HB238 Adsorption on Coinage Metal Surfaces: Insights from a Combined Machine-Learning Interatomic Potential and Bayesian Optimization Framework for Interfacial Electronic Structure

Organic donor–acceptor (D-A) molecules with permanent dipoles are attractive candidates for flexible electronics. Merocyanines represent a well-studied class of such systems, both experimentally and theoretically. In this study, the adsorption of a single merocyanine molecule on Cu(100), Ag(100), and Au(100) surfaces was investigated using a workflow that combines machine learning interatomic potentials with Bayesian optimization. Interatomic potentials based on DFT data were fine-tuned for each substrate and embedded within a Bayesian optimization scheme to efficiently explore lateral translations and rotations. Candidate geometries were subsequently refined via dispersion-corrected DFT relaxations, and the electronic structure was analyzed through projected density of states and core-level spectra simulations. The search identified flat-lying adsorption as the energetically preferred configuration across all coinage metals, with the strongest substrate–molecule coupling on Cu(100). Core-level spectra calculations showed that donor sulfur atoms consistently possessed higher transition energies than acceptor sulfur atoms, and that the donor–acceptor splitting narrowed upon adsorption, indicating significant charge redistribution at the interface. This study provides an efficient route to map complex adsorption landscapes and to elucidate how metal reactivity influences the electronic structure of dipolar organic molecules.