Advances in Gasoline Hydrodesulfurization Catalysts: The Role of Structure-Activity Relationships and Machine Learning Approaches

In response to increasingly stringent environmental regulations, reducing the sulfur content in transportation fuels has become a global priority. This review article provides a comprehensive analysis of recent advances in gasoline hydrodesulfurization (HDS) catalysts, with a particular emphasis on the interplay between structure-activity relationships (SAR) and emerging machine learning (ML) methodologies. The discussion begins with an overview of the fundamentals of gasoline HDS, highlighting the unique challenges associated with achieving deep sulfur removal while preserving fuel octane quality. ML accelerates HDS catalyst design by optimizing MoS(2) morphology (edge/corner ratios) to balance direct desulfurization (DDS) and hydrogenation (HYD). ML models decode synthesis-structure-activity relationships, prioritize key parameters (Co/Mo ratios, sulfidation conditions), and guide experimental iterations, enabling rapid discovery of catalysts with high sulfur removal and minimal olefin loss. Subsequently, the review explores how ML approaches, including random forests, support vector machines (SVM), and deep neural networks, are revolutionizing catalyst design by effectively capturing the complex, nonlinear relationships among multiple reaction parameters. Representative case studies illustrate the successful integration of experimental data with ML models, demonstrating enhanced predictive capabilities and process optimization. Finally, the article discusses current challengessuch as limited high-quality data and the complexity of industrial feedstocksand outlines future research directions aimed at bridging the gap between laboratory-scale innovations and industrial applications.