(ORDO NEWS) — Rare earths are critical components in all kinds of electronics, from smartphones and broadband cables to wind turbines and electric vehicles. But finding useful compounds that can expand the practical use of rare earth elements is notoriously difficult, and the results are unpredictable.
Now, scientists have come up with a clever way to help find new rare earth compounds: A newly developed artificial intelligence system with predictive abilities that will allow us to go beyond what would be possible for people in the lab.
The type of artificial intelligence used here is machine learning: as the name suggests, this is when the software examines a database of information (in this case about rare earth compounds), recognizing patterns and correlations that then allow it to discover new potential matches for that database.
“Machine learning is really important here because when we talk about new compositions, ordered materials are very well known to everyone in the rare earth community,” says materials scientist Prashant Singh of the Ames Lab at Iowa State University.
“However, when you add confusion to known materials, it’s a different matter. The number of compositions gets much larger, often thousands or millions, and you can’t explore all possible combinations through theory or experimentation.”
In materials science, order and disorder refer to how particles are arranged in a material (e.g., in a perfect, crystalline lattice, or in a more chaotic, scattered arrangement), which directly affects the properties and applications of that material.
In this case, the machine learning model was built using a database of rare earth metals and some ideas from density functional theory (DFT), which deals with the analysis of the structure of materials, which is ideal for this kind of research.
The way the model is built means you can quickly test hundreds of permutations and then evaluate the phase stability of each one. In other words, the AI is able to determine if a combination of rare earth elements will be viable, such as if it will fall apart.
These calculations are then supplemented with additional information from the Internet – found using specially designed algorithms – and only after that they are verified and pass several checks to make sure that they remain in reality.
“It’s not really meant to open a specific compound,” says materials scientist Yaroslav Mudryk of the Ames lab. “It’s about how to develop a new approach or a new tool for detecting and predicting rare earths? And that’s what we’ve done.”
Experimental data can also be fed back into the machine learning system, further improving its accuracy and reducing the chance of errors such as creating rare earth compounds that don’t actually work.
At present, the model is still being evaluated and fine-tuned before moving on to search for rare earth compounds, but the researchers promise that this is just the beginning for the newly developed system.
Even better, the team’s methods should work in the future for other elusive types of material. After all, we won’t have to rely so heavily on chance to make these kinds of discoveries.
“Our approach will be useful for the discovery of new and complex rare earth compounds with new functional properties,” the researchers conclude in the published paper.
—
Online:
Contact us: [email protected]
Our Standards, Terms of Use: Standard Terms And Conditions.