Atomistic Modeling of W/h-BN/Graphene Multilayer Coatings for the Protection of the First Wall of a Fusion Reactor: Concept Justification, Validation of Neural Network Potentials, and the First Stage of the Study

The paper considers the degradation of the first wall of a fusion reactor under the combined action of helium fluxes, hydrogen isotopes, and fast neutrons. The standard wall material, tungsten, is destroyed due to helium erosion, hydrogen embrittlement, and neutron cascades. A multilayer protective coating architecture is proposed: a tungsten substrate providing heat removal; a hexagonal boron nitride layer enriched with boron-10 intended for neutron absorption; and an outer graphene barrier limiting the penetration of helium and hydrogen isotopes. The modeling was performed by molecular dynamics with neural network potentials in the GPUMD package. The universal NEP89 potential covering the elements W, B, N, and C was chosen as the main potential. Validation was carried out: the tungsten lattice parameter was reproduced with a deviation of 0%, and the C-C bond lengths in graphene and B-N bonds in h-BN were reproduced with deviations of 0.02% and 0.06%, respectively. It was established that the three-layer system remains structurally stable at 1200 K, and the layers do not mix. The adhesion energy of the W/h-BN interface is -16.48 eV, which indicates stable interlayer bonding. A separate test showed that NEP89 is unsuitable for describing the interaction of He with tungsten: the predicted insertion energy was -41.7 eV, whereas the DFT value lies in the range of +4...+6 eV. This result justifies the need to train a specialized potential for helium, which determines the next task of the study. This work is the first stage of a single study, the results of which will be combined into a final article for a peer-reviewed journal.