Dark matter is supposed to be everywhere. It makes up 85 percent of all matter in the universe. It doesn’t shine, doesn’t interact with light, and yet its gravity is the scaffolding that holds every galaxy together. You can’t see it — but you’re expected to feel its effects in the motion of every star.
So when a galaxy turns up without it, the obvious first instinct is: something must be wrong with the measurement.
That was the response in 2018, when a Yale team led by Pieter van Dokkum first described NGC 1052-DF2 — an ultra-diffuse galaxy the size of the Milky Way but built with roughly 500 times fewer stars, whose stars moved far too slowly to require any dark matter at all. Then came a second: NGC 1052-DF4. Now, in a preprint submitted to arXiv on March 16, 2026, the same team has confirmed a third: NGC 1052-DF9.
All three are lined up.
What the data says
Using the Keck Cosmic Web Imager (KCWI) at the W. M. Keck Observatory in Hawai‘i, astronomers measured how fast the stars inside DF9 are moving relative to one another — a quantity called stellar velocity dispersion. It is a direct proxy for how much gravity is at work inside a galaxy. Dark matter adds mass; mass adds gravity; gravity stirs stars into faster, more agitated orbits. A galaxy without dark matter moves more quietly.
DF9 measured at 6.4 km/s. That is consistent with the 8.3 km/s you would expect from DF9’s stellar mass of about 140 million solar masses acting alone. The dispersion expected if DF9 contained a normal dark matter halo would be 27 ± 3 km/s — more than four times what was observed. The conclusion reached by lead authors Michael Keim, van Dokkum, and their colleagues Zili Shen, Shany Danieli, and Imad Pasha: dark matter is not required to explain the kinematics of DF9.
Three galaxies. The same corner of the sky. No dark matter in any of them.
A trail no one expected
The NGC 1052 group lies roughly 67 million light-years from Earth, anchored by a large elliptical galaxy of the same name. DF2 and DF4 were already anomalous — but over years of follow-up work, they turned out to belong to something stranger still: a narrow, kinematically connected trail of roughly a dozen faint galaxies stretching across the field. Each follows a precise velocity gradient that links them into a single coherent structure. DF9 falls right on that trail, its measured velocity matching the predicted trend almost exactly.
Galaxies do not organize themselves into linear trails by accident.
The leading explanation is the Bullet Dwarf scenario. Around 8 billion years ago, two gas-rich dwarf galaxies collided at enormous speed. Dark matter halos, which interact so weakly with ordinary matter that they pass through most things almost undisturbed, slipped through each other like ghosts. The gas clouds were not so fortunate — they collided, decelerated, and eventually collapsed into new stars. What formed in the collision’s wake was a chain of galaxies built entirely from normal matter, strung along the axis of impact. The dark matter continued outward. The gas, and the stars it made, stayed behind.
DF2, DF4, and now DF9 were born in that wake. Their stars, their quiet motions, and their position along the trail all place them within the predicted aftermath of that single ancient event.
There is a notable irony here. The Bullet Dwarf scenario only works if dark matter is a real, separable substance — exactly as the standard Lambda-CDM cosmological model describes. These dark-matter-free galaxies are not a challenge to dark matter’s existence; they may be among the most direct evidence that it is real. The competing framework, Modified Newtonian Dynamics (MOND), predicts higher velocity dispersions than DF9 shows and offers no natural explanation for a linear trail of dark-matter-free galaxies. Each new confirmation tightens the case against it.
What comes next
DF9 is the third confirmed dark-matter-free galaxy on the trail, but the trail itself contains roughly a dozen faint members, most of them not yet measured. If the Bullet Dwarf scenario is correct, some or all of those remaining galaxies should show similar signatures — stars moving at dispersions consistent with their stellar mass alone, no dark matter required.
Measuring them is the next task. Future observations with the James Webb Space Telescope or the forthcoming Vera C. Rubin Observatory could also probe the stellar populations and structural properties of these galaxies in finer detail, testing exactly how well their ages and morphologies match what a single 8-billion-year-old collision would predict.
The trail will not run out of questions anytime soon.
Somewhere in the direction of Cetus, 67 million light-years away, three galaxies drift through the dark — built entirely of starlight, carrying no trace of the invisible substance that shapes almost everything else.
