Physicists have understood for roughly a hundred and fifty years what two fluids do when they slip past one another at unequal speeds. First the boundary between them buckles, then it curls, and finally it winds up into vortexes. Named the Kelvin-Helmholtz instability and described back in the late 1860s, it explains the ripples wind draws on water and the rows of curved shapes it shears out of clouds.
On the Sun, though, nobody had ever pinned it down.
That gap has now closed. A team headed by David Kuridze and Friedrich Wöger of the National Solar Observatory reports that Kelvin-Helmholtz instabilities are not merely detectable on the solar surface — they are everywhere. The study contends this could reshape our picture of how heat, mass and magnetic energy travel through the Sun’s atmosphere.
They were hiding because they’re tiny
The explanation for a century of invisibility is decidedly unromantic: the plasma whirlpools are small. Too small for any telescope with a mirror below 2 meters to resolve — which, for nearly the whole history of solar physics, described every instrument on the planet.
The Daniel K. Inouye Solar Telescope changed the arithmetic. Opened in Hawaii by the US National Science Foundation, the 4-meter instrument is the largest solar telescope anywhere and began its operational phase in November 2021.
The vortexes were a byproduct of a test
What follows spoils any neat tale of discovery: Kuridze’s group was not looking for vortexes at all. Across a three-minute window on April 14, 2025, they aimed the telescope at an active region close to the middle of the solar disk and recorded at a wavelength of 416 nanometers, using a diagnostic camera setup that the National Solar Observatory and the Max Planck Institute for Solar System Research had built together.
“The main goal was to achieve diffraction-limited performance with the telescope,” Kuridze said.
Diffraction sets a floor on the finest detail any telescope’s optics can physically produce, and just two factors govern it: the size of the mirror and the wavelength being observed. A larger mirror or a shorter wavelength buys finer detail. With Inouye’s mirror locked at 4 meters, the team leaned on the remaining variable.
“416 nanometers is towards the smaller portion of the visible spectrum,” Kuridze said. “We selected it because we wanted to achieve a higher diffraction limit and higher resolution.”
740 frames a second, 19 kilometers of detail
Running at exposures of 100 microseconds, the camera read out 740 frames per second. Each finished image drew on two thousand of those frames via multi-frame blind deconvolution, a method that numerically models the smearing left by Earth’s atmosphere once the telescope’s optics have done their part, then subtracts it.
The result was a movie of the solar surface delivering a fresh frame every two seconds at a spatial resolution near 19 kilometers — precisely the theoretical ceiling for a 4-meter mirror at that wavelength. The team hit it dead on.
“As a byproduct we got these amazing observations, which allowed us to see something which has never been seen before,” Kuridze said.
What a blurry smudge turned into
Granules — the convection cells ferrying heat up from the interior — dominate the solar surface at 416 nanometers, threaded through with concentrated bundles of intense magnetic fields. Where a magnetic bundle meets the surrounding granulation, lower-resolution imagery renders the boundary as smooth and faintly blurred.
The DKIST data shows something else entirely. Those interfaces turn out to consist almost wholly of vortex-like structures and fine dark striations.
Within the field of view the team catalogued 47 vortex-bearing interfaces and gauged how far apart neighboring curls sat: typically 60 to 100 kilometers, with individual vortexes measuring 25 to 170 kilometers across. Because the tiniest examples fell right at the 19-kilometer resolution limit, there is no way yet to say how much smaller the population goes.
Their growth is rapid as well. Tracking the structures, the researchers saw vortexes double in size in less than a minute while travelling along the interfaces at 0.67 to 3 kilometers per second.
Two phases, and one of them is a mess
Asked what an observer hovering above one of these boundaries would witness, Kuridze answers that it depends on the timing, because the instability unfolds in two phases.
“The first is the linear phase, when things are more relaxed, very well organized, regular and beautiful,” he said. “If you see them from close range you will see something like cloud-type things, which are rolling.”
The quiet does not last. “At some point everything turns into a non-linear regime and then things get messy,” Kuridze said. “You are basically getting turbulence, very chaotic turbulence.”
Why the curls form there and nowhere else
Threaded through the plasma, magnetic field lines act like elastic strands that push back when bent. Lying parallel to the flow, they snap a rippling boundary flat again and suppress the instability before it can develop. Oriented across the flow, they have no effect at all.
In the strongly magnetized regions DKIST targeted, the field emerges nearly vertically from the surface while granular flows slip past it horizontally. The strands are simply strung the wrong way to hold anything together, so nothing checks the curls as they grow.
The simulations that made the case
Observing something nobody has observed before is also an excellent way to be fooled by an image-processing artifact, and the team was well aware of it.
“Everything looked like it should be Kelvin-Helmholtz, but of course this is not enough,” Kuridze said. “You need theoretical proof to make sure that it is really Kelvin-Helmholtz.”
Their answer was a simulation: a patch of photosphere about 6 megameters across at a grid spacing of 3.2 kilometers, seeded with a magnetic field map of the very region they had observed. From that they synthesized the images DKIST ought to have registered while looking at the simulated patch, computing 500 spectral points across the observed wavelength band, applying the real interference filter’s transmission profile and degrading the output to match the actual telescope’s resolution.
Appearance and dynamics alike came through in the synthetic images, growth rate, distribution and propagation speeds included. On that basis the team concluded the observations are most likely genuine.
A stirring mechanism nobody budgeted for
Plasma is supposed to sit still under a strong magnetic field. Sunspots make the point plainly: there the field shuts down convection altogether, leaving the surface cool and dark.
An instability that spins up vortexes along the rim of every magnetic element therefore introduces a stirring mechanism where nobody had allowed for one. Magnetized and unmagnetized gas can mix together. Cool material from the fringes of convection cells can seep into the magnetic regions and alter how heat travels just beneath the visible surface. None of that appears in existing models of solar convection.
The corona raises further questions — the Sun’s million-degree outer atmosphere gets part of its heat from field lines that are jostled at their anchor points until they braid into tangles that eventually snap and release energy. Nobody has ever seen the mechanism responsible for that jostling.
“You have these twisting motions everywhere at the surface of the magnetic element, and this twisting motion is nothing else than braiding of the magnetic fields,” Kuridze said.
Three minutes is not a survey
Every bit of this stands on a three-minute observing window plus simulations that carried limits of their own — worth remembering before anyone starts rewriting textbooks.
“One thing we just don’t know is how small these Kelvin-Helmholtz patterns get on the Sun,” Kuridze said. DKIST’s resolving power butts right up against the limit, leaving open the possibility that smaller structures are slipping past unseen. Running simulations finer than the telescope can see is the obvious workaround, and the team attempted it. The attempt has not gone smoothly.
“When you do this at higher resolution in the simulations, some extra physics needs to be involved, and we are not sure exactly how things work in computational simulations when you need to reach those resolutions,” he said. “This is a completely new area, and we need to do more investigation.”
A second gap remains. What DKIST largely supplied were high-resolution visual images, a reasonable stand-in for where the magnetic fields sit and how strong they are — but a stand-in nonetheless. The team’s knowledge of how fast the plasma genuinely moves rests on simulations rather than the telescope, and the magnetic field powering the entire phenomenon has never been measured directly on this scale.
Kuridze’s remedy is more time on the telescope rather than more ingenuity. “If you want to quantify how the magnetic field is evolving in time, what sort of dissipation you have, how much energy is released, how much energy budget there is for eruptions and flares, you need much longer observations, and you need magnetic maps,” he said. “This is the next challenge and the next milestone.”
Nature, 2026. DOI: 10.1038/s41586-026-10871-3