The presence of water (or hydrogen) in nominally-anhydrous mantle minerals, as well as its influence on the chemical and physical properties of these minerals, has been known for almost twenty years. Nevertheless, the total amount of water stored in the mantle is still very poorly constrained and this value is a crucial parameter in understanding the effects of water on the Earth’s dynamics. The poor knowledge of hydrogen concentration in the mantle comes partly from the fact that, strangely, the thermodynamics (solubility, diffusion coefficients, etc.) at mantle conditions of H in olivine, probably the most studied mantle silicate, and its high-pressure polymorphs is also poorly known. In this project we propose to explore this topic by associating experiments with theoretical works. Cutting-edge experiments will be performed, at both room and high pressure, in order to measure the self-diffusion coefficients of H in the three phases, forsterite, wadsleyite and ringwoodite. The pure-Mg end-member of olivine, forsterite, allows to run H self-diffusion experiments at room pressure. Fe-bearing samples would necessitate more time-consuming high-pressure annealing. On the other hand, experiments on wadsleyite and ringwoodite, have to be performed at high pressure, in a large-volume multi-anvil apparatus. H concentration will be measured using infrared spectroscopy, with spectra collected after various annealing durations that will then permit to calculate the hydrogen self-diffusion coefficients. These IR measurements will be obtained on an infrared microscope available in our laboratory and also at the infrared synchrotron facility “Soleil” in Orsay for the shorter profiles. In parallel, computer simulations, atomistic and ab initio, will be used in order to investigate the microscopic processes involved in H diffusion and related point defects in forsterite first, and then wadsleyite and ringwoodite. Atomistic calculations will be done with the code GULP. Because this method is computationally fast with today’s calculators, we will be able to explore a wide range of H-related defect configurations and diffusion paths. Calculations using a quantum approach will be performed with the code SIESTA. These are much more computationally demanding but also much more precise (or exact) than the atomistic approach. They will run on national facility supercomputers and be applied to specific key points found from the GULP calculations. Results of both theoretical and experimental approaches will be compared and will allow to propose a coherent point-defect model explaining H defects and migration in forsterite. We will later investigate if this model also applies, with some adjustments, to Fe-bearing olivine. Theoretical calculations will also be applied to the understanding of the effect of hydrogen on the mobility (and related defects) of other major cations, Mg, Si. The ultimate goal of this project is then to model the upper-mantle electrical conductivitiy profile, linked to H diffusivity via the Nernst-Einstein equation, using our data. We will then compare this model with magnetotelluric measurements. It will follow an estimate of the total amount of water present in the upper mantle, a fundamental result for modelling and understanding mantle dynamics. This project has been already submitted in February 2008 to the programme “Blanc” and was selected in the reserve list.