Abstract
The conventional understanding of bimolecular reactions, which either proceed directly via well-defined transition states or pass through potential energy wells, is well-established. However, increasing attention and interest have been drawn to nontraditional reaction pathways, such as roaming mechanisms. Here, full-dimensional dynamics simulations on a machine learning-based potential energy surface reveal that the Cl + C(2)H(2)→C(2)H+HCl reaction is dominated by two roaming mechanisms-Cl-roaming and H-roaming-rather than direct abstraction. In Cl-roaming, a transient C(2)H(2)Cl adduct forms, allowing Cl to roam and abstract H. In H-roaming, a detached H atom migrates and abstracts Cl. These pathways account for nearly 100% of the total yield, exhibiting distinct energy and angular distributions. These findings challenge the traditional view of the bimolecular reaction with conventional transition states, emphasizing the importance of considering nontraditional pathways in reaction dynamics studies for accurate rate constant predictions and mechanistic insights.