A breakthrough in materials science AI marks a significant shift in how research facilities can approach autonomous discovery. According to a study published in Nature Materials, a research team led by Taeghwan Hyeon, PhD, director of the Center for Nanoparticle Research at the Institute for Basic Science, has developed an artificial intelligence framework that discovers catalysts by integrating knowledge from entirely different material families. This approach aims to address a persistent bottleneck in automated discovery: the restriction of traditional machine learning models to narrow, predefined material domains.
Deep learning model breaks catalyst family boundaries
Discovering new catalysts is one of the central challenges in developing clean-energy technologies such as green hydrogen production. A major hurdle in this field is the oxygen evolution reaction, a slow and energy-intensive step that occurs during water electrolysis. Traditionally, catalyst discovery has remained confined within individual material families, limiting the ability to transfer knowledge across chemically distinct systems.
To bridge these gaps, the research team developed a deep learning model called the Crossbreeding Neural Network. The model learned simultaneously from two distinct catalyst groups: single-atom catalysts supported on carbon materials, and perovskite oxide catalysts. Single-atom catalysts help reveal how individual metal atoms behave on surfaces, while perovskite oxides provide data on how bulk crystal structures influence performance. By cross-referencing these data streams, the artificial intelligence model predicted the performance of a completely new, unexplored material class: single-atom catalysts supported on perovskite oxides.
Validation through experimental synthesis and screening
The framework does not merely rely on computational predictions. The researchers experimentally synthesized and tested 12 of the predicted catalysts within the new material family. The model correctly predicted the activity ranking of all 12 candidates, suggesting that it learned transferable relationships rather than simply memorizing the training data. According to co-first author Junseok Moon, PhD, the model was capable of judging performance variations even within a material family it had never encountered before.
Following this validation, the team expanded the approach to screen multimetallic catalysts containing several different single-metal atoms simultaneously. The Crossbreeding Neural Network computationally screened 8,008 catalyst candidates. It successfully identified a highly promising multimetallic single-atom catalyst containing tungsten, molybdenum, ruthenium, and rhodium atoms anchored on a calcium–praseodymium cobalt iron oxide perovskite support.
Using explainable artificial intelligence techniques, the team visualized how specific atomic environments influence catalytic activity. This allowed them to identify key chemical factors strongly related to activity across both families, including oxidation state, ionic radius, valence d-electron count, electronegativity, and coordination number.
Bridging heterogeneous datasets in laboratory operations
For laboratory directors managing research and development facilities, this development could help improve high-throughput screening workflows and reduce the overhead associated with custom algorithmic development. The work suggests that future AI systems may be able to leverage knowledge across multiple chemical domains rather than requiring separate models for each material family.
Moon noted that when artificial intelligence learns the common language shared across different material families, it can suggest entirely new design directions beyond candidate spaces predefined by humans. The findings demonstrate how AI models can transfer insights between chemically distinct material systems, potentially expanding the scope of autonomous materials discovery. The researchers suggest that similar approaches could eventually be applied to other areas involving the integration of heterogeneous datasets, including battery materials, energy-storage systems, and potentially drug discovery.
This article was created with the assistance of Generative AI and has undergone editorial review before publishing.
