How can artificial intelligence (AI) machine learning models be used to identify new materials? This is what a recent study published in Nature hopes to address as a team of researchers investigated using machine learning algorithms to broaden the scope of discovering and studying new materials. This study has the potential to help researchers develop novel methods for discovering new materials while reducing the labor that has long been required to discover and study them.
For the study, the researchers presented Copilot for Real-world Experimental Scientists (CRESt), which consists of several machine learning models designed to incorporate several more variables than are traditionally used for identifying new materials, including chemical compositions and microstructural images, just to name a few. To enhance the models, the researchers combined them with a method known as Bayesian optimization, which is a statistical method employed to fine-tune machine learning models, enabling them to complete tasks more efficiently.
In the end, CRESt successfully identified a new material formula that could provide almost 10 times the power density per dollar over palladium, which is traditionally an expensive material to identify and extract.
“A significant challenge for fuel-cell catalysts is the use of precious metal,” said Zhen Zhang, who is a PhD student at the Massachusetts Institute of Technology and lead author of the study. “For fuel cells, researchers have used various precious metals like palladium and platinum. We used a multielement catalyst that also incorporates many other cheap elements to create the optimal coordination environment for catalytic activity and resistance to poisoning species such as carbon monoxide and adsorbed hydrogen atom. People have been searching low-cost options for many years. This system greatly accelerated our search for these catalysts.”
This study comes as AI is increasingly being used throughout science, including the discovery of exoplanets by automating the data analysis process and reducing the amount of labor that scientists would traditionally spend analyzing data.
How will machine learning help scientists discover and study new materials in the coming years and decades? Only time will tell, and this is why we science!
As always, keep doing science & keep looking up!
Sources: Nature, EurekAlert!