Pesquisadores’s Emory University combined a custom-designed neural network with experimental laboratory data. The result exposes previously hidden patterns in the interaction of plasma particles with dust. Accuracy exceeds 99% in describing non-reciprocal forces. The work was published in Proceedings of the National Academy of Sciences.
Dust plasma consists of ionized gas that contains microscopic charged particles. Esse system occurs naturally in space, such as in the rings of Saturno, and also in terrestrial environments, such as smoke from forest fires. Scientists tracked the three-dimensional movement of dozens of particles in a controlled vacuum chamber. They then applied artificial intelligence to infer the forces that govern collective behavior.
Modelo AI learns forces between particles with high accuracy
The team separated particle motion into three main components. One of them is the speed-related drag force. Outro involves environmental forces such as gravity. The third captures direct interactions between particles. The neural network trained with real trajectories captured asymmetric details. A leading particle can attract the one behind it, while the trailing particle always repels the leader.
Essa reciprocity does not appear in many-body systems. Researchers compare the phenomenon to two boats crossing a lake and generating waves. The wake of each one affects the other differently depending on their relative position. The model achieved a coefficient of determination greater than 0.99 when predicting particle acceleration.
- The system made it possible to measure loads and armor lengths with unprecedented precision
- Independent Experimentos validated AI-inferred masses
- Resultados contradict classical theoretical assumptions about proportionality between size and load
Justin Burton, professor of experimental physics, highlighted that the method does not work as a black box. The network structure respects known physical constraints and also allows us to discover what was not known.
Suposições classical theories do not resist new data
Previous Teorias assumed that the charge of a dust particle grows proportionally to its radius. The data shows that the relationship is more complex. Ela varies depending on plasma density and temperature. The observed exponent is between 0.30 and 0.80, and increases with background gas pressure.
Outra Common assumption held that forces between particles fall exponentially with distance, regardless of size. The analysis revealed clear particle size dependence on the force decay. Additional Experimentos confirmed these deviations.
Ilya Nemenman, professor of theoretical physics, explained that the high precision made it possible to correct old inaccuracies. The model offers quantitative descriptions that did not previously exist. Wentao Yu, first author, worked on the project as a PhD student at Emory and now pursues research at Caltech. Eslam Abdelaleem, co-author, serves as a postdoctoral fellow at Georgia Tech.
Plasma with dust appears in environments from everyday life to the cosmos
Plasma is called the fourth state of matter because electrons and ions move freely. Ele makes up about 99.9% of the visible universe, from solar winds to lightning. The dust version adds charged grains that change behavior.
Na Lua, weak gravity leaves dust particles floating and sticking to astronauts’ clothes. In wildfires in Terra, charged soot particles can interfere with radio signals used by firefighters. In the laboratory, scientists suspend plastic microspheres in a vacuum chamber and adjust the pressure to simulate real conditions.
The tomographic imaging technique developed in Burton’s laboratory uses a laser sheet that scans the volume. A high-speed camera records images that, when combined, reconstruct 3D trajectories over minutes.
Abordagem paves the way for other many-body systems
The framework developed at Emory runs on a common desktop computer. Ele can be adapted to study colloids in industrial paints and inks or collective interactions in groups of living cells. Nemenman plans to apply similar ideas to the study of collective movement in biological systems during an internship at Alemanha.
Vyacheslav Lukin, program director at National Science Foundation, praised the interdisciplinary collaboration. The advance combines plasma physics and artificial intelligence and can benefit the understanding of living systems.
The research received primary support from NSF, with additional funding from Simons Foundation. The authors emphasize that success depends on careful network design and human interpretation of results.
Detalhes experiment technicians reinforce reliability
The researchers validated the model in two independent ways to infer the mass of each particle. The values coincided and matched direct measurements by optical microscopy. Essa internal consistency increases confidence in inferred forces.
The neural network incorporates physical symmetries and deals with non-identical particles. The training used a limited volume of experimental data, which required a specific architecture. Reuniões weekly sessions over more than a year refined the structure into a relatively simple but powerful model.
Impacto potential goes beyond plasma physics
Cientistas see potential in areas such as industrial materials and biology. In cancer, for example, understanding collective cell interactions can shed light on metastasis processes. The method offers a starting point for inferring laws in systems where direct interactions are difficult to model.
Burton compares the responsible use of AI to the mission of exploring the unknown. Ele believes that the tool, when used well, opens doors to entirely new realms of discovery.
The study demonstrates that artificial intelligence can go beyond analyzing or predicting. Under the right conditions, it helps reveal laws of nature that have remained hidden.