For billions of years, the Milky Way has been quietly destroying things. Globular clusters — ancient, densely packed spheres of hundreds of thousands of stars — drift into the galaxy’s gravitational field and are slowly pulled apart, their members scattered across arcs of sky that can span tens of thousands of light-years. The clusters don’t explode. They dissolve. And what they leave behind are rivers.
Astronomers call them stellar streams: long, thin trails of stars that trace the orbital paths of disrupted clusters, still moving in the same direction, still carrying the memory of a structure that may no longer exist. For decades, finding these rivers required painstaking work, and fewer than 20 had ever been confirmed. In March 2026, a team from the University of Michigan announced they had found 87 more.
The Algorithm That Changed the Count
The discovery comes from Yingtian “Bill” Chen, a doctoral candidate at the University of Michigan, working with senior author Oleg Gnedin and Adrian Price-Whelan of the Flatiron Institute. Their tool — an algorithm called StarStream — takes a fundamentally different approach from earlier searches. Rather than scanning for anything that looks elongated or stream-like, it builds a physical model of how a stream should form, then searches the data for stars that match that prediction. Theory first. Pattern second.
The team applied StarStream to observations from the European Space Agency’s Gaia spacecraft — operational from 2014 to 2025 — which mapped the positions, motions, and brightness of billions of Milky Way stars with unprecedented precision. The results, published in The Astrophysical Journal Supplement Series in March 2026, identified 87 candidate streams, all associated with globular clusters that are still intact. Some of the new streams are shorter than expected. Some are wider. Some are misaligned with their parent cluster’s orbit. All of them are things earlier searches, built for the most obvious patterns, would have passed over entirely.
“It’s like riding a bike with a bag of sand,” said Gnedin, “only the bag has a hole in it. Those grains of sand are like the stars left behind along their trajectory.”
What the Rivers Are Telling Us
Globular clusters are among the oldest objects in the galaxy — dense, gravitationally bound collections of hundreds of thousands of stars that formed in the universe’s earliest chapters, many more than ten billion years ago. They are, in a sense, fossils: survivors of the conditions that preceded everything we recognize as the modern Milky Way. You can read about how those earliest stellar populations came to be in our deep-dive on stellar nurseries — the same physics of collapse and ignition, operating at a time when the universe was barely a billion years old.
As these clusters orbit the galaxy, tidal forces tug at their edges. Stars at the leading and trailing ends of the cluster feel slightly different gravitational pulls and, over vast timescales, drift away, spreading along the orbital path into a stream. The cluster is the source; the stream is the record. What makes this scientifically powerful is that the record is not passive. The shape of a stellar stream — its width, its subtle twists, any gaps carved into its length — is a precise imprint of every gravitational influence it has encountered over billions of years. That includes the dark matter that constitutes roughly 85 percent of the Milky Way’s total mass and which direct observation has never detected. Where dark matter substructure creates a gravitational nudge, the stream bends. Reading those bends is one of the most promising methods astronomers have for mapping what otherwise cannot be seen.
Until StarStream, the sample was too small for that kind of systematic analysis. With 87 candidates — and crucially, streams whose parent clusters are still intact, allowing researchers to observe both the source and the trail simultaneously — the picture shifts. The Milky Way’s outskirts, long thought to be mostly empty space, are turning out to hold a fossil record of everything the galaxy has consumed.
What Comes Next
Not all 87 candidates will be confirmed as genuine streams. They are, for now, the best targets astronomy has for follow-up observation. The Vera Rubin Observatory, now entering operations in Chile, is capable of resolving far fainter structures than Gaia could reach, and is expected to sharpen many of these candidates into clarity. NASA’s Nancy Grace Roman Space Telescope, scheduled for launch in 2027, and the Dark Energy Spectroscopic Instrument (DESI) will add complementary datasets, building toward a map of the galaxy’s dark matter distribution that astronomers have been working toward for decades.
Chen’s StarStream algorithm will continue to run as new data arrives. More streams are almost certainly waiting in the noise — rivers that have been there for billions of years, invisible only because we weren’t asking the right questions.
Somewhere in the halo of the Milky Way, a stream of ancient stars is still tracing the orbit of a cluster that may no longer exist — patient, quiet, and precisely where the physics said it would be.
