Bias-induced NO adsorption and configurational transitions on Cu(110): a machine learning enhanced first-principles grand canonical study.

Precise control over the configurational changes of molecules adsorbed on metal surfaces is critical for advancing molecular-scale technologies, including single-molecule junctions and molecular switches. A bias voltage applied via scanning tunneling microscopy (STM) enables such control by inducing configurational switching through electron injection/extraction. However, gaining atomistic insights into these processes remains experimentally challenging. Herein, we employ grand-canonical self-consistent field-density functional theory (GCSCF-DFT) calculations and a machine-learning-enhanced nudged elastic band (ML-NEB) method to elucidate the NO adsorption on Cu(110) and its configurational change from a side-on (SO-NO) flat-lying to short-bridge (SB) upright geometry under a constant electron chemical potential scheme. Our results reveal pronounced bias-dependent modifications of the electronic structures of adsorbed NO in SB-NO and SO-NO configurations. Moreover, applying a bias potential does not lower the activation barrier for the SO-NO → SB-NO conversion; instead, it facilitates switching through vibrational excitation of the NO molecule. These findings provide a comprehensive atomistic picture of bias-driven molecular reconfiguration on Cu(110) and offer guidance for the rational design of molecular switches.