Mineral phase transformations significantly alter the bulk density and elastic properties of mantle rocks and consequently have profound effects on mantle dynamics and seismic wave propagation. These changes in the physical properties of mantle rocks result from evolution in the equilibrium mineralogical composition, which can be predicted by the minimization of the Gibbs Free Energy with respect to pressure (P), temperature (T), and chemical composition (X). Thus, numerical models that simulate mantle convection and/or probe the elastic structure of the Earth’s mantle must account for varying mineralogical compositions to be self-consistent. Yet coupling Gibbs Free Energy minimization (GFEM) approaches with numerical geodynamic models is currently intractable for high-resolution simulations because execution speeds of widely-used GFEM programs (100–102 ms) are impractical in many cases. As an alternative, this study introduces machine learning models (RocMLMs) that have been trained to predict thermodynamically self-consistent rock properties at arbitrary PTX conditions between 1–28 GPa, 773–2273 K, and mantle compositions ranging from fertile (lherzolitic) to refractory (harzburgitic) end-members defined with a large dataset of published mantle compositions. RocMLMs are 101–103 times faster than GFEM calculations or GFEM-based look-up table approaches with equivalent accuracy. Depth profiles of RocMLMs predictions are nearly indistinguishable from reference models PREM and STW105, demonstrating good agreement between thermodynamic-based predictions of density, Vp, and Vs and geophysical observations. RocMLMs are therefore capable, for the first time, of emulating dynamic evolution of density, Vp, and Vs in high-resolution numerical geodynamic models.

A key feature of subduction zone geodynamics and thermal structure is the point at which the slab and mantle mechanically couple. This point defines the depth at which traction between slab and mantle begins to drive mantle wedge circulation and also corresponds with a major increase in temperature along the slab-mantle interface. Here we consider the effects of the backarc thermal structure and slab thermal parameter on coupling depth using two-dimensional thermomechanical models of oceanic-continental convergent margins. Coupling depth is strongly correlated with backarc lithospheric thickness, and weakly correlated with slab thermal parameter. Slab-mantle coupling becomes significant where weak, hydrous antigorite reacts to form strong, anhydrous olivine and pyroxene along the slab-mantle interface. Highly efficient (predominantly advective) heat transfer in the asthenospheric mantle wedge and inefficient (predominantly conductive) heat transfer in the lithospheric mantle wedge results in competing feedbacks that stabilize the antigorite-out reaction at depths determined primarily by the mechanical thickness of the backarc lithosphere. For subduction zone segments where backarc lithospheric thickness can be inverted from surface heat flow, our results provide a regression model that can be applied with slab thermal parameter to predict coupling depth. Consistently high backarc heat flow in circum-Pacific subduction zones suggests uniformly thin overriding plates likely regulated by lithospheric erosion caused by hydration and melting processes under volcanic arcs. This may also explain a common depth of slab-mantle coupling globally.