Spotlight: Using AI Discovery Engine for Materials Exploration

NIST researcher Gilad Kusne collaborated with the University of Maryland researchers to use machine learning to investigate different materials based on their composition and temperature.

How did they do this? They used an AI algorithm called the Autonomous Materials Search Engine (AMASE) to fire X-ray beams at a wafer, which then bounced off it (shown here).

The wafer, composed of the elements tin and bismuth, contains hundreds of thousands of different ratios of those elements. Scientists call this a combinatorial library.

Each time the algorithm analyzes a point in the wafer, it’s investigating a different material. The process is like picking a book from the library, where each point on the wafer represents a different book to read.

And since researchers are also analyzing factors like temperature, there are now almost an infinite number of materials or books to investigate.

AMASE can investigate these different materials because it’s able to extrapolate or make predictions beyond the available dataset, which, according to Gilad, is something most AI systems are not inherently good at.

Understanding how various types of materials behave under different conditions can lead to a wide range of applications, like automated manufacturing and quantum materials.

Learn more about the method.