Linked influences on slab stagnation: Interplay between lower mantle viscosity structure, phase transitions, and plate coupling

Abstract

An endothermic phase transition in mantle material at 660-km depth constitutes a barrier that in most cases prevents the direct penetration of subducted slabs. Seismic tomography shows that subducted material is in many subduction zones trapped at the bottom of the transition zone, just above the 660-km phase boundary. Recent tomographic models however also report subducted material that penetrates to the shallow lower mantle, and there it is observed to flatten at about 1000-km depth. Models of slab dynamics that generally assume sharp rheological transition at 660-km depth, however, mostly predict slab stagnation at the bottom of the transition zone. Multiple lines of evidence, including recent experiments, indicate that viscosity may gradually increase in the uppermost ∼300 km of the lower mantle, rather than simply changing abruptly at the upper-lower mantle boundary. Here we present the results of a modeling study focused on the effects of rheological transition between upper and lower mantle material on slab deformation and stagnation. We test the effects of smoothing the viscosity increase over 300 km and shifting it to a depth of 1000 km or even deeper. We show that slab ability to penetrate to the lower mantle is mainly controlled by the trench migration rate, which in turn is affected by crustal viscosity. Coupling between the subducting and overriding plates thus plays a key role in controlling slab penetration to the lower mantle and stagnation in the deep transition zone or shallow lower mantle. Models with strong crust and consequently negligible rollback display penetration to the lower mantle without much hindrance and no stagnation above or below the 660-km interface, regardless of viscosity stratification in the shallow lower mantle. Models with weak crust are characterized by fast rollback, and penetration is very limited as slabs buckle horizontally and flatten above the 660-km boundary. Most interesting from the point of view of shallow lower mantle stagnation are models with intermediate crustal viscosity. Here rollback is efficient, though slower than in weak-crust cases. Horizontally lying slab segments are trapped in the transition zone if the sharp viscosity increase occurs at 660 km, but shifting the viscosity increase to 1000 km depth allows for efficient sinking of the flat-lying part and results in temporary stagnation below the upper-lower mantle boundary at about 1000 km depth.