LAUSR

A popular idea is that accretion of sediment at a subduction zone commonly leads to the formation of a subduction channel, which is envisioned as a narrow zone located above a subducting plate and filled with vigorously circulating accreted sediment and exotic blocks. The circulation can be viewed as a forced convection, with downward flow in the lower part of the channel due to entrainment by the subducting plate, and a ‘backflow’ in the upper part of the channel. The backflow is often cited as an explanation for the exhumation of high-pressure/low-temperature metamorphic rocks from depths of 30 to 50 km. Previous analyses of this problem have mainly focused on the restricted case where the walls bounding the flow are artificially held fixed and rigid. A key question is if this configuration can be sustained on a geologically relevant timescale. We address this question using a coupled pair of corner flows. The pro-corner accounts for accretion and deformation directly above the subducting plate, and the retro-corner corresponds to a deformable region in the overlying plate. The two corners share a medial boundary, which is fully coupled but is otherwise free to rotate and deform. Our results indicate that the maintenance of a stable circulating flow in a narrow pro-corner (<15°) requires an unusually large viscosity ratio, μretro/μpro > 103. For lower viscosity ratios, the medial boundary would rotate rearwards, converting the initially narrow pro-corner into an obtuse geometry. For a stable narrow corner, we show that the backflow within the corner is caused by downward convergence of the incoming flow and an associated downward increase in dynamic pressure, which reaches a maximum at the corner point. The total pressure is thus expected to be much greater than predicted using a lithostatic gradient, which means that estimates of depth from metamorphic pressure would have to be adjusted accordingly. In addition, we show that the velocity fields associated with a forced corner flow and a buoyancy-assisted channel flow are nearly identical. As such, structural geology studies are not sufficient to distinguish between these two processes.