US solves 100-year-old mystery of nanocrystals’ atomic structure

For more than a century, scientists have relied on crystallography—analyzing X-ray diffraction patterns—to uncover the atomic structures of materials. This method revolutionized fields from medicine to materials science, famously enabling the discovery of DNA’s double helix.

Yet, crystallography has had a persistent flaw: it works best on large, pure crystals. When only tiny, imperfect nanocrystals are available, the method falls short, leaving the structure of countless materials unknown.

Columbia Engineering researchers used machine learning to mend these persistent issues. Their new algorithm enables the reconstruction of the atomic structure of materials from degraded diffraction patterns of fragments of nanocrystals. A feat that was previously deemed impossible has come true.

“The AI solved this problem by learning everything it could from a database of many thousands of known, but unrelated, structures,” says Simon Billinge, professor of materials science and applied physics and applied mathematics at Columbia Engineering.

“Just as ChatGPT learns the patterns of language, the AI model learned the patterns of atomic arrangements that nature allows.”

Crystallography’s longstanding limitation

Traditional X-ray diffraction techniques rely on pristine, large crystals to generate clear diffraction patterns rich with atomic information. When researchers are limited to powders or suspensions of nanocrystals, the patterns are too degraded to resolve the structure using conventional methods.

This shortcoming has hindered progress in areas ranging from drug development to battery technology to archaeology, where only small or damaged samples are often available.

The Columbia team turned to diffusion generative modeling, an AI technique popularized by image generators like Midjourney and Sora. They trained their model on a dataset of 40,000 known atomic structures, purposely scrambling these structures’ order to teach the AI how to create meaningful order from chaos.

In training, the AI learned to pair poorly resolved diffraction data with the most probable atomic arrangements, witnessing myriad crystal structures throughout the training and reconstructing figures.

These constructs were further polished in a process called Rietveld refinement, which aligned them more precisely to the diffraction data.

“From previous work, we knew that diffraction data from nanocrystals doesn’t contain enough information to yield the result,” Billinge said. “The algorithm used its knowledge of thousands of unrelated structures to augment the diffraction data.”

A leap for material science

“The powder crystallography challenge is a sister problem to the famous protein folding problem where the shape of a molecule is derived indirectly from a linear data signature,” said Hod Lipson, who, with Billinge, co-proposed the study.

The project carries special meaning for Lipson, whose grandfather, Henry Lipson, pioneered early computational crystallography techniques nearly a century ago.

“What particularly excites me is that with relatively little background knowledge in physics or geometry, AI was able to learn to solve a puzzle that has baffled human researchers for a century. This is a sign of things to come for many other fields facing long-standing challenges,” he said.

The algorithm successfully reconstructed the atomic structures of nanocrystals that had previously stumped researchers. The achievement represents a major step forward, potentially unlocking innovations across numerous fields that rely on structural analysis.

Gabe Guo, who led the project as a senior at Columbia, said, “When I was in middle school, the field was struggling to build algorithms that could tell cats from dogs. Now, studies like ours underscore the massive power of AI to augment the power of human scientists and accelerate innovation to new levels.”

The study has been published in Nature Materials.

[ad_1]

_{Images are for reference only.Images and contents gathered automatic from google or 3rd party sources.All rights on the images and contents are with their legal original owners.}

[ad_2]