Bayesian Optimization of Hierarchical ZSM-5 for High-Efficiency Polyolefin Waste Recycling to Light Olefins

Efficient zeolitic recycling of polyolefin waste into C2–C4 light olefins offers an attractive route toward resource recovery; however, conventional zeolite catalysts suffer from severe mass transport limitations that lead to undesired secondary reactions and coke formation. Here, we report a “structure modulation and intelligent optimization” strategy to develop hierarchical ZSM-5 (H-ZSM-5), integrating tailored mesoporosity and crystal size control via Bayesian optimization (BO). Catalytic performance was evaluated in a two-stage system combining thermal pyrolysis and subsequent catalytic cracking. BO efficiently navigates this multivariate design space, identifying an optimal architecture (∼100 nm crystals, 4.0 nm mesopores) that achieved 92.6 wt % total gas yield and 85.1 wt % light olefin selectivity at 500 °C with a low catalyst-to-feed ratio (0.2). The nanoscale H-ZSM-5 also exhibited durability over 40 consecutive cycles and broad compatibility with postconsumer plastic mixtures. Mechanistic studies revealed that the synergy between nanoscale morphology and hierarchical porosity enhances mass transport and acid site accessibility, demonstrating that complex catalyst architectures can be precisely optimized via artificial intelligence strategies, such as BO.