High-Dimensional Atomistic Neural Network Potential to Study the Alignment-Resolved O₂ Scattering from Highly...

Area of Science:

• Computational Chemistry
• Surface Science
• Molecular Dynamics

Background:

• Understanding gas-surface interactions is crucial for catalysis and materials science.
• Accurate potential energy surfaces (PES) are essential for simulating molecular scattering dynamics.
• Previous studies lacked detailed insights into the stereodynamics of O2 scattering from graphite.

Purpose of the Study:

• To construct a high-dimensional, accurate atomistic neural network potential energy surface (ANN-PES) for O2-HOPG interactions.
• To investigate the alignment-dependent scattering dynamics of O2 molecules from a highly oriented pyrolytic graphite (HOPG) surface.
• To compare simulation results with recent experimental data for validation.

Main Methods:

• Development of an ANN-PES using the open-source package (aenet).
• Validation of the ANN-PES through static properties and comparison with ab initio molecular dynamics.
• Quasi-classical molecular dynamics simulations of O2 scattering with specified initial conditions (200 meV, 22.5° incidence, 110-300 K surface temperature).

Main Results:

• O2 scattering from HOPG is a direct process with alignment-dependent angular distributions.
• Perpendicularly aligned O2 molecules lose more translational energy (~20% less) than side-on aligned molecules.
• Perpendicular initial alignment leads to significant rotational excitation, while parallel alignment shows minimal rotational state change.

Conclusions:

• The developed ANN-PES accurately describes O2-HOPG interactions and scattering dynamics.
• Stereodynamics significantly influence energy transfer, confirming the dependence of scattering outcomes on molecular orientation.
• Simulation results align well with experimental observations regarding angular distributions and alignment dependence.