Using open sidewalls for modelling self-consistent lithosphere subduction dynamics

Subduction modelling in regional model domains, in 2-D or 3-D, is commonly done using closed, vertical boundaries. In this paper we investigate the merits of using open boundaries for 2-D modelling of lithosphere subduction but with implication for 3-D modelling. Open sidewalls allow for lateral in- and outflow consistent with the internal dynamics of the model and may simulate the real-mantle environment of subduction much better than closed boundaries, which induce return flows. Our experiments are focused on using open and closed (free-slip) sidewalls while comparing results for two model aspect ratios of 3:1 and 6:1. Slab buoyancy driven subduction with open boundaries immediately develops into strong rollback with high trench retreat velocities. Mantle asthenosphere flow forced by rollback is predominantly laminar and facilitated by the open boundaries. In contrast, free-slip sidewalls proof restrictive on subduction rollback evolution unless the lithosphere plates are allowed to move away from the sidewalls. This, however, initiates return flows pushing both plates toward the subduction zone speeding up subduction. Increasing the aspect ratio to 6:1 does not change the overall flow pattern when using open sidewalls. Again, in contrast, for free-slip boundaries, the slab evolution does change with respect to the 3:1 aspect ratio and does not resemble the 6:1 evolution obtained with open boundaries. We notice a general drop in the amplitude of mantle flow when changing to the 6:1 aspect ratio, which is caused by the increasing shear friction between mantle and lithosphere while the driving slab buoyancy is the same. Based on energy-dissipation arguments we applied a flow speed scaling to convert between flow fields of different model aspect ratios. This proved succesful for the open boundary model. We have also investigated the effect of far-field stress conditions in our open boundary models. Applying realistic normal stress conditions to the strong part of the overriding plate we show that "intra-plate" stresses control subduction dynamics resulting in slab roll-back, stationary or advancing subduction. We conclude that open boundaries are to be preferred for modelling subduction evolution (rollback, stationary or advancing). The relative independence of model aspect ratio avoids the need to place sidewalls at large distance and allows to focus all computational resources on a smaller modelling domain. Open boundaries simulate the natural subduction environment better and avoid the adverse effects (e.g. forced return flows) of free-slip boundaries