Machine learning activation energies of chemical reactions

Abstract

Application of machine learning (ML) to the prediction of reaction activation barriers is a new and exciting field for these algorithms. The works covered here are specifically those in which ML is trained to predict the activation energies of homogeneous chemical reactions, where the activation energy is given by the energy difference between the reactants and transition state of a reaction. Particular attention is paid to works that have applied ML to directly predict reaction activation energies, the limitations that may be found in these studies, and where comparisons of different types of chemical features for ML models have been made. Also explored are models that have been able to obtain high predictive accuracies, but with reduced datasets, using the Gaussian process regression ML model. In these studies, the chemical reactions for which activation barriers are modeled include those involving small organic molecules, aromatic rings, and organometallic catalysts. Also provided are brief explanations of some of the most popular types of ML models used in chemistry, as a beginner's guide for those unfamiliar.

Efficient and Accurate Description of Diels‐Alder Reactions Using Density Functional Theory**

“…[22,23] Another possibility is also to train and/or use a machine learning algorithm for chemical reactions. [24] While some attempts to use this technique have been performed to predict activation energies, [25] it is at a preliminary stage and will need a huge data-set. Very recently, some reactive force-fields were developed with a first application also to a simple Diels-Alder reaction, but, at the present stage, this method needs a specific parametrization for each reaction.…”

A Look Inside the Black Box of Machine Learning Photodynamics Simulations

“…ML models learn the relationship between molecular structures and associated properties (e.g., energies, 10−13 transition dipole moments, 10−13 and oscillator strengths 14,15 ) from a large QC data set. The applications have succeeded in the discovery of new molecules 16 and materials, 17 predicting reaction barriers, 18 finding transition states, 19,20 solving the Schrodinger equation, 21,22 modeling wave functions, 23,24 optimizing density functionals, 25,26 computing IR spectra, 27 UV−vis spectra, 11,13 and NMR spectra, 28 and simulating excited-state dynamics 12,29 and complex photochemical reactions. 1−3 The energy prediction achieves chemical accuracy (1 kcal•mol −1 ), which is required for realistic chemical prediction and comparison with experiments.…”

Prediction of Transition State Structures of General Chemical Reactions via Machine Learning

“…molecular energy and activation barrier) are accurately predicted. [20][21][22][23] Despite those advances of ML techniques, a prediction of 3-dimensional(D) molecular structure is not easily achievable because the inferred 3D structures need to satisfy not only the permutation invariance with respect to the atom order but also physical (rotational and translational invariances) and Euclidean (triangle inequality) constraints.…”