Predicting the Reactivity of Acyclic Silylenes and Germylenes in Hydrogen Activation

Divalent group 14 element compounds (heavier carbenes) have been shown to serve as transition metal mimics and thus can activate small molecules and strong bonds at mild reaction conditions. Incorporating this special ability into novel transition metal-free catalysts requires careful tuning of the molecular properties. Finding the optimal combination of substituents is challenging and an experimentally highly deman¬ding process. In this work, we combine DFT and machine learning methods to predict the energy and activation barrier for the activation of dihydrogen by silylenes and germylenes, to thus facilitate the selection of candidates for small molecule activation in the future. We demonstrate that - based on the analysis of approx. 600 acyclic silylenes generated from 40 different substituents - the energy profiles can be reliably predicted by a simple model using the sum of energy increments of the substituents of the silylene. Furthermore, quantitative structure-activity relationships between the energies and the electronic and steric properties of the silylenes could be established, which enabled the prediction of the activation barrier with a mean average error of less than 9 kJ/mol. The model even enabled an extrapolation to silylenes with substituents not included in the original data set, and thus could predict new silylene structures for small molecule activation. Moreover, the reported procedure can also be applied to germylenes, and their energy profile quantitatively expressed as a function of the energies values of their silicon analogues. Overall, the presented protocol allows for a fast screening and prediction of possible candidates for H2 activation, which will accelerate the development of main group element catalysts in the future.