Machine Learning-Accelerated Prediction of Lewis Acid Site Positioning for Long-Chain Mono-olefin Aromatization on Zn 2+ /HZSM-5 Catalysts

作者：Jing Wang, Yusheng Jia, Rui Li, Yupeng Zhang, Yupeng Zhang, Caiping Ma, Yang Zhang, Yang Zhang, Riguang Zhang, Xiaofeng Li, Baojun Wang, Lixia Ling · 发表于：ACS Catalysis · 年份：2025 · DOI：10.1021/acscatal.5c00459 · 被引用次数：10 · 研究领域：Zeolite Catalysis and Synthesis、Metal-Organic Frameworks: Synthesis and Applications、Machine Learning in Materials Science

Long-chain mono-olefin aromatization is an important reaction during synthesis of coal to aromatics, in which the positioning of Lewis acid in ZSM-5 zeolite determines the performance. This study aims to reveal the relationship between the location of Lewis acid and catalytic performance during the aromatization of C6–C8 mono-olefins and to rapid-screen the optimized positioning of Lewis acid to enhance the efficiency and selectivity of aromatic production. Based on density functional theory (DFT) to obtain high-quality data sets of C6 and C7 dehydrogenation reactions in Zn 2+ /HZSM-5 catalysts, a data-driven ML model is developed to predict the C8 dehydrogenation energy barriers and screen the Lewis acid sites T5-T3 or T11-T3 with high dehydrogenation activity, while DFT confirms the accuracy of the results. The BTX (benzene, toluene, and xylene) selectivity reached 64.1% among various products via microkinetic analysis after investigating the cyclization process and cracking reactions on the Lewis acid site positioning, where the dehydrogenation activity of the catalyst is optimal. The six-membered ring is found to be the key cyclic intermediate due to the pore confinement effect, which effectively stabilizes cyclic intermediates, and toluene and p -xylene show a small steric effect through weak interactions analysis. This can provide important clues and research paradigms for the rational design of ZSM-5 zeolite modified with different metal species.