Scientists have developed a pioneering artificial intelligence-driven technology that reveals the hidden movements of nanoparticles, which are crucial in materials science, pharmaceuticals and electronics. By combining artificial intelligence with electron microscopy, researchers can now visually see atomic-level changes that were previously obscured by noise. This breakthrough provides a clearer understanding of how these tiny particles behave under various conditions, potentially revolutionizing industrial processes and scientific discoveries.

Scientists have developed a new method to reveal how nanoparticles move and change over time. These tiny particles play a vital role in industries such as pharmaceuticals, electronics and energy. The breakthrough, published in Science, combines artificial intelligence with electron microscopy to create detailed visuals of how nanoparticles react under different conditions.

"Nanoparticle-based catalytic systems have a huge impact on society," explains Carlos Fernandez-Granda, professor of mathematics and data science and director of the Center for Data Science at NYU and one of the paper's authors. "It is estimated that 90% of manufactured products involve a catalytic process at some point in their production chain. We have developed an artificial intelligence method that opens a new window into the exploration of atomic-scale structural dynamics in materials."

Left, platinum nanoparticles imaged by electron microscopy. The data have sufficient spatial resolution to display individual atoms. However, due to the high temporal resolution, these images are severely corrupted by noise, but this is necessary to observe the fundamental dynamic behavior of the nanoparticle surface related to its functionality. The picture on the right is the output of an artificial intelligence system that effectively removes noise and reveals the atomic structure of nanoparticles. Image credit: Peter Crozier and Joshua Vincent of Arizona State University and Carlos Fernandez Granda and Srias Mohan of New York University

The research, conducted in collaboration with scientists from Arizona State University, Cornell University and the University of Iowa, blended electron microscopy with artificial intelligence. This powerful combination allows scientists to observe molecular structure and motion in unprecedented detail and speed, down to billionths of a meter.

Peter A. Crozier, professor of materials science and engineering at Arizona State University and one of the authors of this paper, explains: "Electron microscopes can capture images with high spatial resolution, but because the atomic structure of nanoparticles changes extremely quickly during chemical reactions, we need to collect data at extremely high speeds to understand their function. This results in extremely noisy measurements. We developed an artificial intelligence method that can learn how to automatically eliminate this noise, allowing for the visualization of key atomic-level dynamics."

Observing the movement of atoms on nanoparticles is critical to understanding functionality in industrial applications. The problem is that few atoms are visible in the data, so scientists can't determine how they behave -- the equivalent of tracking objects in a video shot at night with an old-fashioned camera. To solve this problem, the authors trained a deep neural network (the computing engine of artificial intelligence) that can "light up" electron microscope images, revealing the underlying atoms and their dynamic behavior.

"The nature of particle changes is extremely diverse, including periods of flux, manifested by rapid changes in atomic structure, particle shape, and orientation; understanding these dynamic changes requires new statistical tools," explained David S. Matteson, professor and associate chair of the Department of Statistics and Data Science at Cornell University, director of the National Institute of Statistical Sciences, and one of the authors of the paper. "This study introduces a new statistical method that uses topological data analysis to quantify fluxes and track the stability of particles as they transition between ordered and disordered states."

Compiled from /ScitechDaily