Scientific Achievement

• Researchers in the Theory of Materials program developed a computationally efficient and transferable dispersion-aware physics-based ML interatomic potential (MLIP) to enable studies of moiré reconstruction in TMD heterostructures

Significance and Impact

• Atomic reconstruction in TMD moiré superlattices with more than 1000 atoms can now be accurately predicted with the accuracy of ab initio van der Waals-corrected density functional theory (DFT)

Research Details

• Developed an approach to train MLIPs using active learning and DFT on small unit cells. MLIP provides new insight into substrate effects on measured atomic-scale corrugation in moiré TMDs

Publication Details

Y. Shaidu, M.H. Naik, S.G. Louie, J.B. Neaton, npj Computational Materials (2025).

DOI:10.1038/s41524-025-01761-9

Work was performed at Lawrence Berkeley National Lab.