AI revealed unexpected new physics in dust plasma

AI revealed unexpected new physics in dust plasma

AI revealed unexpected new physics in dust plasma

A scene inside the laboratory vacuum chamber, where the colloidal particles are suspended in the flat disk, which is bright with laser green light, so as to study the dust plasma. Credit: Burton Lab

Physicians have used the method of learning the machine to identify a surprising new turn on non -reclaimed forces that rule the physical system.

Joint The action of the National Academy of Sciences At the University of Emuri published the results by experimental and theoretical physicists, based on data from a nervous network model and laboratory experiments on dust plasma – which contains suspended dust particles.

This work is one of the relatively few examples not to use AI as a data processing or prediction tool, but also to discover new physical laws that rule the natural world.

“We showed that we can use AI to discover new physics,” says Justin Burton, an EMORY Professor of Experimental Physics and senior co -author of the article. “Our AI method is not a black box: we understand how and why it works. The framework that provides is also universal. It can potentially be applied to many other physical systems to open new paths of discovery.”

pnas The paper still provides a very detailed explanation for the physics of the dusty plasma, which gives precise guesses for non -reclaimed forces.

“We can describe these forces with more than 99 % accuracy,” says Elia Nemanman, an emourist professor of ideological physics and senior author of the dissertation.

“What’s more interesting is that we show that some common ideological ideas about these forces are not right at all. We are able to correct these mistakes because now we can see what is happening in such a good detail.”







The Burton Lab prepared a technique to track the 3D movement of individual particles in the laboratory dust plasma. Researchers were allowed to examine AI from running experiments. Credit: Burton Lab

Researchers hope that their AI approach will serve as a starting point to estimate the rules from the dynamics of a wide range of many physical systems, which contain a large number of communicating particles. Examples of this include from colids to paint, ink and other industrial materials – from the cells of the cells in animals.

The first author of this paper is Vento Yu, who worked on the project as an Emory PhD. The student and is now a post -documentary fellow at the California Institute of Technology. Co -author is Aslam Abdulm, who was also part of the project as an Emuri graduate student and is now a post -documentary fellow in Georgia Tech.

“The project works as an excellent example of an inter -formal cooperation where plasma physics and AI can develop more in the study of new knowledge development systems,” says Viachuslav (Slava) Lukin, the program director of the NSF Plasma Physics Program. “The dynamics of these complex systems dominate the collective interactions that emerging AI techniques can help us better describe, identify, understand, and even overcome.”

Plasma are ionized gases, which means that the charged particles of electrons and ions are transmitted freely, which produces unique properties such as electrical conductivity. Known as the fourth state of the material, the plasma makes it 99.9 % of the visible solar winds from the corona -driven solar winds to the Earth’s lightning bolt, an estimated 99.9 % of the visible universe.

Dust plasma, which adds dust charging particles to the mixture of ions and electrons, is also common in the space and planets.

Lating above the moon level due to weak gravity is an example of charged particles, dust plasma. “That is why when astronauts walk on the moon, their suites are drowning in the dust,” Burton explained.







See the dust spray from Nile Armstrong’s shoes when he pancreases to collect samples of cliffs from the moon surface. Credit: NASA APOLOO 11 Library

An example of dusty plasma on the ground can occur during wildfire when the mascara is mixed with smoke. The charged masculine particles can interfere with radio signals, which affect the communication between firefighters.

Burton studies the physics of dust plasma and amorphosis material. His lab experimented with small, plastic particles suspended in a vacuum chamber filled with plasma as a model of more complex systems. By changing the gas pressure within the chamber, the members of the lab can imitate the real phenomena and study how the system changes when the forces operate.

For the current project, Burton and Yu developed a tomographic imaging technique to track the three -dimensional (3D) movement of particles in the dust plasma. A laser light spreads in the vacuum chamber as the high -speed camera captures images. Snap shots of particles inside the light aircraft are then deposited in a stack, which shows a 3D space of individual particles at a scales for several minutes.

A theoretical bio -physicist looks for laws that are subjected to natural dynamic systems, especially complex biological systems. He is particularly interested in the trend of collective movement, such as how human cells move about the body.

