Although Ziegler–Natta (ZN) catalysts play a major role in the polyolefin market, a true understanding of their properties at the molecular level is still missing. In particular, there is a lack of knowledge on the electronic properties of Ti sites. Theoretical calculations predict that the electron density of the Ti sites in the precatalysts correlates with the activation energy for olefin insertion in the Ti-alkyl bond generated at these sites after activation by Al-alkyls. It is also well known that the effective charge on the Ti sites in the activated catalysts affects the olefin π-complexation. In this contribution, we exploit two electronic spectroscopies, UV–vis and Ti L_2,3-edge near-edge X-ray absorption fine structure (NEXAFS), complemented with theoretical simulation to investigate three ZN precatalysts of increasing complexity (up to an industrial system) and the corresponding catalysts activated by triethylaluminum (TEAl). We provide compelling evidence for the presence of monomeric 6-fold-coordinated Ti⁴⁺ species in all of the precatalysts, which however differ in the effective charge on the Ti sites. We also unambiguously demonstrate that these sites are reduced by TEAl to two types of monomeric 5-coordinated Ti³⁺, either alkylated or not, and that the former are involved in ethylene polymerization. In addition, small TiCl₃ clusters are formed in the industrial catalyst, likely due to the occurrence of severe reducing conditions within the catalyst pores. These data prove the potential of these two techniques, coupled with simulation, in providing an accurate description of the electronic properties of heterogeneous ZN catalysts.

The empirical aspect of descriptor design in catalyst informatics, particularly when confronted with limited data, necessitates adequate prior knowledge for delving into unknown territories, thus presenting a logical contradiction. This study introduces a technique for automatic feature engineering (AFE) that works on small catalyst datasets, without reliance on specific assumptions or pre-existing knowledge about the target catalysis when designing descriptors and building machine-learning models. This technique generates numerous features through mathematical operations on general physicochemical features of catalytic components and extracts relevant features for the desired catalysis, essentially screening numerous hypotheses on a machine. AFE yields reasonable regression results for three types of heterogeneous catalysis: oxidative coupling of methane (OCM), conversion of ethanol to butadiene, and three-way catalysis, where only the training set is swapped. Moreover, through the application of active learning that combines AFE and high-throughput experimentation for OCM, we successfully visualize the machine’s process of acquiring precise recognition of the catalyst design. Thus, AFE is a versatile technique for data-driven catalysis research and a key step towards fully automated catalyst discoveries.