Retention and release of hydrogen isotopes in tungsten plasma-facing components: the role of grain boundaries and the native oxide layer from a joint experiment-simulation integrated approach

International audienceFusion fuel retention (trapping) and release (desorption) from plasma-facing components are critical issues for ITER and for any future industrial demonstration reactors such as DEMO. Therefore, understanding the fundamental mechanisms behind the retention of hydrogen isotopes in first wall and divertor materials is necessary. We developed an approach that couples dedicated experimental studies with modelling at all relevant scales, from microscopic elementary steps to macroscopic observables, in order to build a reliable and predictive fusion reactor wall model. This integrated approach is applied to the ITER divertor material (tungsten), and advances in the development of the wall model are presented. An experimental dataset, including focused ion beam scanning electron microscopy, isothermal desorption, temperature programmed desorption, nuclear reaction analysis and Auger electron spectroscopy, is exploited to initialize a macroscopic rate equation wall model. This model includes all elementary steps of modelled experiments: implantation of fusion fuel, fuel diffusion in the bulk or towards the surface, fuel trapping on defects and release of trapped fuel during a thermal excursion of materials. We were able to show that a single-trap-type single-detrapping-energy model is not able to reproduce an extended parameter space study of a polycrystalline sample exhibiting a single desorption peak. It is therefore justified to use density functional theory to guide the initialization of a more complex model. This new model still contains a single type of trap, but includes the density functional theory findings that the detrapping energy varies as a function of the number of hydrogen isotopes bound to the trap. A better agreement of the model with experimental results is obtained when grain boundary defects are included, as is consistent with the polycrystalline nature of the studied sample. Refinement of this grain boundary model is discussed as well as the inclusion in the model of a thin defective oxide layer following the experimental observation of the presence of an oxygen layer on the surface even after annealing to 1300 K