Résumé
The mantle exposed at slow spreading ridges is pervasively serpentinized, down to ca. 5 km according to geophysical data. The onset and durability of this hydration process require efficient penetration and renewal of fluids at the mineral-fluid interface. However, the mechanisms of fluid penetration are still poorly understood. Moreover, serpentinization is exovolumic, if a mass-conservative system is assumed, or chemical elements are leached out to conserve rock volume. Thus, the extent of serpentinization depends of the system capacity to create space and/or to drive mass transfers. In order to investigate these hydrodynamic and chemical mechanisms, we did a laboratory experiment during which seawater was injected in a sintered San Carlos olivine sample at conditions representative of low temperature ultramafic hydrothermal systems. The percolation-reaction experiment was carried out at 19 MPa and 190 degrees C; the initial water flow was set at 0.2 mL/h then decreased down to 0.06 mL/h after 8 days. During the experiment (23 days), permeability decreased continuously. The composition of the outlet fluid varied strongly during the first 24h of the experiment, then reached equilibrium values. The high Si concentrations in outlet fluids indicated steady olivine dissolution, while their low Fe and Mg concentrations suggested precipitation of Fe- and Mg-rich mineral phases. The reacted sample acquired a reddish brown color, indicating oxidation reactions occurred. Optical observation and SEM imaging revealed the presence of a soft white material filling the pores of the reacted sample. It was identified as a poorly crystallized serpentine type material by AEM/TEM analyses. This proto-serpentine is intimately associated to <100 nm Fe-oxide patches (probably hematite) growing on the olivine surface. We interpret the precipitation of this proto-serpentine together with Fe-oxides throughout the sample as marking the early stages of serpentinization. The fluid composition not being modified by changes in fluid flow, we posit that olivine serpentinization is not controlled by elemental transport and/or by chemical kinetics, but takes place at-equilibrium; the reaction velocity is very fast and independent on the changes in the reaction surface area. Indeed, mass balance calculations indicate that more than 15 wt.% olivine was dissolved while the same mass of proto-serpentine was formed; concurrently, porosity decreased from approximately 12% to 5%. We infer that the structure of the newly formed proto-serpentine resulted in the clogging of fluid paths and explain the decrease of permeability during experiments. Yet the loose structure of the proto-serpentine allows maintaining a connected pore structure ensuring the perennial renewing (diffusion and/or advection) of the reactants at the reaction surfaces. These experimental results are used to constrain numerical reactive transport models and better understand the scale and efficiency of serpentinization reactions (effective reaction rates in porous/fractured media) at the scale of spreading ridges.