Out of the Crystalline Comfort Zone: Sampling the Initial Oxide Formation At Cu(111)

By combining replica exchange molecular dynamics with fast and efficient machine‐learned interatomic potentials, this computational study predicts the complex surface reconstructions accompanying the early Cu(111) surface oxidation. The unprecedentedly exhaustive sampling results in a high‐quality surface phase diagram that not only recovers known reconstructions, but also explains the accompanying disorder and uncovers new entropically‐stabilized low‐coverage phases. Abstract Oxidizing transition metal surfaces are generally characterized by an increasing heterogeneity at simultaneous lowering of crystalline order. This complexity eludes present‐day first‐principles descriptions, with predictive‐quality surface phase diagrams commonly derived from comparing the stability of a small number of ordered surface structural models that are motivated by partial experimental characterization or chemical intuition. Here the computational acceleration brought by machine‐learned interatomic potentials is leveraged for a systematic sampling of the configurational phase space through replica exchange molecular dynamics. Thermodynamic averaging subsequently yields grand‐canonical expectation values for observables like O coverage that account for the disorder and diversity of the sampled structures. Application to the initial oxidation of the Cu(111) surface reveals the (purely entropic) stabilization of sparse O adsorbates at the onset, a plethora of energetically essentially degenerate polymeric –O–Cu–O– ring and chain networks at higher O loading, as well as the presence of experimentally discussed minority species. The in silico surface phase diagram correspondingly shows marked differences to one based merely on established ordered surface reconstructions. By combining replica exchange molecular dynamics with fast and efficient machine-learned interatomic potentials, this computational study predicts the complex surface reconstructions accompanying the early Cu(111) surface oxidation. The unprecedentedly exhaustive sampling results in a high-quality surface phase diagram that not only recovers known reconstructions, but also explains the accompanying disorder and uncovers new entropically-stabilized low-coverage phases. Abstract Oxidizing transition metal surfaces are generally characterized by an increasing heterogeneity at simultaneous lowering of crystalline order. This complexity eludes present-day first-principles descriptions, with predictive-quality surface phase diagrams commonly derived from comparing the stability of a small number of ordered surface structural models that are motivated by partial experimental characterization or chemical intuition. Here the computational acceleration brought by machine-learned interatomic potentials is leveraged for a systematic sampling of the configurational phase space through replica exchange molecular dynamics. Thermodynamic averaging subsequently yields grand-canonical expectation values for observables like O coverage that account for the disorder and diversity of the sampled structures. Application to the initial oxidation of the Cu(111) surface reveals the (purely entropic) stabilization of sparse O adsorbates at the onset, a plethora of energetically essentially degenerate polymeric –O–Cu–O– ring and chain networks at higher O loading, as well as the presence of experimentally discussed minority species. The in silico surface phase diagram correspondingly shows marked differences to one based merely on established ordered surface reconstructions. Advanced Science, Volume 12, Issue 48, December 29, 2025.