Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 12: Morphology Prediction at Interfaces
O 12.4: Vortrag
Montag, 7. März 2016, 15:45–16:00, S051
Machine learning of surface adsorbate structure — •Milica Todorović1, Michael Gutmann2, Jukka Corander2, and Patrick Rinke1 — 1Aalto University, Espoo, Finland — 2University of Helsinki, Helsinki, Finland
To efficiently search many atomistic configurations on large length scales, we developed a parameter-free machine learning tool for studying organic/inorganic interfaces. Our preferred Bayesian optimisation approach relies on probabilities to construct model functions, which are then iteratively refined by input of real data points balancing exploitation with exploration. A Bayesian optimisation algorithm was interfaced with both classical potential and density-functional theory codes to enable iterative learning (on-the-fly) of potential energy surfaces (PES) on large supercomputers without human input. For additional efficiency, the method exploits structural rigidity of molecular groups, reducing the degrees of freedom and making the learning process analogous to “molecular LEGO“. We present a proof-of-concept test based on the alanine molecule and an application featuring electron donor C60 molecules on the (101) surface of TiO2 anatase. The Bayesian optimisation structure search (BOSS) acquires PES information fast and is particularly efficient in pinpointing the global minimum structure. This versatile scheme for global minimum search could be extended beyond interface packing considerations to address complex configurational problems across scientific disciplines.