Quantifying Vein Network Permeability in Dehydrated Serpentinites Using Thermodynamics and Generative AI

Abstract

Fluids released from subducting hydrated rocks influence volcanism, tectonics, and geochemical cycling, but the mechanisms of fluid escape in subduction zones remain poorly understood. We address this issue by investigating the Erro-Tobbio meta-serpentinites (ET-MS), Italy, exhumed serpentinite rocks that preserve extensive dehydration vein networks formed by the porosity-generating breakdown of antigorite and brucite. We characterized the structure and morphology of these self-organized vein networks and evaluated their hydrodynamic properties using a novel approach. Specifically, we combined X-ray tomography and drone imagery with generative machine learning, electron microscopy, and equilibrium thermodynamics to model and analyze fluid pathways in the ET-MS. In both natural and simulated samples, these dehydration vein networks act as efficient drainage systems, enabling rapid fluid percolation even at porosities below 1%. The maximum network permeability is (Formula presented.), several orders of magnitude higher than that of intact serpentinite. Fe-rich olivine and monticellite occur alongside relict brucite and magnetite in these veins. This assemblage indicates that the high permeability arises from porosity localized along brucite- and magnetite-rich veins, where infiltration of reducing fluids enhanced dehydration reactions. These findings demonstrate that serpentinite dehydration in subduction zones can produce flow-optimized vein structures that efficiently channel fluids at low porosity, potentially influencing fluid migration on local to regional scales before widespread dehydration occurs.