The cover artwork represents the H2-activation transition state with Vaska’s complex, and the machine learning models (Gaussian processes) used to predict the energy barrier. The cover is associated to the recently published article “Machine learning dihydrogen activation in the chemical space surrounding Vaska's complex”. This work was the fruit of an interdisciplinary collaboration between the Chemistry Dpt. at the Univ. of Toronto (Pascal Friederich, Gabriel dos Passos Gomes and Alán Aspuru-Guzik), the Dpt. of Mathematics at the Univ. of Oslo (Riccardo De Bin) and the Hylleraas Centre (David Balcells). The cover artwork was made by the co-author Gabriel dos Passos Gomes.
In this article, the authors showed that models based on neural networks and Gaussian processes could predict the H2-activation barriers for a large family of iridium complexes (ca. 2000) derived from Vaska’s complex. The highlight of this work is the possibility of achieving high accuracy (ca. 1 kcal/mol) with Gaussian processes trained with small datasets computed at the DFT level. The use of molecular fingerprints in the features representing the complexes facilitated the chemical interpretation of the predictions.
The article is open-access, and the datasets are freely available from both GitHub and Dataverse.