Dec 8, 2023

MIT developed ML-tool for material prediction beats human wisdom

Posted by in category: materials

The tool uses first-principle calculations to arrive at material behavior in a few thousand steps rather than a million possibilities.

How well did the system perform?

The researchers have not yet put the system under extensive analysis but have seen some encouraging results in the limited cases so far.

For instance, when the framework was applied to a strontium titanium oxide or SrTiO3, perovskite material that has been studied using conventional methods for over three decades, the researchers found two new atomic configurations that have not been described before. Interestingly, one of the configurations identified earlier was deemed to be unstable by the framework.

Leave a reply