Buoyancy, bending, and seismic visibility in deep slab stagnation

The petrological consequences of deep subhorizontal deflection (" stagnation" ) of subducting slabs should affect both apparent seismic velocity structures and slab morphology. We construct kinematic thermal models of stagnant slabs and perform thermodynamic modeling of the consequent perturbation of high-pressure phase transitions in mantle minerals, focusing upon Japan as our study area. We calculate associated thermo-petrological buoyancy forces and bending moments which (along with other factors such as viscosity variations and rollback dynamics) may contribute to slab deformation. We consider effects of variations in depth of stagnation, post-stagnation dip angle, phase transition sharpness, transition triplication due to multiple intersection of geotherms with phase boundaries, and potential persistence of metastable phases due to kinetic hindrance. We also estimate seismic velocity anomalies, as might be imaged by seismic tomography, and corresponding seismic velocity gradients, as might be imaged by receiver-function analysis. We find that buoyant bending moment gradients of petrological origin at the base of the transition zone may contribute to slab stagnation. Such buoyancy forces vary with the depth at which stagnation occurs, so that slabs may seek an equilibrium slab stagnation depth. Metastable phase bending moment gradients further enhance slab stagnation, but they thermally decay after ∼600-700 km of horizontal travel, potentially allowing stagnant slabs to descend into the lower mantle. Stagnant slabs superimpose zones of negative velocity gradient onto a depressed 660-km seismic discontinuity, affecting the seismological visibility of such features. Seismologically resolvable details should depend upon both stagnation depth and the nature of the imaging technique (travel-time tomography vs. boundary-interaction phases). While seismic tomography appears to yield images of stagnant slabs, discontinuity topography beneath Japan resolved by migrated receiver functions appears to be consistent with slab penetration of the transition zone. However, model slabs which bottom around ∼780-810 km and then bend upwards by a few degrees can match both the tomographic and receiver-function images.