Emory University Physicists Use AI to Discover New Physical Laws

ByMason Reed

April 28, 2026

Researchers at Emory University utilized a custom neural network to uncover hidden non-reciprocal forces in dusty plasma, challenging long-standing theoretical assumptions about the fourth state of matter.

In a significant departure from the ‘black box’ nature of modern artificial intelligence, researchers at Emory University have demonstrated that machine learning can do more than just process data—it can reveal the fundamental laws of nature. By focusing on dusty plasma, often called the fourth state of matter, a team led by physicists Justin Burton and Ilya Nemenman has successfully modeled complex particle interactions with over 99% accuracy, overturning several long-held theoretical assumptions in the process.

The study, published in PNAS and supported by the National Science Foundation and the Simons Foundation, utilized a custom-built neural network to analyze the motion of particles within ionized gas. Unlike traditional AI models that require massive datasets, this physics-based network was designed to operate with limited experimental data. It focused on ‘non-reciprocal forces,’ a phenomenon where one particle influences another differently than it is influenced in return—much like the asymmetrical wake created by two boats moving across a lake.

Experimental physicist Justin Burton and lead author Wentao Yu, now a postdoc at Caltech, developed a tomographic laser-sheet imaging system to track dozens of particles in three dimensions. This high-precision tracking allowed the AI to identify that a leading particle in a plasma field attracts a trailing one, while the trailing particle repels the leader. Furthermore, the AI corrected common misconceptions regarding charge-size proportionality and how quickly forces decay over distance, proving that these relationships are far more complex than scientists previously believed.

The implications of this research extend far beyond the laboratory. Dusty plasma is found throughout the universe, from the rings of Saturn to the moon’s surface, where it causes dust to cling to astronaut suits. On Earth, these charged particles appear in wildfires and can disrupt critical radio communications. By understanding the collective motion of these particles, researchers hope to apply the same AI framework to industrial materials like inks and paints, and even to biological systems such as the movement of metastatic cancer cells.

While the AI performed the heavy lifting of the calculations on a standard desktop computer, the researchers emphasized that human oversight remains the bedrock of scientific integrity. The project required over a year of refining the neural network’s structure to ensure it followed physical rules while remaining free to explore unknown variables. As the technology matures, it offers a decentralized tool for innovation that respects the traditional scientific method while accelerating the pace of discovery in both the physical and life sciences.

Leave a Reply

Your email address will not be published. Required fields are marked *