“The general questions of how the whole system creates the whole system is very important,” said Nemanman. “For example, in cancer, you want to understand that the conversation of cells becomes a metastatic, breaking down from some of them and moving to a new place.”

Although Niman often cooperates with researchers of Life Sciences, the project with the Burton Lab provides the opportunity to get into a much easier system than a living. It offered an ideal opportunity to try to use AI to investigate the dynamics of the mass movement to learn new physics.

“There are very few examples about how AI is revolutionizing in science,” says Nemanman.

AI, one of the most famous examples of Chatgpt, train on a large amount of information available on the Internet to predict appropriate text in response to an indicator.

“When you are investigating something new, you don’t have too much data for AI training.” “This meant that we had to design a nervous network that could be trained in a small amount of data and still learn something new.”

Burton, Neman, Yu and Abdulm met in a weekly conference room to discuss the issue.

“We needed to create a network to comply with the necessary rules, while it is still allowed to detect and estimate unknown physics,” Burton explained.

“At these weekly meetings, we had more than a year after talks,” said Nemanman. “Once we brought the right network structure for training, it turned out to be easier.”

Physicians removed restrictions to create three -independent contribution models in particle movement by the nerve network: the effect of speed, or the drag force; Environmental forces, such as gravity; And particle forces.

A trained, AI model, trained at 3D Particular Trackers, calculated hereditary balance, non -inferior particles and learned the effective non -reclaimed forces among particles with the utmost accuracy.

To explain these non -reciprocal forces, the researchers use the imitation of the two boats moving across the lake, causing waves. The weekend of each boat affects the second boat’s movement. A boat dispute depends on their positions – the other can withdraw or attract the boat – for example, whether boats are traveling side by side or behind each other.

“In a dust plasma, we explain how a well -known particle attracts the previous particle, but the particle of the back always repels the well -known,” Niman said. “This trend was expected by some people, but now we have an exact estimate for it that has not existed before.”

AI revealed unexpected new physics in dust plasma

Experience and data workflow review. Credit: The action of the National Academy of Sciences (2025) DOI: 10.1073/pnas.2505725122, https://www.pnas.org/doi/10.1073/pnas.2505725122

Their searches also correct some wrong assumptions about dust plasma.

For example, a longtime theory said that in the exact proportion of particle radius, the radius of the dust particle, the so much charge that was trapped on this particle. “We have shown that this theory is not right,” says Nemanman. “Although it is true that the bigger the particle is, the increase is not necessarily proportional to the radius. It depends on the density and temperature of the plasma. “

Another theory said that the forces between the two particles fall rapidly in the distance between the two particles directly, and the element through which it falls does not depend on the size of the particle. The new AI method shows that the drop -off -offs depend on particle size.

Researchers confirmed their results through experiments.

Their physics -based nervous network runs on a desktop computer and offers a universal, theoretical framework to open the mystery about other complex, many body systems.

For example, Nymans are waiting for the coming professor at the Constitutions School of Commonwealth in Germany. The school collects an intercontinental approach to studying the growing field of collective behavior, everything from birds to fish and human mobs.

He says, “I will teach students all over the world how to use AI to assess the physics of the collective movement.

Although their AI framework has the ability to evaluate new physics, human physicists need to design the right structure for the nerve network and translate and verify the resulting data.

“Critical thinking is needed in the manufacture and use of AI tools through real -growing ways in science, technology and humanity,” says Burton.

He feels hopeful about the possibility of AI to benefit society.

“I think about it as a star track slogan, go with pride where no one has before,” says Burton. “Used properly, AI can open the door to the entire new circle to discover.”

More information:
Physics -filled machine learns indicate unexpected physics in dusty plasma, The action of the National Academy of Sciences (2025) DOI: 10.1073/pnas.2505725122, www.pnas.org/doi/10.1073/pnas.2505725122

Provided by Emuri University

Reference: AI has unexpectedly revealed new physics in Dust Plasma (2025, August 1).

This document is subject to copyright. In addition to any fair issues for the purpose of private study or research, no part can be re -reproduced without written permission. The content is provided only for information purposes.

Share this article

Leave a Reply

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