The decomposition dynamics of vibrationally excited syn-CH3CHOO to form vinoxy + hydroxyl (CH2CHO + OH) radicals or to recombine to form glycolaldehyde (CH2OHCHO) are characterized using statistically significant numbers of molecular dynamics simulations using a full-dimensional neural-network-based potential energy surface at the CASPT2 level of theory. The computed final OH-translational and rotational state distributions agree well with experiments and probe the still unknown O-O bond strength DeOO for which best values from 22 to 25 kcal/mol are found. OH-elimination rates are consistent with experiments and do not vary appreciably with DeOO due to the non-equilibrium nature of the process. In addition to the OH-elimination pathway, OH roaming is observed following O-O scission, which leads to glycolaldehyde formation on the picosecond time scale. Together with recent work involving the methyl-ethyl-substituted Criegee intermediate, we conclude that OH roaming is a general pathway to be included in molecular-level modeling of atmospheric processes. This work demonstrates that atomistic simulations with machine-learned energy functions provide a viable route for exploring the chemistry and reaction dynamics of atmospheric reactions.
Read full abstract