A team of scientists at Los Alamos National Laboratory has developed a novel machine learning framework capable of modeling the chaotic movements of particles in turbulent flows. This advance is significant because predicting the behavior of particles entrained by turbulence—whether dust in a tornado or sugar grains in a cup of coffee—has historically been a formidable challenge, especially at large scales. The research, published in *Proceedings of the National Academy of Sciences*, represents a major step towards a deeper understanding of this ubiquitous phenomenon in physics and engineering.

Turbulence is one of the most complex unsolved problems in classical physics, characterized by its unpredictability and the wide range of spatial and temporal scales involved. Traditional models often struggle to capture the detailed dynamics of particles within these chaotic flows, limiting our ability to predict and control processes ranging from atmospheric pollutant dispersion to mixing in chemical reactors. This new machine learning-based approach offers a promising avenue to overcome these limitations by learning directly from data rather than relying exclusively on approximate physical equations.

The developed framework is the first of its kind to use machine learning to model particle movement in turbulence at scale. Although the summary does not detail the specific methodology or quantitative results, the implication is that this model can identify patterns and relationships within turbulence data that are difficult to discern with conventional methods. The ability to more accurately predict particle behavior in turbulent environments has broad implications, from improving climate models and weather forecasting to the more efficient design of vehicles and industrial processes. This work opens the door to future research that could further refine these models and apply them to a variety of complex scenarios in science and engineering.