University of Science and Technology of China (USTC)announced that Prof. Weixue Li's group utilized artificial intelligence (AI) has achieved important results in basic research on catalysis.
Interpretable artificial intelligence reveals the nature of "metal-carrier interactions".
This study establishes a controlling equation between the metal-carrier interaction and the fundamental properties of the material in experimental data through interpretable AI techniquesIn this work, the essential factors determining the metal-carrier interactions were revealed, a principled criterion for strong metal-metal interactions was proposed, and the challenge of metal catalysts coated with oxide carriers was solved.
Note: Explainable Artificial Intelligence (XAI) refers to the ability of an intelligent body to communicate clearly and effectively with users, affected persons, decision makers, developers, etc. of an AI system in a way that is explainable, comprehensible, and interactive between humans and computers, in order to gain human trust and at the same time meet regulatory requirements.
This latest study aggregates experimental interfacial interaction data from multiple papers covering 25 metals and 27 oxides. Using interpretable AI algorithms, the study constructs a feature space of up to 30 billion expressions through iterative mathematical operations using material properties as the basic features. The study utilizes compression-aware algorithms, combined with domain knowledge and theoretical derivation, to filter out physically clear and numerically accurate descriptors, and establish the controlling equations between metal-carrier interactions and material properties.
The above results will help optimize the design of catalysts with high activity, high selectivity and high stability.Promising to accelerate the discovery of new catalytic materials, new catalytic reactionsThe company's mission is to contribute to the green upgrading and sustainable development of energy, environment and materials.
at the same time,This study demonstrates that interpretable AI algorithms can build mathematical models from experimental data, unearth hidden physical laws, build theories with predictive power, and accelerate the process of scientific principle discovery.It will promote the deep integration of AI technology and chemistry research, providing new perspectives and possible solutions for realizing important scientific problems and technological innovation breakthroughs.
