Scientists from England have presented a new method in which AI quickly determines the properties of materials. The system works in the same way as a council of scientists.
The researchers explained that X-ray diffraction (XRD) analysis plays a critical role in the discovery of new materials. For example, it is used to analyze the crystal structures of new materials to find out where they can be applied. However, scientists note that these methods often require manual processing, they are time-consuming and not scalable. “New methods are needed to discover and optimize new materials faster through autonomous high-throughput experiments,” they note.
Now, a team at the University of Liverpool has unveiled a method where artificial intelligence can be used to quickly and accurately analyze X-ray data. This is an artificial intelligence agent called Crystallography Companion Agent (XCA). The XCA can perform offline phase identification from XRD data during measurement. The agent is suitable for both organic and inorganic materials. These measurements are made possible by large-scale simulations of physically correct X-ray diffraction data that are used to train the algorithm.
A unique feature of the agent, which the team has adapted to their task, is that he knows how to doubt. Usually, neural networks make the final decision, even if the data does not support a certain conclusion, while the scientist would not be sure of his conclusions and would discuss the results with other researchers.
“This decision-making process in the new model is similar to the discussion of scientists, only several AI models are involved,” the researchers note. “That said, the process doesn’t need manual, human-labeled data and works reliably with many sources of experimental complexity.”