Physicists at Emory University used a custom neural network to uncover modern physical laws governing dusty plasma, a system of ionized gas with suspended dust particles.
The AI model described non-reciprocal forces—where one particle affects another differently than the reverse—with over 99% accuracy, correcting long-standing theoretical assumptions.
By combining machine learning with precise 3D tracking of particle trajectories from laboratory experiments, the team validated AI inferences against real-world data.
The research, published in PNAS, demonstrates that AI can move beyond data analysis to discover entirely new laws of nature in complex many-body systems.
How the AI model revealed hidden force laws in dusty plasma
The neural network was trained on 3D particle motions to infer effective forces between charged dust particles, accounting for symmetries and nonidentical particles in the system.
It learned the effective nonreciprocal forces with exquisite accuracy, enabling precise measurements of particle charge and screening length that deviated from conventional plasma physics theory.
This approach incorporated physical intuition into the machine learning structure, ensuring the model provided interpretable insights rather than operating as a black box.
Why this method could apply to other complex systems
The researchers believe their framework is universal and could be extended to other many-body systems, from colloids to living cells, where collective interactions dominate behavior.
By revealing previously hidden rules in dusty plasma, the method opens new routes to discover physical laws in systems ranging from industrial materials like paint and ink to biological assemblies.
The study’s first author, Wentao Yu, is now a postdoctoral fellow at Caltech, and co-author Eslam Abdelaleem is a postdoctoral fellow at Georgia Tech.
Funding and institutional support behind the discovery
The research was primarily supported by the National Science Foundation, with additional funding from the Simons Foundation.
Vyacheslav Lukin, program director for the NSF Plasma Physics program, noted that the project exemplifies how advances in plasma physics and AI may lead to further progress in studying living systems.
What is dusty plasma and why is it considered the fourth state of matter?
Dusty plasma is a mixture of ions, electrons, and macroscopic charged particles found in space and planetary environments, often referred to as the fourth state of matter alongside solid, liquid, and gas.
How accurate was the AI model in describing forces in dusty plasma?
The model described non-reciprocal forces with an accuracy of more than 99%, as stated by Ilya Nemenman, Emory professor of theoretical physics and co-senior author of the paper.