Methylnaphthalene Isomerization and Diffusion Behavior over H-BEA Zeolites Using Molecular Dynamics Simulations with Machine Learning Potentials

The monomolecular isomerization of methylnaphthalene, which is crucial for understanding alkyl aromatic ring reactions in the petrochemical industry, has been studied using ab initio and machine learning-based molecular dynamics simulations. The diffusional free energy landscapes suggest that the free energy barriers for methylnaphthalene translocation between straight channels are comparable when either both channels share a Brønsted acid site or neither possesses one. Furthermore, ab initio molecular dynamics simulations coupled with metadynamics provided detailed free energy landscapes for the monomolecular isomerization reaction network. The methyl transfer process is identified as the most feasible pathway, representing the rate-limiting step with a maximum free energy barrier of 124 kJ/mol. Interestingly, the more stable Wheland intermediate does not participate in the predominant reaction pathway. Utilizing eXtreme Gradient Boosting and Shapley Additive exPlanations, a strong correlation between the charges of molecular moieties and free energy barriers has been established. The present study provides significant insights into the mechanisms underlying the isomerization reactions of polyaromatics, which are pivotal in the development of efficient zeolite catalysts for petrochemical processes.