Controls on the Migration of Fluids in Subduction Zones

Arc volcanism associated with subduction is generally considered to be caused by the transport in the slab of hydrated minerals to sub-arc depths. In a qualitative sense it appears clear that progressive dehydration reactions in the down-going slab release fluids to the hot overlying mantle wedge, causing flux melting and the migration of melts to the volcanic front. However, the quantitative details of fluid release, migration, melt generation and transport in the wedge remain poorly understood. In particular, there are two fundamental observations that defy quantitative modeling. The first is the location of the volcanic front with respect to intermediate depth earthquakes (e.g. 100+/-40 km; England et al., 2004, Syracuse and Abers, 2006) which is remarkably robust yet insensitive to subduction parameters. This is particularly surprising given new estimates on the variability of fluid release in global subduction zones (e.g. van Keken et al. 2011) which show great sensitivity of fluid release to slab thermal conditions. Reconciling these results implies some robust mechanism for focusing fluids and/or melts toward the wedge corner. The second observation is the global existence of thermally hot erupted basalts and andesites that, if derived from flux melting of the mantle requires sub-arc mantle temperatures of 1300 degrees C over shallow pressures of 1-2 GPa which are not that different from mid-ocean ridge conditions. These observations impose significant challenges for geodynamic models of subduction zones, and in particular for those that do not include the explicit transport of fluids and melts. We present a range of high-resolution models that include a more complete description of coupled fluid and solid mechanics (allowing the fluid to interact with solid rheological variations) together with rheologically consistent solution for temperature and solid flow. Focusing on end-members of a global suite of arc geometries and thermal histories we discuss how successful these interactions are at focusing both fluids and hot solids to sub-arc regions worldwide. We will also evaluate the efficacy of current wet melting parameterizations in these models. When driven by buoyancy alone, fluid migrates through the mantle wedge along a near vertical trajectory. Only interactions with the solid flow at very low values of permeability or high values of fluid viscosity can cause deviations from this path. However, in a viscous, permeable medium, additional pressure gradients are generated by volumetric deformation due to variations in fluid flux. These pressure gradients can significantly modify the fluid flow paths. At shallow depths, compaction channels form along the rheological contrast with the overriding plate while in the mantle wedge itself porosity waves concentrate the fluid. When considering multiple, distributed sources of fluid, interaction between layers in the slab itself can also cause significant focusing. As well as permeability, rheological controls and numerical regularizations place upper and lower bounds on the length-scales over which such interactions occur further modifying the degree of focusing seen. The wide range of behaviors described here is modeled using TerraFERMA (the Transparent Finite Element Rapid Model Assembler), which harnesses the advanced computational libraries FEniCS, PETSc and SPuD to provide the required numerical flexibility.