Machine-learning molecular dynamics simulation reveals lower supercritical silicate-water fluid viscosity, indicating rapid ascent from subducting slabs to the crust in 8,000 years.
Key Points
To determine the viscosity of supercritical NaAlSi3O8–H2O fluids across a broad range of pressures, temperatures, and water concentrations to understand volatile-driven magma transport in subduction zones.
Simulated supercritical NaAlSi3O8–H2O mixtures (0–100% H2O) using molecular dynamics powered by an ab initio-quality machine learning potential.
Tested conditions across temperatures of 1200–3000 K (at 2 GPa) and pressures of 1–5 GPa (at 2500 K).
Applied an empirical viscosity model to calculate Darcian fluid percolation velocity and slab-to-Moho transit times in subduction zones.
Demonstrated that water breaks down the silicate melt network at the atomic scale, yielding lower fluid viscosities than predicted by prior models.
Found that fluid percolation velocity increases during ascent from the slab into the mantle wedge and slows during subsequent upward migration.
Calculated fluid ascent times from the subducting slab to the Moho of approximately 8,000 years, satisfying geochemical U-Th disequilibria constraints.