AI finds non‑reciprocal forces in plasma
- Emory University physicists reported that a physics-tailored neural network learned force laws inside dusty plasma from 3D particle tracks, then experimentally confirmed uneven particle-to-particle interactions in the lab. - In a Proceedings of the National Academy of Sciences paper published July 31, 2025, the model fit the data with R² above 0.99 and found charge and screening values that diverged from standard assumptions. - The result adds precision to models of dusty plasmas found in space and industry, and gives researchers a template for other many-body systems. (pnas.org)
Plasma is a gas so energized that electrons break free, leaving a soup of charged particles. In dusty plasma, that soup also carries tiny grains of dust that push and pull on one another. (pnas.org) Those interactions are hard to measure because the plasma itself acts like an invisible middleman. A particle can change the local flow of ions, and that altered flow can push a neighboring particle differently in return. (pnas.org) (nature.com) A team at Emory University used a neural network built around known physical constraints to infer those forces directly from laboratory data. The paper, led by Wentao Yu with Eslam Abdelaleem, Ilya Nemenman and Justin C. Burton, appeared in Proceedings of the National Academy of Sciences on July 31, 2025. (pnas.org) (pubmed.ncbi.nlm.nih.gov) The model was trained on 3D trajectories of particles moving inside a laboratory dusty plasma. It learned effective nonreciprocal forces with R² greater than 0.99, meaning its predictions closely matched the measured motion. (pnas.org) Nonreciprocal means the force from particle A on particle B is not simply mirrored by the force from B on A. In ordinary textbook mechanics, action and reaction balance inside an isolated system, but here the surrounding plasma carries part of the momentum and energy. (nature.com) (pnas.org) The Emory group did not discover that asymmetry exists in principle. Earlier experiments and theory had already shown nonreciprocal interactions in complex plasmas, including a 2020 Scientific Reports study that called direct force measurements an unsolved problem. (nature.com) What the new paper adds is a much sharper map of those forces in a many-particle experiment. The authors said the model was accurate enough to infer particle masses in two independent ways and to extract charge and screening length from the motion itself. (pnas.org) (news.emory.edu) That precision also exposed gaps in common approximations used for dusty plasmas. The paper reports large deviations from standard theoretical assumptions about how particle charge and screening should scale. (pnas.org) Dusty plasmas matter outside the lab because they appear in places such as Saturn’s rings and interstellar space, and they also show up in technological plasma processes. The paper says the same machine-learning framework could be adapted to other many-body systems, from colloids to living tissues. (pnas.org) (news.emory.edu) Burton said the method is not a black box and that the team understands how and why it works. For this result, the headline is narrower than “new physics” in general: AI helped turn particle motion into a tested force law for dusty plasma. (news.emory.edu)
