A more flexible neural network (NN) method using the fundamental invariants (FIs) as the input vector is proposed in the construction of potential energy surfaces for molecular systems involving identical atoms. Mathematically, FIs finitely generate the permutation invariant polynomial (PIP) ring. In combination with NN, fundamental invariant neural network (FI-NN) can approximate any function to arbitrary accuracy. Because FI-NN minimizes the size of input permutation invariant polynomials, it can efficiently reduce the evaluation time of potential energy, in particular for polyatomic systems. In this work, we provide the FIs for all possible molecular systems up to five atoms. Potential energy surfaces for OH₃ and CH₄ were constructed with FI-NN, with the accuracy confirmed by full-dimensional quantum dynamic scattering and bound state calculations.

A global potential energy surface for the H₂ + OH ↔ H₂O + H reaction has been constructed using the neural networks method based on ∼17 000 ab initio energies calculated at UCCSD(T)-F12a/AVTZ level of theory. Time-dependent wave packet calculations showed that the new potential energy surface is very well converged with respect to the number of ab initio data points, as well as to the fitting process. Various tests revealed that the new surface is considerably more smooth and accurate than the existing YZCL2 and XXZ surfaces, representing the best available potential energy surface for the benchmark four-atom system. Equally importantly, the number of ab initio energies required to obtain the well converged potential energy surface is rather limited, indicating the neural network fitting is a powerful method to construct accurate potential energy surfaces for polyatomic reactions.

Access via Vibration: Molecular beam studies over the past decade have elucidated many subtle quantum mechanical factors governing the influence of vibrational excitation on the outcome of elementary chemical reactions. However, these studies have generally had to focus on reagents that can be easily made to vibrate by direct absorption in the infrared (IR). Wang et al. (p. 1499 ) show that a variation on stimulated Raman pumping can efficiently excite the IR-inactive stretch vibration in the diatomic molecule, hydrogen deuteride (HD). As a result, they can probe the influence of that vibration on the outcome of the HD + F reaction. Through a combined spectroscopic and theoretical investigation, they uncover Feshbach resonances along the reaction coordinate that are only accessible through vibrational preexcitation.

A few very brief pauses in the action: Chemical reactions proceed by the cumulative effect of trillions upon trillions of collisions between atoms and molecules. Usually, a given collision bounces the participants right back out again, either in their original form or with the atoms shuffled around into distinct products. In certain cases, the reacting partners experience a brief lull, termed a resonance, before they rearrange. Yang et al. report the discovery of particularly short-lived resonances in certain reactive collisions of chlorine atoms with vibrationally excited hydrogen deuteride (HD). Their results suggest that similar, as yet overlooked, resonances may lurk in other reactions of vibrationally excited molecules. Science , this issue p. 60

We report a new global potential energy surface of the HOCO system based on the F12 correction of unrestricted coupled-cluster with singles doubles and approximative triples using the augmented correlation-consistent polarized valence triple-zeta basis set (UCCSD(T)-F12/AVTZ), fitted by using the neural networks. Quantum dynamics calculations confirmed the satisfactory convergence of surface with respect to the number of data points and fitting process. It is found that the total reaction probabilities and complex-formation probabilities obtained on the present surface differ considerably with those obtained on the potential energy surface recently reported by Li et al. [J. Chem. Phys. 136, 041103 (2012)]. Various comparisons revealed that the present surface is substantially more accurate than that surface, representing the best available potential energy surface for this benchmark complex-forming four-atom system.

We report a permutationally invariant global potential energy surface (PES) for the H + CH₄ system based on ∼63 000 data points calculated at a high ab initio level (UCCSD(T)-F12a/AVTZ) using the recently proposed permutation invariant polynomial-neural network method. The small fitting error (5.1 meV) indicates a faithful representation of the ab initio points over a large configuration space. The rate coefficients calculated on the PES using tunneling corrected transition-state theory and quasi-classical trajectory are found to agree well with the available experimental and previous quantum dynamical results. The calculated total reaction probabilities (J_tot = 0) including the abstraction and exchange channels using the new potential by a reduced dimensional quantum dynamic method are essentially the same as those on the Xu-Chen-Zhang PES [Chin. J. Chem. Phys. 27, 373 (2014)].