Highly efficient PCCP/Cr(III) ethylene tetramerization catalyst designed using artificial neural network approach

Chromium complexes are feasible for inducing ethylene selective oligomerization to generate valuable linear α-olefins, such as 1-hexene and 1-octene. The current theoretical calculation techniques can effectively accelerate the design of molecular catalysts. In this work, we combined density functional theory (DFT) and artificial neural network (ANN) methods to model vinyl-bridged PCCP/Cr active species and used an ANN-based model to predict and screen the designed novel PCCP ligands. DFT was used to optimize the structure of active species and calculate electronic properties. The datasets were established using steric descriptors, electronic descriptors, and experimental data. The ANN-based models were trained based on the established datasets. We found that the optimized ANN-based models ANN-HT and ANN-OT, respectively for the prediction of 1-hexene and 1-octene selectivity, provide satisfactory predictive performance, with R² values of 0.962 for ANN-HT and 0.976 for ANN-OT. We investigated the contribution of steric and electronic descriptors on catalytic performance, and the results showed that steric descriptors have a greater contribution to controlling product selectivity compared to electronic descriptors. The similarity between the ethylene oligomerization test results of the experimentally prepared L²⁸/Cr catalyst and the ANN-based model prediction results proved the accuracy of the models. The experimentally prepared L²⁸/Cr exhibited catalytic activity of 2181 kg∙g⁻¹∙h⁻¹ and 65.8 % 1-C₈ selectivity under suitable reaction conditions.