Accident-tolerant hybrid ceramics for fusion breeding blanket

• Li–Be–X–O is synthesized based on the prediction by the machine learning model. • Steam exposure tests at 1200 °C verified the accident tolerance of hybrid ceramics. • Two-phase hybrid ceramics can breed sufficient fuel tritium without metallic Be. • Hybrid ceramics allows a novel blanket design with enhanced safety margins. In a water-cooled ceramic breeding blanket for fusion reactors, hydrogen gas generation by steam oxidation of metallic Be compounds (i.e. neutron multiplier) in a loss-of-coolant accident (LOCA) raises major safety concerns. Li–Be hybrid ceramics has a potential to reduce hydrogen generation significantly, however, the stable compositions and structures for quaternary compositions have not been comprehensively understood. Herein, we report machine-learning based prediction, synthesis, structure, and properties of chemically stabile two-phase Li–Be–X–O hybrid ceramics. The steam exposure tests demonstrated a negligibly small H 2 generation from the two-phase powder of Li 2 BeSiO 4 and 5 at.% BeO below 1200 °C. The stability is explained by the intrinsic ionic/covalent bonding characters and little capacity for further oxidation by steam. Neutronic calculations with simplified one-dimensional model show that the two-phase hybrid cermaics has a sufficient tritium breeding capability without having metallic Be-based multiplier in the blanket. The hybrid ceramics is the first example of multi-functional oxide to breed sufficient fuel tritium with no metallic neutron multiplier, which allows a novel design of ceramic breeding blanket with enhanced safety margins during in-box LOCA.