Material scientists are making remarkable advances in the search for new materials with specific properties, and the University of Rochester is leading the way by using machine learning to fast-track their discovery. Crystalline materials have a well-ordered, repeating crystal lattice structure, and the specific arrangement of these lattices gives the material its properties. For instance, if a material needs to be strong, heat-resistant, and lightweight, the lattice must be just right in order to achieve the desired outcome.
To find out the properties of new materials, scientists use a process called X-ray diffraction (XRD). This involves taking a small sample of the material and grinding it into a fine powder, then exposing it to X-rays. As the X-rays hit the atoms in the material, they are diffracted in various directions depending on the atomic arrangement, creating a pattern on a detector that scientists need to analyze in order to infer the properties of the material.
The problem is that XRD produces huge amounts of data that are beyond the ability of humans to process efficiently. This is where machine learning comes in. Led by materials science PhD student Jerardo Salgado, a study developed deep learning models to automate the classification of materials based on their XRD patterns, using convolutional neural networks (CNNs) – a type of neural network that is excellent at performing image recognition and classification tasks. The models were trained on a large dataset of synthetic XRD patterns, representing a variety of experimental conditions and material types.
Project lead Niaz Abdolrahim, a mechanical engineering professor at the University of Rochester said, “There is a lot of materials science and physics hidden in each one of these images, and terabytes of data are being produced every day at facilities and labs worldwide.” He added, “Developing a good model to analyze this data can really help expedite materials innovation, understand materials at extreme conditions, and develop materials for different technological applications.”
Using machine learning models to filter XRD data could speed up the development of materials for faster electronics, better batteries, and even everyday items with enhanced durability, functionality, and sustainability. The research team at the Center for Matter at Atomic Pressures is particularly excited about the application of machine learning, as it could help them not only discover ways to create new materials, but also learn about the formation of stars and planets.
The use of AI to free scientific minds from the drudgery of data analysis will mean that their creative thinking can be better directed at designing the materials that will shape our future. Exciting times lie ahead for materials science! With machine learning to help, advances in the creation of new materials could be made at lightning speed.