Which tells you more about the Sun: a granule 2,000 kilometers wide, or the 20-kilometer curl breaking along its edge?
This is FreeAstroScience. What follows is a small result measured at the very edge of what any telescope can resolve: the instability behind the Sun’s fringed granule edges, the 65-kilometer scale it works at, and the one claim the Nature paper still refuses to make.
On the Sun’s visible surface, the 20-kilometer curl carries the new information rather than the 2,000-kilometer granule beneath it. Those curls are magnetized Kelvin-Helmholtz instabilities, shear-driven whirls that grow where neighboring plasma layers slide past each other at different speeds, and the Daniel K. Inouye Solar Telescope resolved them at a spatial resolution of 19 kilometers on 14 April 2025. They mix magnetized and unmagnetized plasma wherever granules press against a strong magnetic field.
For as long as anyone has photographed the Sun in close-up, the edges of its granules looked soft. Not sharp, not fringed, just a blur where one bubble of rising plasma met the next. Kuridze and colleagues give a plain reason for that softness: the characteristic scale of the structures involved sits below what telescopes with apertures under 2 meters can resolve. Blur was the instrument, not the star.
What did DKIST actually see on 14 April 2025?
Between 21:38 and 21:41 universal time, the Daniel K. Inouye Solar Telescope was pointed at active region NOAA 14060, close to disk center at a viewing angle of roughly 14 degrees. A diagnostic camera setup called FastCam, built by the National Solar Observatory together with the Max Planck Institute for Solar System Research, read a 2K by 1K sensor at 740 frames per second on a 100-microsecond exposure.
Each usable science frame was then rebuilt from 2,000 calibrated camera frames by multi-frame blind deconvolution, a technique that models the leftover atmospheric distortion and removes it numerically. What survived that process covers about 5,800 by 4,350 kilometers of solar surface at a resolution of 19 kilometers, which is the Rayleigh diffraction limit of a 4-meter aperture at the observing wavelength of 416 nanometers. The reconstructed sequence, along with its documentation, is available from the DKIST FastCam data archive.
Michiel van Noort of the Max Planck Institute, who gathered the data and ran the reduction and image restoration, put the difficulty flatly. Resolving structures around 20 kilometers in size, he said, is “at the limit of what even the world’s largest solar telescope and state-of-the-art simulations can achieve.” His institute offers a comparison for what 19 kilometers means from Earth, and it is a good one: picking out a 1-euro coin from 180 kilometers away.
Getting a clean look at the Sun has never been comfortable for an instrument, which is part of why our earlier piece on Parker Solar Probe measuring the Sun’s electric field was worth writing. DKIST works from the ground instead, and pays for that in atmosphere. During the best periods of this run the Fried parameter sat around 12 centimeters.
Why a shear layer curls
Two fluids sliding past each other at different speeds do not stay politely separated. Shear at the boundary amplifies tiny disturbances into waves, and the waves roll over into vortices. Lord Kelvin and Hermann von Helmholtz worked the mechanism out in papers from 1868 and 1871, and it has been turning up in new places ever since.
Max Planck’s release lists some of those places:
- The surface of a lake.
- Ocean waves as they break.
- Cloud formation in Earth’s atmosphere.
- The atmospheres of Jupiter and Saturn, where neighboring bands slide past one another at different speeds.
- The boundary where the solar wind meets a planetary magnetosphere.
A magnetic field can shut the instability down, but only from one direction. The component running parallel to the shear flow resists the bending; the component perpendicular to it provides no stabilization at all. In the plage region DKIST looked at, the strong fields stand nearly vertical, which puts them perpendicular to the horizontal shear layers that converging granular flows build around them. So the field dominating this scene is precisely the field least able to stop the curling, and that is why the instability is everywhere in these images rather than occasional.
Linear theory ties the fastest-growing wavelength to the thickness of the velocity shear layer. Measured in the simulations, that layer runs about 12 kilometers thick, and the theory then predicts most unstable wavelengths somewhere between 60 and 100 kilometers.
How fast do these vortices grow?
During its linear phase a Kelvin-Helmholtz ripple grows exponentially, so its growth rate can be measured directly by tracking how far the corrugated boundary is displaced sideways from frame to frame. Doing that on the DKIST sequence gives rates between 0.014 and 0.054 per second.
