Machine learning perspective: Revealing deep mechanisms and new advances in adsorption and catalysis of gaseous molecules
Yue Zhu, Fengyu Gao, Lei Yi, Honghong Yi, Qingjun Yu, Shunzheng Zhao, Yuansong Zhou, Ya Wang, Xiaolong Tang
Abstract
In the field of chemical and environmental engineering, research on the adsorption and catalytic processes of gas molecules has long been constrained by the efficiency bottleneck of traditional trial and error methods and empirical analysis. With the rapid development of artificial intelligence technology, machine learning (ML), leveraging its powerful data processing and pattern recognition capabilities, has provided a revolutionary tool for material design and optimization. This paper systematically reviews the latest progress of ML in the research of gas-molecule adsorption and catalytic materials, focusing on three core aspects: formulation screening, material property correlation, and descriptor construction. By integrating experimental conditions and material characteristics, ML models can efficiently predict adsorption/catalytic performance and uncover key descriptors to guide the design of new materials. In formulation screening, ML can identify key factors from complex multivariable systems and optimize the composition, synthesis methods, and reaction conditions of catalysts and adsorbents. Regarding material property correlation, it can deeply analyze the internal relationships between the physical-chemical and reaction properties of materials. In the field of descriptor construction, innovative designs of electronic, elemental, and structural descriptors offer a new perspective for understanding material properties and catalytic reactions. Although ML shows great potential in this field, it also faces many challenges, such as data quality and scarcity, model interpretability and generalization ability, and descriptor universality. This paper not only reviews key application cases of ML in gas pollutant treatment but also proposes the urgent need for interdisciplinary collaboration and the construction of standardized databases. Despite being in the early stage of research, the ML-driven Materials Genome Initiative has demonstrated the potential to revolutionize traditional research and development models, opening up a new path for solving environmental and energy problems.