Giving AI a “crystal clear” view of the materials world

Scientists have developed an advanced artificial intelligence (AI) model that could revolutionize the search for new materials by providing a "crystal clear" view of their structures. The researchers at MIT created a vision-language model that combines 3D atomic structures with natural language descriptions, allowing AI to better understand the properties and potential applications of materials.

Unlike existing large language models that rely heavily on text from published papers, this multimodal AI model synthesizes narratives about crystal structures, making it more effective at predicting material functions. The model was trained on a dataset containing AI-generated descriptions of crystal structures, which helped it learn the connection between atomic arrangements, bonding patterns, and material properties.

When tested, the model outperformed text-only models in identifying materials suited for solid-state lithium-ion battery electrolytes, including those previously identified as promising candidates. The researchers believe their approach can accelerate materials science by providing a more comprehensive understanding of the relationship between a material's structure and its properties.