Integrating X-ray Diffraction with AI and First-Principles Methods for Materials Discovery

Experimental x-ray diffraction (XRD) provides a critical data stream for autonomous materials discovery, where rapid, reliable interpretation of structural information is essential. This talk will cover a series of computational methods that utilize XRD data in materials structure solution and discovery efforts. Early work [1] combined diffraction data, crystallographic symmetry, and density functional theory (DFT) energetics in first-principles-assisted structure solution (FPASS), enabling automated resolution of complex and metastable crystal structures with minimal human intervention. Subsequent work [2] automated the FPASS method and validated/utilized this approach across a wide range of compounds. We have also utilized XRD data in prototype-based and database-driven approaches to solve hundreds of previously unsolved experimental compounds at low computational cost, thereby expanding the searchable space for high-throughput computation.[3] Finally, in more recent work we extend these tools to a phase-mapping framework for high-throughput combinatorial XRD, where domain knowledge from crystallography, thermodynamics, and chemistry is explicitly encoded into optimization and machine-learning models.[4] By integrating experimental constraints directly into AI algorithms, these approaches enable physically grounded interpretation of diffraction data and accelerate computational materials discovery efforts.

1. Meredig, B.; Wolverton, C. Nature Materials 12, 123–127 (2013).

2. Ward, L.; Michel, K.; Wolverton, C. Physical Review Materials 1, 063802 (2017).

3. Griesemer, S. D.; Ward, L.; Wolverton, C. Physical Review Materials 5, 105003 (2021).

4. Yu, D.; Griesemer, S.; Liu, T.-C.; Wolverton, C.; Zhu, Y. npj Computational Materials 11, 354 (2025).