Machine learning of activation energy prediction for extended element

Abstract

Activation energy is an important kinetic parameter in determining the reaction rate, mechanism, and product. Estimating the activation energy of a reaction is a very complex and time-consuming task both experimentally and computationally. Despite the activation energy can be quickly predicted through machine learning, there are significant limitations in the type of reaction or elements included in the reaction. In this dissertation, the predictable elements are expanded by generating 4,552 reaction data on involving silicon, yield MAE of 3.09 kcal mol$^{-1}$ and RMSE of 5.33 kcal mol$^{-1}$. In addition, the model is calibrated by the uncertainty quantification of prediction results.