The knowledge of surface structures is crucial for accurate predictions of many material properties. To produce cheaper, more efficient, and sustainable materials for important technologies, like heterogenous catalysis (HC), we need precise theoretical tools for recovering surface structures. HC synthesis of important chemicals consumes 1% of energy production, and it is also essential for the transition from fossil fuels to sustainable ones. In TRSS, we aim to deliver a novel AI tool, a graph neural network connected with reinforcement learning, for assembling surface structures from atoms and bonds, almost like assembling Lego. In connection with other AI technologies, like machine learning force fields, we will rapidly create and screen a vast number of surface models. Through collaboration with atom-resolving microscopy, we will be able to tune and test our AI tools, which will significantly accelerate materials engineering and thus our transformation towards green technologies.

2026

2029