Machine Learning-Assisted Development of a Ga2O3-Modified Cu/SiO2 Catalyst for Efficient Hydrogenation of Dimethyl Oxalate to Ethylene Glycol

Low ethylene glycol (EG) selectivity and insufficient stability in the complex multistep hydrogenation of dimethyl oxalate (DMO) have invariably stimulated a research focus of copper-based catalysts. Conventional catalyst development heavily relies on extensive trial-and-error experimentation and serendipitous discovery. In order to efficiently predict high-performance catalysts, this study constructs a machine learning (ML) model based on the K-means clustering algorithm to successfully screen out the potential Ga promoter. Experimental results demonstrate that a 0.5Ga-Cu/SiO2 catalyst achieves nearly 100% DMO conversion and 98.59% EG selectivity at a relatively low temperature (180 °C) and high weight hourly space velocity (1.5 h−1), while maintaining continuous operational stability for over 240 h. Physical characterizations reveal that Ga2O3 promotes the dispersion of copper species and reduces local electron density, thereby modulating the balance between Cu0 and Cu+ dual active sites. Moreover, density functional theory calculations indicate that the interfacial interaction between Ga2O3 and copper species enhances the substrate dissociation adsorption. This study validates the utility of ML in screening advanced copper-based catalysts for efficient DMO hydrogenation to EG and provides a valuable framework for the future catalyst design in the thermocatalytic field.