Emory AI finds one‑way plasma forces
- Emory physicists reported in a 2025 PNAS paper that a physics-tailored machine-learning model inferred force laws in laboratory dusty plasma from 3D particle trajectories. - The model learned nonreciprocal particle-to-particle forces with R² above 0.99 and exposed deviations from standard dusty-plasma assumptions about charge and screening length. - The work adds a rare case of AI extracting physical laws from real experiments, not just fitting data. (emory.edu)
Dusty plasma is an ionized gas with tiny charged grains mixed in, and Emory physicists used machine learning to infer how those grains push and pull on one another. (pmc.ncbi.nlm.nih.gov) The paper, “Physics-tailored machine learning reveals unexpected physics in dusty plasmas,” was published in *Proceedings of the National Academy of Sciences* on July 31, 2025. The authors are Wentao Yu, Eslam Abdelaleem, Ilya Nemenman, and Justin C. Burton of Emory University. (pmc.ncbi.nlm.nih.gov) (nature.com) In a simple textbook picture, forces between two objects are equal and opposite. In dusty plasma, the surrounding charged gas can distort that symmetry, so one particle’s effect on another does not have to be matched in the same way in reverse. (pmc.ncbi.nlm.nih.gov) (arxiv.org) The Emory team trained its model on three-dimensional particle trajectories from laboratory experiments, not on synthetic data alone. The system was built to respect known physics constraints while still learning the effective interparticle forces from motion. (pmc.ncbi.nlm.nih.gov) (arxiv.org) The model fit the measured forces with R² greater than 0.99, according to the paper. Emory said that let the researchers describe the nonreciprocal forces with accuracy above 99%. (pmc.ncbi.nlm.nih.gov) (emory.edu) The researchers also said the learned force laws disagreed with some standard assumptions used in dusty-plasma theory. In particular, the model enabled more precise estimates of particle charge and screening length than the usual simplified approximations. (pmc.ncbi.nlm.nih.gov) (arxiv.org) Dusty plasmas are not just a lab curiosity. The paper notes they appear in planetary and space environments, and Emory pointed to examples ranging from Saturn’s rings to Earth’s ionosphere. (pmc.ncbi.nlm.nih.gov) (emory.edu) The authors framed the result as a method for many-body systems, where lots of particles interact at once and standard pair-by-pair equations can miss the full picture. Emory said the same framework could be adapted to systems such as colloids and clusters of living cells. (pmc.ncbi.nlm.nih.gov) (emory.edu) Wentao Yu was the paper’s first author and had worked on the project as an Emory doctoral student before moving to the California Institute of Technology for a postdoctoral fellowship. Eslam Abdelaleem, another co-author, later moved to a postdoctoral role at the Georgia Institute of Technology, Emory said. (emory.edu) The closing claim is narrower than “AI discovered a new law of nature” and more concrete than a black-box prediction. Emory’s result is that a physics-constrained model recovered measurable force laws from messy experimental motion and found where the standard picture breaks. (pmc.ncbi.nlm.nih.gov) (emory.edu)
