Only 18 such “stellar streams”—objects that can help reveal clues about dark matter and the Milky Way’s evolution—had been known previously

Stellar streams are trails of stars that astronomers can study to solve mysteries about the history of our Milky Way galaxy and, potentially, the dark matter that helps shape the cosmos despite eluding direct observation.

These streams have largely been left behind by small dwarf galaxies and globular clusters, which have since petered out of existence. But there is a rare and important third type of stellar stream: streams from globular clusters that still survive.

“Previously, we knew of fewer than 20. Now we’ve found 87,” said Yingtian “Bill” Chen of the University of Michigan and an author of a new study in The Astrophysical Journal Supplement Series.

The small sample size made it difficult to disentangle which features may be idiosyncratic features of a certain stream vs. signatures with galactic implications. Now, while earning his doctorate in astronomy, Chen developed an algorithm that has more than quadrupled the number of known stellar streak candidates from still-extant globular clusters.

Although not all of the new candidates may eventually prove to be actual stellar streams, the research provides new targets for the next generation of telescopes to study.

This study, supported by NASA, was powered by data from the European Space Agency’s Gaia spacecraft, which operated from 2014 to 2025, observing billions of stars in the Milky Way. Now, more powerful tools are coming online that will enable astronomers to examine the newly discovered candidates in more detail to verify which ones contain salient clues about dark matter and our galactic history.

“Of these 87 candidates, we have relatively low confidence in some of them because of background contamination. But that can be largely improved with future observations,” Chen said. “Gaia is relatively old, but there will be new surveys including NASA’s Roman Space Telescope, the Vera Rubin Observatory and the Dark Energy Spectroscopic Instrument, or DESI.”

The Rubin Observatory started collecting data last summer and the Nancy Grace Space Telescope is currently scheduled to launch in 2027. DESI began its survey in 2021.

Clusters and streams

One of the biggest challenges in finding stellar streams is that they are simply hard to see. There are hundreds of billions of stars contained in the famous spiral structure that many of us think of when we hear the words, “Milky Way.” But orbiting the Milky Way itself are dwarf galaxies, which have far fewer stars, and globular clusters, which are even smaller still.

It’s these comparatively minuscule stellar congregations that give rise to stellar streams. The tidal interplay between these tinier groups and the much larger Milky Way can result in stars being plucked from the smaller stellar swarms along their orbital paths.

“It’s like riding a bike with a bag of sand, only the bag has a hole in it,” said Oleg Gnedin, a senior author of the new study and a U-M professor of astronomy. “Those grains of sand are like the stars left behind along their trajectory.”

The shapes and sizes of these streams contain clues about the gravitational energy the clusters experience and how the Milky Way’s mass, which includes an ample amount of dark matter, is distributed. The first stellar streams, observed decades ago, were left by dwarf galaxies and were larger and more spread out compared to streams from globular clusters, which were discovered more recently.

These smaller streams have almost exclusively been discovered serendipitously by astronomers who happened to spot them in images from missions like Gaia, Gnedin said. The U-M team took a more systematic approach, first developing a physical model to account for how these streams form. Chen then used that model for a computer algorithm called StarStream to analyze Gaia’s data for signs of streams within this new, physical context.

“It turns out that it’s a lot easier to find things when you have a theoretical expectation of what you’re looking for when you have a simple phenomenological picture,” Gnedin said.

While the team was excited to find this many candidates, the researchers are already looking forward to what their algorithm can do with the next generation of data.

“It’ll be very easy to adjust the algorithm to future missions,” Chen said. “Once we have the data, it will be very straightforward to apply it.”

Adrian Price-Whelan, an associate research scientist at the Flatiron Institute also contributed to the study.