Zeolites are crucial in industrial catalysis, renowned for their unique microporous structures and versatile catalytic properties. However, accurately simulating zeolite-catalyzed processes poses significant challenges due to their spatiotemporal complexity, which requires capturing both atomic-level interactions and macroscopic phenomena. This review examines recent advancements in realistic simulations of zeolite catalytic processes, focusing on techniques such as machine learning potentials (MLPs), enhanced sampling methods, and kinetic Monte Carlo (KMC) simulations. These computational strategies have substantially improved the accuracy and efficiency of catalytic reaction simulations, addressing the traditional limitations associated with complex systems like zeolites. MLPs offer precise potential energy surfaces at lower computational costs, enabling extended molecular dynamics simulations. Enhanced sampling techniques, including umbrella sampling and metadynamics, effectively explore rare events and complex energy landscapes, although their success depends on the careful selection of collective variables (CVs). KMC simulations further enhance our understanding by modeling long-term molecular events, such as diffusion and reaction kinetics, at larger spatial and temporal scales. Despite notable progress, challenges remain, particularly regarding CV selection and KMC’s reliance on accurate first-principles data. The integration of machine learning approaches, such as automated CV selection and transfer learning for MLP refinement, presents promising solutions to these issues. This review highlights these advancements and their potential to revolutionize the study of zeolite catalytic processes, bridging the gap between theoretical modeling and experimental observations and contributing to the design of more effective and sustainable catalysts.

中文翻译：

---

迈向沸石催化过程的真实模拟：迷你回顾


沸石在工业催化中至关重要，以其独特的微孔结构和多功能催化性能而闻名。然而，由于其时空复杂性，准确模拟沸石催化的过程带来了重大挑战，这需要捕获原子级相互作用和宏观现象。本综述研究了沸石催化过程真实模拟的最新进展，重点介绍了机器学习电位 （MLP）、增强采样方法和动力学蒙特卡洛 （KMC） 模拟等技术。这些计算策略大大提高了催化反应模拟的准确性和效率，解决了与沸石等复杂系统相关的传统限制。MLP 以较低的计算成本提供精确的势能表面，从而支持扩展的分子动力学模拟。增强的采样技术，包括伞式采样和元动力学，有效地探索了罕见事件和复杂的能源景观，尽管它们的成功取决于对集体变量 （CV） 的仔细选择。袋鼠妈妈再生模拟通过在更大的空间和时间尺度上模拟长期分子事件（例如扩散和反应动力学）来进一步增强我们的理解。尽管取得了显著进展，但挑战仍然存在，尤其是在简历选择和 KMC 对准确第一性原理数据的依赖方面。机器学习方法的集成，例如用于 MLP 细化的自动 CV 选择和迁移学习，为这些问题提供了有前途的解决方案。 本文重点介绍了这些进步及其彻底改变沸石催化过程研究的潜力，弥合了理论建模和实验观察之间的差距，并有助于设计更有效和可持续的催化剂。