Posted in | News | Materials Research

Accelerating Scientific Discovery with AI

How can scientific discoveries based on large volumes of experimental data be accelerated by artificial intelligence (AI)? This can be achieved in heterogeneous catalysis, according to a recent study led by Prof. Weixue Li from the University of Science and Technology of China (USTC) of the Chinese Academy of Sciences, published in Science.

The researchers developed a comprehensive theory of metal-support interaction (MSI), a key aspect of catalysis, by combining interpretable AI with domain knowledge, experimental data, and first-principles simulations.

Supported metal catalysts are widely used in industrial chemical production, petrochemical refining, and environmental control systems like exhaust catalysts. MSI influences interfacial activities, such as charge transfer, chemical composition, perimeter sites, particle shape, and suboxide encapsulation, in addition to stabilizing dispersed catalysts. As a result, modifying MSI is one of the few ways to enhance catalyst performance.

For example, strong MSIs have recently gained attention as they are believed to drive many important interfacial processes. They were first used to characterize the encapsulation of supported metal nanoparticles by suboxide layers at elevated temperatures, a discovery made in 1978.

Due to the complex processes and delicate interfaces involved, there are still fundamental uncertainties regarding the nature of (strong) MSIs and their impact on interfacial processes in general, and encapsulation in particular.

Prof. Li's team achieved a breakthrough on this topic in this study. Li hypothesized the existence of a simple formula to precisely characterize and predict the strength of MSI. The team then compiled reliable experimental data from earlier groundbreaking studies that covered 27 oxides and 25 metals.

The researchers developed a formula to predict MSI based on simple, accessible material properties by combining domain expertise, theoretical derivation, and advanced interpretable machine-learning techniques.

This formula revealed that the total metal-metal and metal-oxygen interactions across the interface determine MSI strength. Interestingly, the metal-metal interaction—an unexpectedly significant factor that had not been recognized before—contributes much more to MSI than the well-known metal-oxygen interaction.

The universality of the suggested formula is impressive. It can be applied to metal single-atom catalysts, metal-supported oxide catalysts, and oxide-supported metal nanoparticle catalysts. This discovery opens the door to a new understanding and engineering of support effects by emphasizing the critical role that metal-metal contact plays.

According to additional large-scale molecular dynamics simulations based on neural network potentials, the metal-metal interaction also determines the proportion of metal-metal bonds at the encapsulation interface and the kinetic rates of oxide encapsulation over metal catalysts.

Based on these findings, the team proposed a strong metal-metal interaction principle that explains encapsulation. This principle accounts for nearly all reported encapsulation events and predicts the emergence of novel systems that have yet to be identified. Other metal compound supports can also be encapsulated using this principle, providing theoretical guidance for interface engineering and design.

On the basis of these results, the group put out a strong metal-metal interaction principle that describes encapsulation and not only accounts for almost all reported encapsulation events but also predicts the emergence of novel systems that have not yet been identified. Other metal compound supports can also be encapsulated using the strong metal-metal interaction principle, which offers theoretical direction for interface engineering and design.

Yadong Li, Professor, Tsinghua University

Prof. Weixue Li, a Professor at the University of Science and Technology of China, added, “This breakthrough is expected to accelerate the discovery of new catalytic materials and reactions, advancing the field of catalysis in energy, environment, and material science, thereby contributing to the sustainable development of society.

This study demonstrates the possibility of combining domain expertise and interpretable AI algorithms to create mathematical models and derive scientific principles from enormous volumes of historical experimental data. In this approach, the study provides a new viewpoint on scientific discovery in chemistry in the era of “AI for Science.”

Journal Reference:

Wang, T. et. al. (2024) Nature of metal-support interaction for metal catalysts on oxide supports. Science. doi.org/10.1126/science.adp6034

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.