Predicting Rate Constants of Hydroxyl Radical Reactions with Alkanes Using Machine Learning

Abstract

The hydrogen abstraction reactions of the hydroxyl radical with alkanes play an important role in combustion chemistry and atmospheric chemistry. However, site-specific reaction constants are difficult to obtain experimentally and theoretically. Recently, machine learning has proved its ability to predict chemical properties. In this work, a machine learning approach is developed to predict the temperature-dependent site-specific rate constants of the title reactions. Multilayered neural network (NN) models are developed by training the site-specific rate constants of 11 reactions, and several schemes are designed to improve the prediction accuracy. The results show that the proposed NN models are robust in predicting the site-specific and overall rate constants.

Fast Predictions of Reaction Barrier Heights: Toward Coupled-Cluster Accuracy

“…Conventional quantum chemistry and TST workflow for predicting a high-pressure limit rate coefficient is shown by the dark arrows. Several methods have been proposed in literature (and in the present work, shown by orange arrows) to accelerate steps in the process: (a) Initial conformer generation; ,− ,− (b) TS guess generation; (c) estimating barrier heights from 2D representation; ,,,,− (d) estimating rate coefficient from 2D representation; ,,− (e) semiempirical optimization; (f) TS guess generation; − ,− (g) Estimating barrier heights from 3D representation; , (h) estimating rate coefficient from 3D representation; , (i) accelerating TS optimization; − (j) estimating barrier height from unoptimized TS guess; (k) estimating rate coefficient from optimized TS; (l) estimating barrier height from optimized TS. , …”

“…In other work, Komp et al trained a FFN on ∼1.5 million data points to predict the product of k ∞ ( T ) with the reactant partition function for 1D barrier problems. FFNs have also been fit to the rate coefficients from small data sets of hydroxyl radical reactions , and ionic liquids . In all these cases, the data sets employed did not cover much reaction space.…”

Supervised Machine Learning Algorithms for Predicting Rate Constants of Ozone Reaction with Micropollutants

“…In recent years, driven by the fast development of artificial intelligence and data science, multiple machine learning methods, such as MLR, SVM, DT, RF, DNN, extremely randomized trees (ET), ridge regression (RR), and artificial neural network (ANN) have been employed in diverse fields and shown the strength in calculating performance. − It includes but not limited to planning chemical syntheses, designing new materials and catalysts, , evaluating the properties of chemicals, − predicting reaction performance, calculating the kinetic parameter , and so on. In assessing the environmental risk of compounds, the MLR method is commonly used to develop QSPR models, the nonlinear machine learning models are scarce.…”

Transfer Learning Approach to Multitarget Temperature-Dependent Reaction Rate Prediction

“…The use of artificial neural networks (ANNs) on fixed chemical descriptors has been applied to room-temperature gas-phase OH reactions . More recently, ANNs were used to predict temperature-dependent site-specific rates for alkanes + OH reactions; however, the study was limited to only 11 reactions . Machine learning models on fixed fingerprints were also used for the prediction of rate constants of aqueous OH and SO 4 reactions. , Performance of such models on small data sets could be enhanced by combining small data sets and transferring knowledge as recently shown for SO 4 , HClO, O 3 , and ClO 2 .…”