Quantum and quasi-classical dynamics of the OH + CO → H + CO2reaction on a new permutationally invariant neural network potential energy surface

Exploring the frontiers of condensed-phase chemistry with a general reactive machine learning potential

Atomistic simulation has a broad range of applications from drug design to materials discovery. Machine learning interatomic potentials (MLIPs) have become an efficient alternative to computationally expensive ab initio simulations. For this reason, chemistry and materials science would greatly benefit from a general reactive MLIP, that is, an MLIP that is applicable to a broad range of reactive chemistry without the need for refitting. Here we develop a general reactive MLIP (ANI-1xnr) through automated sampling of condensed-phase reactions. ANI-1xnr is then applied to study five distinct systems: carbon solid-phase nucleation, graphene ring formation from acetylene, biofuel additives, combustion of methane and the spontaneous formation of glycine from early earth small molecules. In all studies, ANI-1xnr closely matches experiment (when available) and/or previous studies using traditional model chemistry methods. As such, ANI-1xnr proves to be a highly general reactive MLIP for C, H, N and O elements in the condensed phase, enabling high-throughput in silico reactive chemistry experimentation.

High vibrational excitation of the reagent transforms the late-barrier H + HOD reaction into an early-barrier reaction

Polanyi's rules predict that a late-barrier reaction yields vibrationally cold products; however, experimental studies showed that the H2 product from the late-barrier H + H2O(|04⟩-) and H + HOD(vOH = 4) reactions is vibrationally hot. Here, we report a potential-averaged five-dimensional state-to-state quantum dynamics study for the H + HOD(vOH = 0-4) → H2 + OD reactions on a highly accurate potential energy surface with the total angular momentum J = 0. It is found that with the HOD vibration excitation increasing from vOH = 1 to 4, the product H2 becomes increasingly vibrationally excited and manifests a typical characteristic of an early barrier reaction for vOH = 3 to 4. Analysis of the scattering wave functions revealed that vibrational excitation in the breaking OH bond moves the location of dynamical saddle point from product side to reactant side, transforming the reaction into an early barrier reaction. Interestingly, pronounced oscillatory structures in the total and product vibrational-state-resolved reaction probabilities were observed for the H + HOD(vOH = 3, 4) reactions, in particular at low collision energies, which originate from the Feshbach resonance states trapped in the bending/torsion excited vibrational adiabatic potential wells in the entrance region due to van der Waals interactions.

Theoretical Insights into the Dynamics of Gas-Phase Bimolecular Reactions with Submerged Barriers

Much attention has been paid to the dynamics of both activated gas-phase bimolecular reactions, which feature monotonically increasing integral cross sections and Arrhenius kinetics, and their barrierless capture counterparts, which manifest monotonically decreasing integral cross sections and negative temperature dependence of the rate coefficients. In this Perspective, we focus on the dynamics of gas-phase bimolecular reactions with submerged barriers, which often involve radicals or ions and are prevalent in combustion, atmospheric chemistry, astrochemistry, and plasma chemistry. The temperature dependence of the rate coefficients for such reactions is often non-Arrhenius and complex, and the corresponding dynamics may also be quite different from those with significant barriers or those completely dominated by capture. Recent experimental and theoretical studies of such reactions, particularly at relatively low temperatures or collision energies, have revealed interesting dynamical behaviors, which are discussed here. The new knowledge enriches our understanding of the dynamics of these unusual reactions.

