Developing new materials, like metallic glasses, has traditionally been a slow and costly process by trial-and-error experiments. But what if we could map the hidden relationships among elements to guide our search?
A research group, led by Dr. Yuan-Chao Hu from Songshan Lake Materials Laboratory, has done just that. In a study published in National Science Review , they constructed the ‘material networks’ for metallic glasses—materials that are as strong as metal but can be shaped like plastics. In this network, similar to social network, elements are ‘friends’ (nodes) and a developed alloy among them is a ‘friendship’ (link or triangle).
By analyzing this network’s structure, the team found that fully connected groups of elements, or ‘cliques,’ are excellent predictors for new, multi-component metallic glasses. This insight can drastically reduce the vast search space of possible combinations of chemical elements.
The most striking finding came from creating a ‘dynamic network’ that includes the history of metallic glass discovery. They found that many ‘new’ alloys developed over recent decades were already encoded in the network structure from the earlier data.
“This suggests that traditional research, guided by existing knowledge and intuition, can fall into an ‘innovation trap,’” said Yuan-Chao Hu, the corresponding author of the study. “We might be unconsciously confirming what the network already knows, rather than finding truly novel combinations.”
The study also revealed that the material networks have unique rules. Unlike the scale-free internet network, the metallic glass network is ‘abnormally scale-free.’ This is because it grows under its inherent ‘physical constraints’—there is a limited number of elements, but the number of combinations (metallic glasses) between them grows much faster.
This network-driven approach offers a new pathway for navigating the complex world of material discovery, potentially accelerating the development of metallic glasses and other advanced materials for future technologies.
National Science Review
Data/statistical analysis