Establishment of a general theory of metal-support interactions: Advances in catalytic reactions using AI

How can artificial intelligence (AI) help accelerate scientific discoveries based on vast amounts of experimental data? A new study by Professor Li Weixue’s team at the University of Science and Technology of China (USTC) at the Chinese Academy of Sciences , we show how this can be achieved with heterogeneous catalysts. The results were published in the journal Science.

Researchers establish a general theory of metal-support interaction (MSI), one of the most important pillars of catalysis, by integrating interpretable AI with experimental data, domain knowledge, and first-principles simulations I did.

Supported metal catalysts are widely used in petrochemical refining, industrial chemical production, and environmental control systems such as exhaust catalysts.

MSI helps stabilize dispersed catalysts and influences interfacial processes such as charge transfer, chemical composition, surrounding sites, particle morphology, and suboxide encapsulation. Therefore, MSI tuning is one of the few strategies to improve catalyst performance.

For example, strong MSI, originally used to describe the encapsulation of supported metal nanoparticles by a suboxide layer at high temperatures (discovered in 1978), has recently attracted much attention and has been It is believed to be the origin of the remarkable interfacial processes of

However, due to the (strong) nature of MSI and the complex interfaces and complex processes involved, fundamental questions remain regarding the impact of MSI on interfacial processes in general and encapsulation in particular.

In this study, Professor Lee’s team has made a breakthrough on this issue. Lee envisioned that there should be a simple formula to accurately describe and predict MSI strength. The team then compiled consistent experimental data from previous seminal work on 25 metals and 27 oxides.

The team used advanced interpretable machine learning algorithms, combined with domain knowledge and theoretical derivations, to identify equations that predict MSI from basic, easily available material parameters.

This simple equation reveals that the MSI strength is the sum of metal-metal and metal-oxygen interactions across the interface. Metal-metal interactions are a surprisingly new but previously unrecognized quantity that contributes significantly to MSI compared to the widely accepted metal-oxygen interactions.

The proposed formula shows remarkable universality. It can be applied not only to oxide-supported metal nanoparticle catalysts, but also to metal monatomic catalysts and metal-supported oxide catalysts.

This finding highlights that metal-metal interactions are a key component of support effectiveness and opens new avenues for understanding and engineering such supports.

Moreover, large-scale molecular dynamics simulations based on neural network potentials show that this metal-metal interaction also determines the rate of oxide encapsulation on the metal catalyst and the proportion of metal-metal bonding at the encapsulation interface. It became clear.

Based on these findings, the research team proposed a principle of strong metal-metal interactions to explain the occurrence of encapsulation. This not only explains almost all observed encapsulation phenomena, but also predicts new systems yet to be discovered.

The principle of strong metal-metal interactions in encapsulation can also be applied to the encapsulation of other metal compound supports and provides theoretical guidance for interfacial design and engineering.

“This result solves a major fundamental scientific problem in heterogeneous catalysis and brings immense potential to the design of efficient supported catalysts,” said Professor Li Yadong of Tsinghua University.

“This breakthrough is expected to accelerate the discovery of new catalytic materials and reactions, advancing the field of catalysis in energy, environment and materials science, thereby contributing to the sustainable development of society. ” said Professor Li Weixue.

This study demonstrates the potential of integrating interpretable AI algorithms with domain knowledge to build mathematical models and extract scientific principles based on vast amounts of historical experimental data. This study thus provides a new perspective on scientific discoveries in chemistry in the era of “AI for science.”