If you have an ACS member number, please enter it here so we can link this account to your membership. (optional)

ACS values your privacy. By submitting your information, you are gaining access to C&EN and subscribing to our weekly newsletter. We use the information you provide to make your reading experience better, and we will never sell your data to third party members.


Inorganic Chemistry

AI finds single-molecule magnets

Researchers use deep learning to find single-molecule magnets in a haystack of crystal structures

by Fionna Samuels
March 8, 2024

Line structure of a dysprosium-salen coordination complex.
A metal-salen complex this new deep learning program accurately predicted to be a single-molecule magnet from a pool of 20,000 crystal structures

The binary nature of single-molecule magnets—metal coordination complexes with switchable magnetic polarity—makes them good candidates for designing new technologies like miniaturized memory storage devices and quantum computers. Each material’s crystal structure determines its specific properties and potential uses. But first, scientists must find a single-molecule magnet find among a sea of other, nonmagnetic metal complexes.

Now a team at the Tokyo University of Science has designed an artificial intelligence program “able to predict the [magnetic] properties of a metal complex solely from images,” according to lead investigator Takashiro Akitsu (IUCrJ 2024, DOI: 10.1107/S2052252524000770). His team created a training dataset by sifting through 800 papers describing metals coordinated with salen-type ligands—organic ligands with N2O2donor sites—and converting the complexes’ crystal structures into 3D images. Some complexes exhibited magnetism, while others did not. After training on this dataset, the deep learning algorithm was able to correctly identify single-molecule magnets from a pool of 20,000 metal complexes with similar, salen-type ligands in the Cambridge Structural Database.

“I see this as an augmenting tool rather than a tool that’s going to substitute for a solid-state chemist,” says Prasanna Balachandran, a computational materials scientist at the University of Virginia unaffiliated with the work. However, he says it’s possible that a deep learning program was unnecessary because there are other less computationally intensive machine learning techniques available. Ultimately, Balachandran hopes this work nucleates more collaboration between the data and materials science communities. By embracing the technology, he says, researchers could quickly and cheaply discover new single-molecule magnets.



This article has been sent to the following recipient:

Chemistry matters. Join us to get the news you need.