Rates in units of per second are hard to feel. Converting them into a timescale is one division, because the e-folding time is the reciprocal of the growth rate:
t = 1 / γ
Here t is the time a ripple takes to grow its amplitude by a factor of e, roughly 2.7, and γ is the measured growth rate in units of per second. Feed in the fastest observed value, 0.054 per second, and t works out at about 18 seconds. At the slow end, 0.014 per second gives around 71 seconds. Which means the whole three-minute observing run spans only about ten e-folding times at the fast end, and that is why a sequence this short was long enough to watch these structures being born, roll over and merge.
Synthetic images from the simulation land in the same territory, 0.027 to 0.059 per second. Apparent vortex speeds line up too: 0.67 to 3.0 kilometers per second measured on the Sun, against 1.6 to 2.8 kilometers per second in the model. Both sets overlap. Growth rates are the unsettling part of this paper, because something that reorganizes itself in under twenty seconds is not a background detail in a star’s magnetic life.
The simulation and the telescope do not quite agree
Forty-seven Kelvin-Helmholtz interfaces were identified by eye across the DKIST field and measured for wavelength, defined as the median separation between the intensity dips running along each interface. That distribution fits a log-normal with a peak near 65 kilometers, with individual values spread from 25 to 170 kilometers. Sixty-five kilometers is about three percent of the width of a 2,000-kilometer granule, and that ratio is the whole reason these structures stayed invisible for so long.
Run the same procedure on 94 instability occurrences in the synthetic filtergrams and the histogram peaks at 49 kilometers instead.
Both distributions peak well above the 19-kilometer diffraction limit, which is the reassuring half of the result, since a histogram piling up against the resolution floor usually means the instrument is writing the answer. But observation and model disagree by a third on where the peak sits, and the authors decline to draw the tidy conclusion. They write that they refrain from concluding that smaller characteristic scales do not exist on the Sun, given the remaining differences between the two distributions. That is an honest sentence, and it is the one most coverage of this paper will skip.
Knowing what the model actually is matters here. MURaM ran a plage region inside a box measuring 6.144 by 6.144 by 2.048 megameters on a grid spacing of 3.2 kilometers, and the emergent 416-nanometer intensity was synthesized with the Rybicki-Hummer radiative transfer code before being blurred down to match DKIST. Anyone who wants to check the physics rather than take it on trust can read the MURaM radiative MHD code, which is public.
What the curls do to the Sun’s magnetic field
Vortices this size do more than look interesting. They mix. Using the vortices as the source of a turbulent diffusivity, η ≈ (1/3)uλ, with a typical length scale λ of 65 kilometers and an average shear velocity u of 3 kilometers per second, the paper arrives at roughly 0.65 × 1012 square centimeters per second.
Where that number applies is what makes it interesting. It applies inside magnetic field regions stronger than equipartition, the places where turbulence is normally suppressed and mixing is not supposed to happen. Low-entropy material from the edge of a granule gets dragged into the magnetized plasma, and magnetized plasma gets stirred out into its surroundings.
Depth changes the picture again. In the simulations these vortices develop across a vertical range running from about 100 to 400 kilometers below the visible surface, and they grow stronger the deeper you look, because magnetic stabilization weakens where the field is less vertical and the plasma beta is higher. Down there the instability starts doing something structural rather than cosmetic: it splits what looks like one monolithic magnetic element at the surface into several separate strands underneath.
The same mechanism finally accounts for the fine dark striations that high-resolution images have shown at the edges of magnetic elements for decades without anyone establishing where they came from. A corrugated boundary shifts the height at which the emergent intensity forms, and the striations follow from that shift. Twisted field lines, meanwhile, are the standard reservoir for the small energy releases that keep the corona hot, which is where the story stops being about the photosphere alone. When we covered the X-class flare that was the strongest since 2017, the energy behind it was described as stored in twisted magnetic field. Nobody could then say what did the twisting. These vortices are a candidate.
Three minutes of one active region
Limits here are real, and the paper does not hide them. What DKIST delivered is a three-minute time series of a single active region at a single wavelength, recorded on one day through variable seeing. Forty-seven vortices were picked out by manual visual inspection rather than by an automated detector, so the sample carries whatever a careful human eye carries.
