Now, a scientific team led by scientists from the SLAC National Accelerator Laboratory, the National Institute of Standards and Technology (NIST) and Northwestern University of the United States Department of Energy has reported that they have found a shortcut to discover and improve metallic glass, and can develop new materials with less time and cost.
200 times faster to discover new materials
Ideally, when two or three metals are fused together, an alloy that looks like metal will be obtained, and its atoms will be arranged into rigid geometry.
Using a new system combined with machine learning in Stanford synchrotron radiation light source, the scientific team can quickly screen hundreds of sample materials, enabling the team to discover the composition of metallic glass made of three new mixtures, which is 200 times faster than before.
Chris Wolfton, a professor at Northwestern University, is a pioneer in the use of computers and artificial intelligence to predict new materials and one of the collaborators of the paper. He said that it usually takes 10 or 20 years for new materials to complete the process from discovery to commercial use. “This achievement has greatly shortened the time spent in the discovery of new materials.”
The prospect of material science will change
In the past half century, scientists have only studied the composition of about 6000 kinds of metallic glass, and this new system can produce and screen 20000 kinds of components.
Although other teams are also using machine learning prediction to find different kinds of metallic glasses, this time the scientists quickly verified and predicted through experiments, and then circulated the results to the next round of machine learning and experiments, which is the unique feature of this progress.
In fact, this method can be used in various experiments, especially in finding materials, such as metallic glass and catalysts. Jason Hitch-Shimur, NIST material research engineer, said that AI will change the prospects of material science.
Provide practical tools for global scientists
This paper is the first scientific achievement of this project funded by the US Department of Energy. SLAC is cooperating with Citrine Informatics, a Silicon Valley AI company, to change the way of finding new materials and provide practical tools for scientists around the world.
The company was founded by former graduate students from Stanford University and Northwestern University. They created a material science data platform, in which the data in electronic forms and laboratory notes are stored in a consistent format, so it can be used for the learning of artificial intelligence systems.
Recently, the speed of evaluating new materials is very slow. Even though five potential types of metallic glass can be detected every day, it still takes a thousand years to study every possible combination of metallic glass to overcome toxic and expensive ingredients, or remove fragile properties.
Wolfton said that the ultimate goal is to enable scientists to obtain the direct feedback results in the machine learning model and prepare another set of samples to be tested in the next day or even the next hour.