Numerical simulation of ocean/continent convergent systems : influence of subduction geometry and mantle wedge hydration on crustal recycling

Studies on high-pressure (HP) and ultrahigh-pressure (UHP) rocks exposed in orogenic belts linked to collisional margin show that nappes of oceanic and continental deeply subducted crust can be exhumed to shallow structural levels. In particular, during ocean-continent-type subduction, the crustal material dragged into subduction channel is composed chiefly by ocean and trench sediments, crustal slices belonging to subducting plate [microcontinent (Ring & Layer, 2003) or linked to early continental collision (Chemenda et al., 1995)] or crustal slices tectonically eroded from the overriding plate (ablative subduction) (Tao & O’Connell, 1992, Marotta & Spalla, 2007). Several models have been developed, during last 20 years, to analyse exhumation of subducted crustal material. They can be resume on five main mechanisms: a) crustal-mantle delamination (Chemenda et al., 1995), b) slab break-off (Ernst et al., 1997), c) slab retreat (Ring & Layer, 2003) and roll-back slab (Brun & Faccenna, 2008) and d) decoupling of two main ductile layers (Yamato et al., 2008), in which the exhumation is mainly driven by negative buoyancy and/or faulting and e) subduction-channel flow (Gerya & Stockhert, 2005) in which the exhumation is driven by the upwelling flow developed in low-viscosity mantle wedge. Only channel flow takes into account recyrculation of crustal slices dragged to high depth by ablation in pre-collisional subduction zones. To study the effects of subduction rate, slab dip and mantle rheology changes on channel flow efficiency a parametric analysis is made. We present the results of a set of numerical simulation with different subduction rates, slab dips and mantle rheology represented by dry dunite and dry olivine flow laws. Numerical model predictions are finally compare to some PT paths obtained from ancient and actual subduction zones with different slab dips and convergence velocities. A general good agreement between natural data and model predictions emerges from the comparison: exhumation rates obtained from complete PTt-paths (total exhumation rates) are more compatible with natural rates rather than maximum exhumation rates; the thermal states predicted by ablative subduction simulations with a hydrated mantle wedge justify the natural PT estimates obtained on continental crust units involved in ocean/continent subduction systems. For these reasons, we propose ablative subduction of the upper continental plate linked to hydrated mantle wedge as a good alternative pre-collisional mechanism, with respect to the collisional mechamisms as the slab break-off, slab-retreat and roll-back slab