Accelerating Hierarchical ZSM-5 Engineering via Bayesian Optimization-Guided Discovery

Abstract

Bayesian optimization (BO) was used to accelerate the synthesis of hierarchical ZSM-5 with a balanced micro–mesoporous structure. Using mesoporosity and microporosity as dual objectives, a Gaussian process regression surrogate guided three successive BO iterations informed by 15 initial experiments. The optimized sample (HZ-R0.55T50t2) exhibited the highest hierarchy factor (0.17), featuring similar mesoporosity (0.44) but higher microporosity (0.38) than the benchmark, indicating reduced diffusional resistance with preserved framework. Characterizations revealed that NaOH induced framework dissolution, whereas TPAOH moderated desilication; their mixture synergistically created uniform mesopores while maintaining crystallinity. 27Al and 29Si NMR confirmed realumination and silanol generation, and acidity analysis showed redistributed acid sites. Sensitivity analysis identified the TPAOH fraction and temperature as the dominant factors influencing the hierarchy factor. This study establishes a data-driven workflow for zeolite design, demonstrating that BO effectively accelerates hierarchical structure optimization while minimizing the experimental effort.