More important than any of that: the observations never measured the horizontal plasma velocity. That quantity comes entirely from the simulation, where it runs parallel to the sharp magnetic interface and carries a strong gradient there, which is what creates the optimal conditions for the instability. So the ingredient that makes the mechanism work is modeled, not seen.
Two smaller caveats deserve carrying with you. Because of the Wilson depression, the surface where optical depth reaches unity sits at different geometrical heights inside and outside magnetic concentrations, so “the solar surface” here is a corrugated reference rather than a plane. And MURaM used a numerical diffusivity that the authors justify at this grid spacing while stating outright that a spatially dependent Spitzer diffusivity would be required for higher-resolution runs.
Does this explain nanoflares?
Press release and paper part company slightly at this point, and the gap is worth naming. The Max Planck text says the vortices “could be the driving force that routinely triggers the twisting” of the field lines whose snapping produces nanoflares. Wording in the paper is narrower: the observed vortices “could provide an efficient source of small-scale, omnipresent magnetic flux braiding.” Braiding is a precondition for the energy release. It is not the release.
A second difference is framing. Read the release and you get the discovery of something never before visible. Open the paper and you get experimental confirmation of a prediction first made in 1993 by Karpen and colleagues and developed in 2004 by Kolesnikov and colleagues. Both descriptions are accurate. Only one of them tells you that theorists called this thirty-three years ago and then waited for an aperture large enough to check.
One thread here gets left out on purpose. The paper notes that similar striated structures appear in molecular clouds, in the Polaris flare and in Taurus, and suggests the same instability may help create them. That is a different subject with its own literature, and folding it in would mean summarizing work we have not read. It gets its own article or it gets nothing.
Sami Solanki, who directs the Max Planck Institute for Solar System Research and co-authored the paper, framed the whole thing as a statement about scale: minute processes at the limit of what can be resolved “significantly determine the nature of our star.” Readers of our reflection on whether the Sun is a king or just another star will recognize the move. Whether it holds depends on whether these vortices really are as ubiquitous as three minutes of one plage region suggest.
Where the smooth edge went
Nineteen kilometers of resolution, reached on 14 April 2025 and published on 5 August 2026, turned a soft granule edge into a row of breaking waves. Wavelengths across 47 measured interfaces peak at 65 kilometers, the ripples e-fold in about 18 seconds at the fast end, and they stir magnetized and unmagnetized plasma together in exactly the regions where turbulence was supposed to be suppressed. What the paper will not say is that nothing smaller exists down there.
We wrote it out at this length because most coverage will hand you the euro-coin comparison and stop, while the histogram is the part that lets you judge the claim yourself. Turning hard physics into words you already own is the whole job here. Do not let the part of you that asks questions doze off, because that is the hour monsters keep. Argue with our reading of the press release if you think we were unfair to it. When someone points the same camera at a quiet-Sun field instead of a plage, we will run these numbers again on this page. FreeAstroScience, Rimini. Gerd Dani.
Sources
- Kuridze, D., Woger, F., van Noort, M., Rempel, M., Cameron, R., Rimmele, T., Solanki, S. K., Jaeggli, S. A., Tritschler, A., Uitenbroek, H., Przybylski, D. & Boboltz, D. A. (2026). Ubiquitous Kelvin-Helmholtz instabilities driving plasma mixing on the Sun. Nature, Springer Nature. Received 13 March 2026, accepted 30 June 2026, published online 5 August 2026. Open access. DOI: 10.1038/s41586-026-10871-3. https://www.nature.com/articles/s41586-026-10871-3
- DKIST FastCam dataset (van Noort setup), National Solar Observatory and Max Planck Institute for Solar System Research. Observations of 14 April 2025. https://dkist.virtualsolar.org/vanNoortfastcam/
- MURaM radiative magnetohydrodynamic code, National Center for Atmospheric Research. Source repository used for the simulations in Kuridze et al. (2026). https://github.com/NCAR/MURaM_main
- Max Planck Society (2026). Tiny vortices discovered on the sun’s surface. Published 5 August 2026 via Phys.org, edited by Lisa Lock, reviewed by Robert Egan. https://phys.org/news/2026-08-tiny-vortices-sun-surface.html
