FASTER (Fuel And STructural matERials modelling)
Research on sustainable, safe and economic electricity generation is a field of high importance and impact. Key technologies under research and development are the next generation nuclear power systems (Gen-IV) and fusion reactors. Nuclear energy generates a large part of the electricity in the world and is currently undergoing an expansion and development phase. Many of the issues on which development is needed is related to the materials to be used in future nuclear systems. In this project, we aim to investigate the behavior under irradiation of key nuclear materials. Studying transuranic bearing nitride fuels is essential for developing reactor systems where spent fuel can be burnt in order to increase efficiency and drastically decrease the lifetime of the radioactive waste. For these fuels to be licensed, they have to be characterized and radiation riven processes in them have to be understood. State of the art first principles methods (DMFT, DFT+U, etc) will be applied in order to study these materials. Furthermore, we aim to investigate the evolution of the microstructure in ferritic and ferritic/martensitic steels (as structural materials), austenitic steels, oxide dispersion strengthened steels and FeCrAl alloys (for the fuel cladding). We build multi-scale models that are used to predict the long term behavior of these materials under different conditions, including varying levels of neutron irradiation. The long term models will be based on characteristics determined using state of the art first principles methods. In order to build a generic model for microstructure evolution with and without an irradiation field, a neural network coupled to a stochastic long term evolution model will be trained using first principles calculations. A few selected cases of particular interest will be used as primary case studies, including a validation case. This type of method, which is only possible to perform with the aid of peta-scale computing, can be a game changer for materials modeling, especially for nuclear applications. Even with a large scale SNIC allocation we cannot complete this study, but with the added support from external HPC resources we are making good progress.