Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 33: Machine Learning in Dynamics and Statistical Physics I
DY 33.13: Vortrag
Donnerstag, 20. März 2025, 12:45–13:00, H47
Molecular Dynamics of Endohedral CaX@C60 Fullerenes: Reproducing Correlated Movement Features Using Machine Learning Applications — •Mihaela Cosinschi1,3, Amanda Teodora Preda1,3, Calin Andrei Pantis Simut1,3, Nicolae Filipoiu1,3, Ioan Ghitiu4, Mihnea Alexandru Dulea3, Andrei Manolescu5, and George Alexandru Nemnes1,2,3 — 1University of Bucharest, Faculty of Physics, Magurele, Romania — 2Research Institute of the University of Bucharest, Bucharest, Romania — 3Horia Hulubei National Institute for Physics and Nuclear Engineering, Magurele, Romania — 4National Institute for Laser, Plasma and Radiation Physics, Magurele, Romania — 5Department of Engineering, School of Technology, Reykjavik University, Reykjavik, Iceland
Fullerenes are allotropes of carbon with remarkable properties due to their high degree of symmetry, cage-like structures and ability to support addition of internal or external atoms. In the present work, we have conducted a molecular dynamics (MD) study on the C60 fullerene containing one to four encapsulated calcium atoms. All-atomistic ab initio DFT methods were employed to perform calculations through the Orca MD Module, albeit at a high computational cost. Results proved that the internal atoms adopt minimal-energy configurations and exhibit coupled motion, maintaining constant characteristics after a period of equilibration. Furthermore, we have built an artificial neural network (ANN) that can pick up the dynamics patterns and recreate trajectories to reasonable accuracy, allowing for MD calculations in significantly shorter times, even under small perturbations.
Keywords: fullerene; potential mapping; ab initio DFT; molecular dynamics; machine learning