Accurate rovibrational energies of ozone isotopologues up to J = 10 utilizing artificial neural networks

In recent years, ozone and its isotopologues have been a topic of interest in many fields of research, due to its importance in atmospheric chemistry and its anomalous isotopic enrichment-or the so-called "mass-independent fractionation." In the field of potential energy surface (PES) creation, debate over the existence of a potential barrier just under the dissociation threshold (referred to as a "potential reef") has plagued research for some years. Recently, Dawes and co-workers [Dawes, Lolur, Li, Jiang, and Guo (DLLJG) J. Chem. Phys. 139, 201103 (2013)] created a highly accurate global PES, for which the reef is found to be replaced with a (monotonic) "plateau." Subsequent dynamical calculations on this "DLLJG" PES have shown improved agreement with experiment, particularly the vibrational spectrum. However, it is well known that reaction dynamics is also highly influenced by the rovibrational states, especially in cases like ozone that assume a Lindemann-type mechanism. Accordingly, we present the first significant step toward a complete characterization of the rovibrational spectrum for various isotopologues of ozone, computed using the DLLJG PES together with the ScalIT suite of parallel codes. Additionally, artificial neural networks are used in an innovative fashion-not to construct the PES function per se but rather to greatly speed up its evaluation.