Nuclear fusion and astrophysics are related fields. That these fields can learn from each other is demonstrated in a recent Nature publication co-authored by DIFFER researcher MJ Pueschel. Corresponding author Bindesh Tripathi links the mechanism of magnetic-field generation to a three-decades-old study of dynamo action in fusion plasmas, albeit giving a new interpretation. Another connection to fusion plasmas is provided by the notion that turbulence is created not by a single impulse but by continuous forcing.
"In this collaboration, we managed to combine our knowledge on fluid turbulence with that on nuclear fusion and astrophysics. With the help of record-breaking computer simulations, this allowed us to understand where magnetic fields in the Universe are coming from – a long-standing scientific mystery and an ingredient in enabling the Universe to produce intelligent life,” says Pueschel.
Read the publication on Nature: https://www.nature.com/articles/s41586-025-09912-0
The following text was written by the press office of Wisconsin-Madison:
All celestial bodies — planets, suns, even entire galaxies — produce magnetic fields, affecting such cosmic processes as the solar wind, high-energy particle transport, and galaxy formation. Small-scale magnetic fields are generally turbulent and chaotic, yet large-scale fields are organized, a phenomenon that plasma astrophysicists have tried explaining for decades, unsuccessfully.
In a paper published 12 January 2026 in Nature, a team led by scientists at the University of Wisconsin–Madison have run complex numerical simulations of plasma flows that, while leading to turbulence, also develop structured flows due to the formation of large-scale jets. From their simulations, the team has identified a new mechanism to describe the generation of magnetic fields that can be broadly applied, and has implications ranging from space weather to multimessenger astrophysics.
“Magnetic fields across the cosmos are large-scale and ordered, but our understanding of how these fields are generated is that they come from some kind of turbulent motion,” says the study’s lead author Bindesh Tripathi, a former UW–Madison physics graduate student and current postdoctoral researcher at Columbia University. “Given that turbulence is known to be a destructive agent, the question remains, how does it create a constructive, large-scale field?”
Before working on three-dimensional (3D) magnetic fields, Tripathi investigated systems with hydrodynamic flows and two-dimensional (2D) magnetic fields. After staring at the movies and images of 3Dl magnetic turbulence, he noticed similarities in the shapes of large-scale flows and large-scale magnetic field structures. But it wasn’t as simple as applying fluid dynamic theory to magnetic field generation: the former may be solved as a 2D problem, whereas the latter must be solved in 3D, making it a much more complex, difficult-to-solve problem.
Tripathi and his colleagues decided to tackle the problem with two key changes from previous research.
The first difference was the input: a constantly replenished velocity gradient. A cyclist hitting a curb head-on, say, experiences a velocity gradient: the wheels stop, but momentum can cause the cyclist to fly over the handlebars. Velocity gradients exist throughout the universe; for example, within different layers of the sun or when two neutron stars merge. The team reasoned that this gradient is likely important to include while studying 3D magnetic fields.
Second, they ran perhaps the most complex simulation to date of magnetic fields in the presence of an unstable velocity gradient — 137 billion grid points in 3D space. Altogether, they ran around 90 simulations, generating 0.25 petabytes of data and using nearly 100 million CPU hours on the Anvil supercomputer at Purdue University.
“We start our simulations with a flow that has a velocity gradient, then we add some tiny perturbations, like moving one fluid particle infinitesimally, we let that perturbation propagate over the system and grow, and then analyse the data over time,” Tripathi says. “Initially, these perturbations lead to turbulent flows and magnetic fields in small-scale structures, then, over time, they emerge into larger, ordered structures.”
When Tripathi ran the same simulations where the initial velocity gradient had decayed over time, the simulation only produced the chaotic, small-scale patterns. “So that’s really the main key: to have a steady, large-scale gradient in velocity,” he emphasizes.
Adds Paul Terry, physics professor at UW–Madison and senior author of the study: “Magnetic field generation via dynamos has been extensively studied for 70 years, with the frustrating result that the generated fields almost always end up at small scales and highly disordered, unlike observations. This work, therefore, potentially resolves a long-standing issue.”
Though the theory cannot be tested in the distant universe, a lab-based experiment does support the team's findings: in 2012, colleagues at the Wisconsin Plasma Physics Laboratory were trying to better understand the nature of the magnetic field generation process in a laboratory experiment, but their data did not fit any of the previous models. Tripathi and colleagues’ new theory of magnetic field generation more closely matches the experimental data and helps to resolve the confounding findings.
“This work has the potential to explain the magnetic dynamics relevant in, for example, neutron star mergers and black hole formation, with direct applications to multimessenger astronomy,” Tripathi says. “It may also help better understand stellar magnetic fields and predict gas ejections from the sun toward the earth.”
This work was supported by the National Science Foundation (2409206) and U.S. Department of Energy (DE-SC0022257) through the DOE/NSF Partnership in Basic Plasma Science and Engineering. Anvil at Purdue University was used through allocation TG-PHY130027 from the Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) program, which is supported by National Science Foundation (2138259, 2138286, 2138307, 2137603 and 2138296).
Go to the News page.