ADVANCED MATERIALS & PROCESSES | NOVEMBER/DECEMBER 2025 5 RESEARCH TRACKS SUPERCOMPUTERS SUPPORT AI MATERIALS DISCOVERY With help from supercomputers at the DOE’s Argonne National Laboratory, researchers at the University of Cambridge are developing AI tools that automatically mine scientific journal articles to build structured materials databases. These datasets are then used to train specialized language models designed to streamline materials research. “The aim is to have something like a digital assistant in your lab,” says Jacqueline Cole, head of molecular engineering at Cambridge. “A tool that complements scientists by answering questions and offering feedback to help steer experiments and guide their research.” Cole’s recent work has focused on developing smaller, faster, and more efficient AI models to support materials research, without the massive computing costs typically required to train large language models (LLMs) from scratch. To bypass this costly pretraining process, Cole and her colleagues developed a method for generating a large question and answer (Q&A) dataset from domain-specific materials data. Using new algorithms and their ChemDataExtractor tool, they converted a database of photovoltaic materials into hundreds of thousands of Q&A pairs. This process, known as knowledge distillation, captures detailed materials information in a form that off-the-shelf AI models can easily ingest. Cole’s team used the Q&A pairs to fine-tune smaller language models, which went on to match or outperform much larger models trained on general text, achieving up to 20% higher accuracy in domain-specific tasks. While their study focused on solar-cell materials, the team believes the approach could be applied broadly to other research areas. anl.gov. SOFTNESS IN AMORPHOUS MATERIALS Researchers in Japan from the University of Osaka, the National Institute of Advanced Industrial Science and Technology (AIST), Okayama University, and the University of Tokyo are figuring out why glass and other amorphous materials deform more easily in some regions than others. By applying a mathematical method called persistent homology, the team demonstrated that these soft regions are governed by hidden hierarchical structures, where ordered and disordered atomic arrangements coexist. Amorphous materials including glass, rubber, and certain plastics lack the long-range order of crystalline solids. However, they are not completely random, as they possess medium-range order (MRO), subtle atomic patterns that extend over a few nanometers. MRO has long been suspected to play a critical role in determining the physical properties of amorphous materials, particularly their mechanical responses. Yet due to the complexity of atomic networks, conventional analysis methods have been unable to clarify how MRO relates to regions that deform more easily than their surroundings. The structural origins of mechanical softness in amorphous solids remained elusive until now. In amorphous silicon, the team discovered hierarchical ring structures— smaller rings with irregular edge lengths nested inside larger rings. This coexistence of order and disorder means that softness emerges not from randomness alone, but from constraints imposed by MRO interwoven with local disorder. The discovery establishes a clear principle: Mechanically soft regions arise where disorder is embedded within MRO. This finding provides a practical guideline for developing amorphous solids that are both flexible and strong, suitable for applications from displays and coatings to energy devices. www.osaka-u.ac.jp. Schematic shows how scientific literature is mined using ChemDataExtractor to build materials databases. Persistence diagram obtained from the structure of amorphous silicon, examples of local ring structures corresponding to each point in the diagram, and representative structures including atoms with large nonaffine displacements.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==