META-LEARN

Quantum-mechanical simulations of microstructural entities with 10⁶ atoms have a computational cost that made them inapplicable to relevant engineering applications. Machine-learned interatomic potentials (MLIPs) enable, for the first time, these large-scale simulations with near-quantum-mechanical accuracy. However, the construction of an MLIP for a specific material is limited to highly experienced domain experts from different disciplines: simulation, machine learning, engineering, physics, and material science. To tackle this limitation, in this project, titled META-LEARN, we envision the creation of a knowledge graph that can store this expert knowledge and allow for automatized assistance to MLIP creators. We will study both the creation of such a knowledge graph and the methods to facilitate the design process of MLIPs.

Project Details

This  project is funded by an ERC grant awarded to Prof. Dr. rer. nat. Blazej Grabowski, head of the Department of Materials Design of the Institute for Materials Science of the University of Stuttgart. The Analytic Computing department supports the research group of Prof. Grabowski with expertise on ontology and knowledge engineering.

Funding Period: January 2026 - December 2030.
Source of Funding: European Research Council #101200433.

Research Team

• Prof. Dr. rer. nat. Blazej Grabowski (Head of the Materials Design Department)
• Dr. Xi Zhang (Materials Design)
• Wenchuan Liu (Materials Design)