2020 Virtual AIChE Annual Meeting

(106b) Machine Learning Quantum Tunneling in the Kinetics of Chemical Reactions

Author

The computation of reaction rate constants has been tackled using a multitude of theories ranging from classical transition state theory1 to the fully quantum flux-flux correlation function approach.2 The main cost of these methods is related to the need to explore potential energy surfaces, PES, either at the level of geometry optimizations, minimum energy path searches or to describe dynamics in time. In this context, machine learning algorithms which accelerate the search for reactive pathways or the prediction of rate constants are of great interest.3,4 We have constructed a database of over 130000 reaction rate constants5 obtained via quantum calculations of the reaction rate constants for particles crossing one dimensional single and double symmetric and asymmetric barriers. A deep neural network, DNN, was trained on the database with the ADAM optimizer and used to predict reaction rate constants. The most significant input features were identified using the Pearson correlation coefficient, and a grid search analysis was carried out for a coarse optimization of the DNN model hyperparameters. The network was trained for hundreds of epochs. The predicted results from the test set are in good agreement with the exact values. This approach shows promise for the application of machine learning to predict reaction rate constants.
  1. Pechukas, P. Transition State Theory. Annu. Rev. Phys. Chem. 32, 159–177 (1981).
  2. Tromp, J. W. & Miller, W. H. The reactive flux correlation function for collinear reactions H + H2, Cl + HCl and F + H2. Faraday Discuss. Chem. Soc. 84, 441–453 (1987).
  3. Amabilino, S. et al. Training neural nets to learn reactive potential energy surfaces using interactive quantum chemistry in Virtual Reality. J. Phys. Chem. A 123, 4486–4499 (2019).
  4. Ulissi, Z. W. et al. Machine-learning methods enable exhaustive searches for active Bimetallic facets and reveal active site motifs for CO2 reduction. ACS Catal. 7, 6600–6608 (2017).
  5. Valleau, S. Machine learning quantum tunneling in the kinetics of chemical reactions. In preparation. (2020).