Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 17: Development of Calculation Methods

MM 17.6: Vortrag

Mittwoch, 19. März 2025, 11:45–12:00, H22

Beyond-DFT Machine-Learning Interatomic Potentials and Applications to Covalent-Organic Frameworks — •Yuji Ikeda, Axel Forslund, and Blazej Grabowski — University of Stuttgart, Germany

Covalent-organic frameworks (COFs) are nanoporous crystalline materials composed of covalent organic secondary building units (SBUs), primarily composed of light elements such as C, N, O, and H. Many COFs exhibit a quasi-two-dimensional layered structure, stabilized by van der Waals (vdW) interactions. Machine-learning interatomic potentials (MLIPs) offer an exciting opportunity to explore COFs, enabling access to extended time and length scales in molecular dynamics (MD) simulations. To accurately model vdW interactions, MLIPs must be trained on datasets that include these effects, often derived from vdW-DFT functionals. However, vdW-DFT methods are essentially semi-empirical, with parameters calibrated for experimental agreement, raising concerns about their transferability. Instead, we propose generating training data from post-Hartree-Fock methods, such as coupled-cluster (CC) calculations, which are non-empirical and provide beyond-DFT accuracy. By utilizing MLIPs trained on these high-accuracy datasets, we aim to investigate the structural properties of COFs in unprecedented detail.

Keywords: Machine-learning interatomic potentials (MLIPs); Covalent-organic frameworks (COFs)