Full-dimensional quantum mechanical calculations for the tunneling behavior of HOCO dissociation to H + CO2

The tunneling behavior during HOCO dissociation to H + CO₂ was investigated by full-dimensional quantum mechanical calculations based on an accurate global potential energy surface. The tunneling lifetimes for the low-lying 1500 vibrational states were calculated using the low-storage filter diagonalization method after a 1 million-step Chebyshev propagation. In the calculated energy range, the lifetimes of different vibrational states with similar energy are found to differ by 3-4 orders of magnitude, and the lower limit for these tunneling lifetimes is consistent with the experimental results reported by Continetti et al. For the given vibrational progressions, the lifetime of the vibrational states decreases with the increasing energy level, which is consistent with the results of 1D simulation calculations reported by Bowman, but the declining curve obviously fluctuates, and the declining slope is significantly different from that obtained by 1D simulation. Due to a difference in the effective barrier width, the mode-specific behavior of the tunneling effect is manifested in that the O-C-O' and H-O-C bend modes lead to the largest enhancement and an inhibitory effect on the tunneling process, respectively.

Machine Learning for Chemical Reactions

Machine learning (ML) techniques applied to chemical reactions have a long history. The present contribution discusses applications ranging from small molecule reaction dynamics to computational platforms for reaction planning. ML-based techniques can be particularly relevant for problems involving both computation and experiments. For one, Bayesian inference is a powerful approach to develop models consistent with knowledge from experiments. Second, ML-based methods can also be used to handle problems that are formally intractable using conventional approaches, such as exhaustive characterization of state-to-state information in reactive collisions. Finally, the explicit simulation of reactive networks as they occur in combustion has become possible using machine-learned neural network potentials. This review provides an overview of the questions that can and have been addressed using machine learning techniques, and an outlook discusses challenges in this diverse and stimulating field. It is concluded that ML applied to chemistry problems as practiced and conceived today has the potential to transform the way with which the field approaches problems involving chemical reactions, in both research and academic teaching.

An accurate full-dimensional potential energy surface for the reaction OH + SO → H + SO2

An accurate full-dimensional PES for the OH + SO ↔ H + SO₂ reaction is developed by the permutation invariant polynomial-neural network approach.

Neural Network Potential Energy Surfaces for Small Molecules and Reactions

We review progress in neural network (NN)-based methods for the construction of interatomic potentials from discrete samples (such as ab initio energies) for applications in classical and quantum dynamics including reaction dynamics and computational spectroscopy. The main focus is on methods for building molecular potential energy surfaces (PES) in internal coordinates that explicitly include all many-body contributions, even though some of the methods we review limit the degree of coupling, due either to a desire to limit computational cost or to limited data. Explicit and direct treatment of all many-body contributions is only practical for sufficiently small molecules, which are therefore our primary focus. This includes small molecules on surfaces. We consider direct, single NN PES fitting as well as more complex methods that impose structure (such as a multibody representation) on the PES function, either through the architecture of one NN or by using multiple NNs. We show how NNs are effective in building representations with low-dimensional functions including dimensionality reduction. We consider NN-based approaches to build PESs in the sums-of-product form important for quantum dynamics, ways to treat symmetry, and issues related to sampling data distributions and the relation between PES errors and errors in observables. We highlight combinations of NNs with other ideas such as permutationally invariant polynomials or sums of environment-dependent atomic contributions, which have recently emerged as powerful tools for building highly accurate PESs for relatively large molecular and reactive systems.

Universal crossed beam imaging studies of polyatomic reaction dynamics

Crossed-beam imaging studies of polyatomic reactions show surprising dynamics not anticipated by extrapolation from smaller model systems.

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

Several symmetry strategies have been compared in fitting full dimensional accurate potentials for reactive systems based on a neural network approach.

Combined Experimental-Theoretical Study of the OH + CO → H + CO2 Reaction Dynamics

A combined experimental-theoretical study is performed to advance our understanding of the dynamics of the prototypical tetra-atom, complex-forming reaction OH + CO → H + CO₂, which is also of great practical relevance in combustion, Earth's atmosphere, and, potentially, Mars's atmosphere and interstellar chemistry. New crossed molecular beam experiments with mass spectrometric detection are analyzed together with the results from previous experiments and compared with quasi-classical trajectory (QCT) calculations on a new, full-dimensional potential energy surface (PES). Comparisons between experiment and theory are carried out both in the center-of-mass and laboratory frames. Good agreement is found between experiment and theory, both for product angular and translational energy distributions, leading to the conclusion that the new PES is the most accurate at present in elucidating the dynamics of this fundamental reaction. Yet, small deviations between experiment and theory remain and are presumably attributable to the QCT treatment of the scattering dynamics.

