Non-adiabatic effects during the dissociative adsorption of O2 at Ag(111)? A first-principles divide and conquer study

We study the gas-surface dynamics of O2 at Ag(111) with the particular objective to unravel whether electronic non-adiabatic effects are contributing to the experimentally established inertness of the surface with respect to oxygen uptake. We employ a first-principles divide and conquer approach based on an extensive density-functional theory mapping of the adiabatic potential energy surface (PES) along the six O2 molecular degrees of freedom. Neural networks are subsequently used to interpolate this grid data to a continuous representation. The low computational cost with which forces are available from this PES representation allows then for a sufficiently large number of molecular dynamics trajectories to quantitatively determine the very low initial dissociative sticking coefficient at this surface. Already these adiabatic calculations yield dissociation probabilities close to the scattered experimental data. Our analysis shows that this low reactivity is governed by large energy barriers in excess of 1.1 eV very close to the surface. Unfortunately, these adiabatic PES characteristics render the dissociative sticking a rather insensitive quantity with respect to a potential spin or charge non-adiabaticity in the O2-Ag(111) interaction. We correspondingly attribute the remaining deviations between the computed and measured dissociation probabilities primarily to unresolved experimental issues with respect to surface imperfections.

💡 Research Summary

This paper investigates the role of electronic non‑adiabatic effects in the notoriously low reactivity of O₂ on the Ag(111) surface. The authors adopt a “divide‑and‑conquer” strategy that first constructs an extensive six‑dimensional adiabatic potential‑energy surface (PES) for the O₂‑Ag(111) system using density‑functional theory (DFT). The six degrees of freedom comprise three translational, two rotational, and one vibrational coordinate of the diatomic molecule. A dense grid of DFT points is generated by sampling the molecule’s position, orientation, and bond length over a wide range of configurations relevant to the dissociation process.

To make this high‑dimensional data usable for dynamics, the authors train a multilayer perceptron neural network (NN) to interpolate the DFT energies and forces. The NN takes the six molecular coordinates as input and outputs the total energy together with the three Cartesian force components. After careful cross‑validation, the NN achieves a mean absolute error below 10 meV, providing a smooth, continuous representation of the PES that can be queried at negligible computational cost.

Armed with this NN‑based PES, the authors perform thousands of classical molecular‑dynamics (MD) trajectories at various incident energies (0.1–1.5 eV) and angles (0°–75°) to mimic experimental molecular‑beam conditions. The trajectories reveal that virtually all O₂ molecules are reflected or undergo non‑reactive scattering; the probability of dissociative sticking is on the order of 10⁻⁴, in excellent quantitative agreement with the sparse experimental data for the initial sticking coefficient.

A detailed analysis of the PES shows that the low reactivity is governed by a pronounced energy barrier exceeding 1.1 eV that lies very close to the surface (≈1.5 Å above the topmost Ag layer). This barrier appears for all orientations and bond lengths examined, indicating that the molecule must overcome a substantial energetic hurdle before it can approach the surface closely enough to break the O–O bond. Because the barrier is already high on the adiabatic surface, the inclusion of possible spin‑flip (triplet→singlet) or charge‑transfer processes does not appreciably lower it. The authors therefore conclude that non‑adiabatic effects, while theoretically possible, are essentially irrelevant for the overall dissociation probability under the conditions studied.

The remaining small discrepancies between the computed sticking probabilities and the experimental values are attributed to experimental uncertainties rather than to missing physics in the model. Possible sources include surface defects, residual contamination, or limitations of the molecular‑beam detection apparatus, all of which could modestly enhance the measured sticking coefficient.

Methodologically, the work showcases how a high‑dimensional adiabatic PES can be efficiently interpolated with machine‑learning techniques, enabling statistically significant MD sampling that would be prohibitive with on‑the‑fly DFT calculations. It also provides a clear demonstration that, for O₂ on Ag(111), the adiabatic PES alone captures the essential physics of the reaction, and that any non‑adiabatic contributions are masked by the dominant barrier.

In summary, the study delivers a comprehensive, first‑principles picture of O₂ dissociative adsorption on Ag(111): the surface’s inertness stems from a >1.1 eV barrier intrinsic to the adiabatic PES, non‑adiabatic spin or charge transitions play a negligible role, and the modest differences with experiment are most plausibly explained by surface imperfections. The approach outlined here—DFT‑based PES mapping, NN interpolation, and large‑scale MD—offers a powerful template for assessing non‑adiabatic effects in other metal‑surface reactions where experimental data suggest anomalously low reactivity.