This research demonstrates how Large Language Models can be integrated with experimental data to accelerate materials discovery by identifying causal relationships in complex materials systems.

• Combines domain knowledge with LLM capabilities to create enhanced causal discovery frameworks
• Successfully applied to ferroelectric materials (Sm-doped BiFeO3) to identify key property determinants
• Outperforms traditional statistical methods in identifying relevant causal factors
• Provides a pathway to optimize material synthesis conditions for desired properties

For engineering teams, this approach offers a powerful new tool to reduce experimental iterations and accelerate materials development cycles, particularly in discovering novel materials with specific functional properties.

Causal Discovery from Data Assisted by Large Language Models