Full-dimensional quantum mechanics calculations for the spectroscopic characterization of the isomerization transition states of HOCO/DOCO systems

Full-dimensional quantum mechanics calculations were performed to determine the vibrational energy levels of HOCO and DOCO based on an accurate potential energy surface. Almost all of the vibrational energy levels up to 3500 cm^-1 from the vibrational ground state were assigned, and the calculated energy levels in this work are well in agreement with the reported results by Bowman. The corresponding full dimensional wavefunctions present some special features. When the energy level approaches the barrier height, the trans-HOCO and cis-HOCO states strongly couple through tunneling interactions, and the tunneling interaction and Fermi resonance were observed in the DOCO system. The energy level patterns of trans-HOCO, cis-HOCO and trans-DOCO provide a reasonable fitted barrier height using the fitting formula of Field et al., however, a discrepancy exists for the cis-DOCO species which is considered as a random event. Our full-dimensional calculations give positive evidence for the accuracy of the spectroscopic characterization model of the isomerization transition state reported by Field et al., which was developed from one-dimensional model systems. Furthermore, the special case of cis-DOCO in this work means that the isotopic substitution can solve the problem of the accidental failure of Field's spectroscopic characterization model.

A neural network potential energy surface for the F + CH4reaction including multiple channels based on coupled cluster theory

We report here a new global and full dimensional potential energy surface (PES) for the F + CH₄reaction.

Accurate potential energy surfaces for hydrogen abstraction reactions: A benchmark study on the XYG3 doubly hybrid density functional

The potential energy surface (PES) for the H + CH₄ system has been constructed with the recently developed XYG3 doubly hybrid functional, while those with the standard B3LYP hybrid functional, and the Møller-Plesset perturbation theory up to the second order (MP2) are also presented for comparison. Quantum dynamics studies demonstrated that satisfactory results on the reaction probabilities and the rate coefficients can be obtained on top of the XYG3-PES, as compared to the results based on the highly accurate, yet expensive, CCSD(T)-PES (Li et al., J. Chem. Phys. 2015, 142, 204302). Further investigation suggested that the XYG3 functional is useful in providing accurate rate coefficients for some larger systems involving H atom abstractions. © 2017 Wiley Periodicals, Inc.

Quantum Chemical Study of Supercritical Carbon Dioxide Effects on Combustion Kinetics

In oxy-fuel combustion, the pure oxygen (O₂), diluted with CO₂ is used as oxidant instead air. Hence, the combustion products (CO₂ and H₂O) are free from pollution by nitrogen oxides. Moreover, high pressures result in the near-liquid density of CO₂ at supercritical state (sCO₂). Unfortunately, the effects of sCO₂ on the combustion kinetics are far from being understood. To assist in this understanding, in this work we are using quantum chemistry methods. Here we investigate potential energy surfaces of important combustion reactions in the presence of the carbon dioxide molecule. All transition states and reactant and product complexes are reported for three reactions: H₂CO + HO₂ → HCO + H₂O₂ (R1), 2HO₂ → H₂O₂ + O₂ (R2), and CO + OH → CO₂ + H (R3). In reaction R3, covalent binding of CO₂ to the OH radical and then the CO molecule opens a new pathway, including hydrogen transfer from oxygen to carbon atoms followed by CH bond dissociation. Compared to the bimolecular OH + CO mechanism, this pathway reduces the activation barrier by 5 kcal/mol and is expected to accelerate the reaction. In the case of hydroperoxyl self-reaction 2HO₂ → H₂O₂ + O₂ the intermediates, containing covalent bonds to CO₂ are found not to be competitive. However, the spectator CO₂ molecule can stabilize the cyclic transition state and lower the barrier by 3 kcal/mol. Formation of covalent intermediates is also discovered in the H₂CO + HO₂ → HCO + H₂O₂ reaction, but these species lead to substantially higher activation barriers, which makes them unlikely to play a role in hydrogen transfer kinetics. The van der Waals complexation with carbon dioxide also stabilizes the transition state and reduces the reaction barrier. These results indicate that the CO₂ environment is likely to have a catalytic effect on combustion reactions, which needs to be included in kinetic combustion mechanisms in supercritical CO₂.

