In an unprecedented scientific breakthrough, a team of researchers from Columbia University in the United States has managed to solve a mystery that has puzzled scientists for over a hundred years—thanks to artificial intelligence. This milestone could pave the way for major advances in technology and materials science.
The scientists successfully identified the precise atomic structure of nanocrystals—ultra-small particles widely used in electronics, advanced materials development, archaeology, and artifact analysis.
These tiny particles have long posed a challenge due to their lack of regular crystalline structure, which rendered traditional X-ray diffraction techniques ineffective. Those techniques rely on large, well-ordered crystals to produce distinct patterns that reveal atomic arrangements. In contrast, nanocrystals scatter rays in complex patterns that are difficult to decode.
Nevertheless, the research team developed a new algorithm called PXRDnet, powered by advanced AI, capable of analyzing unclear diffraction patterns using a database of tens of thousands of crystal structures.
By training the algorithm on these patterns, the team was able to infer the probable atomic arrangements of the studied materials without needing direct physical information. Professor Simon Billinge expressed his amazement at AI’s ability to learn the laws permitted by nature—enabling it to overcome barriers that have hindered scientists for decades.
The researchers also highlighted that the algorithm can analyze nanocrystals as small as 10 angstroms—ten thousand times thinner than a human hair—marking a significant leap forward in understanding materials at the atomic level without the need for expensive tools or perfect crystals.
Team leader Gabi Go noted that using AI in this context represents a tremendous evolution, pointing out that not long ago, the same technology struggled to tell a cat from a dog.
Professor Hod Lipson added that what’s truly impressive is AI’s ability to solve complex problems that have eluded human experts for over a century—despite lacking physical intuition or scientific knowledge.
This achievement confirms that AI is more than just an assistive tool—it is a powerful force capable of delivering groundbreaking scientific solutions across multiple fields, opening new horizons for discoveries that may reshape scientific research as we know it.