Shi Yin, Jinming Mu, Xudong Zhu, Lixin He
This paper presents a novel approach to crystal structure prediction using large language models and constrained optimization to improve symmetry inference and enforce physical validity, achieving state-of-the-art results without relying on existing databases.
Researchers have developed a new method for predicting the structure of crystals, which are important for discovering new materials and understanding their properties. Traditional methods often rely on existing databases or templates, which can limit the discovery of new structures. This new approach uses advanced language models to better understand chemical information and generate detailed patterns directly from the chemical composition. By combining this with a process that ensures the physical and mathematical consistency of the crystal structures, the method can predict new and unique structures more accurately than before.