Dynamics of transient speciesviaanion photodetachment

Recent experimental and theoretical advances in transient reaction dynamics probed by photodetachment of polyatomic anions are reviewed.

Recent advances in quantum scattering calculations on polyatomic bimolecular reactions

This review surveys quantum scattering calculations on chemical reactions of polyatomic molecules in the gas phase published in the last ten years.

Crossed beam polyatomic reaction dynamics: recent advances and new insights

This review summarizes the developments in polyatomic reaction dynamics, focusing on reactions of unsaturated hydrocarbons with O-atoms and methane with atoms/radicals.

Chemical Reaction CO+OH(•) → CO2+H(•) Autocatalyzed by Carbon Dioxide: Quantum Chemical Study of the Potential Energy Surfaces

The supercritical carbon dioxide medium, used to increase efficiency in oxy combustion fossil energy technology, may drastically alter both rates and mechanisms of chemical reactions. Here we investigate potential energy surface of the second most important combustion reaction with quantum chemistry methods. Two types of effects are reported: formation of the covalent intermediates and formation of van der Waals complexes by spectator CO2 molecule. While spectator molecule alter the activation barrier only slightly, the covalent bonding opens a new reaction pathway. The mechanism includes sequential covalent binding of CO2 to OH radical and CO molecule, hydrogen transfer from oxygen to carbon atoms, and CH bond dissociation. This reduces the activation barrier by 11 kcal/mol at the rate-determining step and is expected to accelerate the reaction rate. The finding of predicted catalytic effect is expected to play an important role not only in combustion but also in a broad array of chemical processes taking place in supercritical CO2 medium. It may open a new venue for controlling reaction rates for chemical manufacturing.

Recent Advances in Quantum Dynamics of Bimolecular Reactions

In this review, we survey the latest advances in theoretical understanding of bimolecular reaction dynamics in the past decade. The remarkable recent progress in this field has been driven by more accurate and efficient ab initio electronic structure theory, effective potential-energy surface fitting techniques, and novel quantum scattering algorithms. Quantum mechanical characterization of bimolecular reactions continues to uncover interesting dynamical phenomena in atom-diatom reactions and beyond, reaching an unprecedented level of sophistication. In tandem with experimental explorations, these theoretical developments have greatly advanced our understanding of key issues in reaction dynamics, such as microscopic reaction mechanisms, mode specificity, product energy disposal, influence of reactive resonances, and nonadiabatic effects.

Quantum dynamics of polyatomic dissociative chemisorption on transition metal surfaces: mode specificity and bond selectivity

Recent advances in quantum dynamical characterization of polyatomic dissociative chemisorption on accurate global potential energy surfaces are critically reviewed.

High-Level, First-Principles, Full-Dimensional Quantum Calculation of the Ro-vibrational Spectrum of the Simplest Criegee Intermediate (CH2OO)

The ro-vibrational spectrum of the simplest Criegee intermediate (CH2OO) has been determined quantum mechanically based on nine-dimensional potential energy and dipole surfaces for its ground electronic state. The potential energy surface is fitted to more than 50 000 high-level ab initio points with a root-mean-square error of 25 cm(-1), using a recently proposed permutation invariant polynomial neural network method. The calculated rotational constants, vibrational frequencies, and spectral intensities of CH2OO are in excellent agreement with experiment. The potential energy surface provides a valuable platform for studying highly excited vibrational and unimolecular reaction dynamics of this important molecule.