamics as a 2-D process and further simplify the problem into either a density driven model (no heat transfer) or a corner-flow problem (thermal convection) [Christensen, 2001; Enns et al., (in revision); van Keken, 2003]. Recent 3-D global models of density driven flow incorporating a history of plate motion (Conrad and Lithgow-Bertelloni, 2002) have succussfully ruled out slab "suction" (basal shear traction induced by downward flow of the slabs) as a major driving force, but exact partitioning of the remaining forces acting on the slab remain unconstrained. A survey of trenches around the world reveals that over half of the slabs presently subducted in the upper mantle have a discontinuous edge (either a slab tip on a young slab, or the side edge of a slab with finite width) around which mantle can flow: prime examples being slabs in the Mediterranean and Carribean. However, even slabs with a wide lateral extent (and where a 2-D approximation may seem appropriate), show signs of having 3-D complexity. For example, on the surface Tonga appears relatively symmetric, but when the history of subduction is considered, the slab has a twisted, 3-D structure due to significant eastward retreat of just the northern part of an originally N-S oriented trench edge. Similarly the widest slabs, South American and Kamchatka, show seismic anisotropy attributed to trench parallel mantle flow (Russo and Silver, 1994; Peyton, et al., 2001, respectively), while the Aleutian trench has oblique subduction varying in magnitude from west to east, and medium width Central American slab likely has a slab window allowing 3-D flow (Johnston and Thorkelson, 1997). Recent laboratory experiments of subduction have demonstrated the full complexity of flow occuring in 3-D geometry (Kincaid and Griffiths, 2003; Schellart, 2004), owing to the analog slab having a lateral extent smaller than the width of the box. These experiments clearly show subduction of a finite-width slab will generate a flow of material from behind the slab around both the side edges and under the nose of the slab into the mantle wedge. This rollback induced flow establishes a positive feedback with backward hinge migration on the surface, and has significant consequences for the composition and dynamics of the mantle wedge. Here we present results of 3-D numerical experiments aimed to quantify the partitioning between different forces acting on such a slab. These experiments include a high viscosity slab (relative to background mantle), a high viscosity lower mantle and a computational domain large enough so that the flow induced by subduction of a finite-width slab is not constrained by the side or bottom boundaries. We provide a self-consistent force balance and integrate the forces acting over the different portions of the slab, thereby partitioning such forces into specific components. We quantify the force due to rollback-induced flow, and signify its importance as a driving force relative to the other forces present: a net slab pull force, a force responsible for bending the slab at the subduction hinge, and a resistive force due to shear traction on the upper, lower, and nose (if present) surfaces of the subducted slab.