Ab Initio Potential Energy Surfaces and Quantum Dynamics for Polyatomic Bimolecular Reactions

Abstract

There has been great progress in the development of potential energy surfaces (PESs) and quantum dynamics calculations in the gas phase. The establishment of a fitting procedure for highly accurate PESs and new developments in quantum reactive scattering on reliable PESs allow accurate characterization of reaction dynamics beyond triatomic systems. This review will give the recent development in our group in constructing ab initio PESs based on neural networks and the time-dependent wave packet calculations for bimolecular reactions beyond three atoms. Bimolecular reactions of current interest to the community, namely, OH + H, H + HO, OH + CO, H + CH, and Cl + CH, are focused on. Quantum mechanical characterization of these reactions uncovers interesting dynamical phenomena with an unprecedented level of sophistication and has greatly advanced our understanding of polyatomic reaction dynamics.

Potential Energy Landscape of CO Adsorbates on NaCl(100) and Implications in Isomerization of Vibrationally Excited CO

“…This systematic data sampling approach has been validated in various gas-phase and molecule-surface interaction systems. [37][38][39] Finally, 16196 symmetry unique points were sampled.…”

A critical comparison of neural network potentials for molecular reaction dynamics with exact permutation symmetry

“…NRPIP-NN and FI-NN are then compared by fitting PESs for OH + CO, H + H 2 S and H + NH 3 , whose FIs are available. 46 As has been mentioned above, the form of the invariant polynomials is not unique. Taking A 2 B, A(1)A(2)B(3), as an example, FIs and NRPIPs are given explicitly in Table 2.…”

“…To overcome this possible problem, FI-NNs can be used with the input layer FIs containing only the smallest possible number of generating invariants. 45,46 For the PIP-NN, NRPIP-NN and FI-NN fitting methods, several different NN architectures with two or three hidden layers have been tested. For each architecture, 50-200 different NNs have been optimized differing in the initial randomly assigned fitting parameters.…”

“…Because the number of the FIs increases mildly with the degree, as in the NRPIP mentioned above, the architecture of the FI-NN is more flexible resulting in a faster evaluation of the energy, in particular, for complicated systems, than for PIP-NNs. 46 All methods, PIP-NN, NRPIP-NN and FI-NN, employ invariant polynomials as the input layer of the NN. As systems become more and more complicated, it is harder and harder, and even impractical for high-dimensional systems beyond approximately 10 like atoms, to employ these NN approaches due to the vast increase in the number of the polynomials, and it is difficult to determine the least order of the PIPs.…”

“…As systems become more and more complicated, it is harder and harder, and even impractical for high-dimensional systems beyond approximately 10 like atoms, to employ these NN approaches due to the vast increase in the number of the polynomials, and it is difficult to determine the least order of the PIPs. In the HD-NNP method, on the other hand, such systems can directly be addressed with the same rigorous inclusion of the permutation invariance, although, with a few exceptions, [46][47][48][49][50][51] to date it has been mainly applied to condensed systems.…”