Catalytic Methylation of <i>m</i>-Xylene, Toluene, and Benzene Using CO<sub>2</sub> and H<sub>2</sub> over TiO<sub>2</sub>-Supported Re and Zeolite Catalysts: Machine-Learning-Assisted Catalyst Optimization
Kah Wei Ting, Haruka Kamakura, Sharmin Sultana Poly, Motoshi Takao, S. M. A. Hakim Siddiki, Zen Maeno, Koichi Matsushita, Ken‐ichi Shimizu, Takashi Toyao
Abstract
Catalytic methylation of aromatic hydrocarbons using CO2 and H2 as a methylating agent was conducted over a combination of TiO2-supported Re (Re(1)/TiO2; Re = 1 wt %) and H-β (SiO2/Al2O3 = 40) in a batch reactor. Catalytic methylation of m-xylene was performed, and this catalyst combination demonstrated excellent performance for the synthesis of methylbenzenes, giving a high yield of total methylated products (10 and 57%, as calculated on the basis of CO2 and m-xylene, respectively), while generating relatively small amounts of byproducts such as demethylated and dearomatized products as well as CO and CH4 in the gas phase under the investigated reaction conditions (pCO2 = 1 MPa, pH2 = 5 MPa, T = 240 °C, t = 20 h). Our catalysts were also found to perform well for the methylation of toluene, providing a high yield and high selectivity for methylated products compared with the other investigated catalyst combinations. In addition to conducting conventional-type catalyst research, we used a data science approach based on machine learning techniques to identify important input variables that govern the catalytic performance, enabling optimization of the catalyst for the methylation reaction. Compared with the catalysts optimized using the conventional approach, the improved Re/TiO2 catalyst with a Re loading amount of 1.8 wt %, which was optimized with the aid of ML, exhibited greater activity toward the methylation of benzene using CO2/H2.