Journal: Journal of Chemical Physics
DOI: 10.1063/1.5036602
PubMed: 30007387
Published: July 12, 2018
Authors
C. Petty, R. F. Spada, F. Machado, B. Poirier
Abstract
Ozone and its isotopologues have long been of interest in atmospheric chemistry and for their anomalous isotopic enrichment — the so-called “mass-independent fractionation.” Debate has persisted about the existence of a potential barrier just under the dissociation threshold (the “potential reef”). This work uses the highly accurate Dawes-Lolur-Li-Jiang-Guo (DLLJG) potential energy surface, which replaces the reef with a monotonic plateau, to compute the first significant characterization of the rovibrational spectrum for various ozone isotopologues.
Artificial neural networks are used innovatively — not to construct the PES itself, but to dramatically speed up its evaluation during dynamical calculations. Calculations use the ScalIT suite of parallel codes.
Notes
- First comprehensive rovibrational characterization of ozone isotopologues using the DLLJG PES
- Novel use of neural networks for PES acceleration (not construction) — an early ML-in-chemistry application
- Directly relevant to atmospheric ozone chemistry and isotope fractionation puzzles
- Collaboration between Texas Tech (Poirier group) and Brazilian ITA group (Spada, Machado)