In 2015, when scientists, for the first time ever, directly detected gravitational waves—ripples in spacetime—from colliding black holes, the result came as a shock to some astronomers. Based on previous studies of black holes using x-rays, many experts expected each member of a merging pair would typically weigh about 10 times the sun’s mass, but the 2015 merger featured twin giants three times that heavy. And the Laser Interferometer Gravitational-Wave Observatory (LIGO), which enabled the discovery, has been spotting bizarrely big pairs ever since.

Black holes can certainly get supermassive—leviathans weighing billions of suns lurk at the hearts of most large galaxies. The question is: How could they grow so huge? Much of their bulk must be acquired after birth as they feast on gas and stars, but some theories suggest that mergers or chains of mergers may form a supermassive black hole’s initial seed. Black hole matchmaking in the loneliness of space is not easy, though, so astrophysicists still puzzle over what circumstances could bring the objects together.

An emerging theory holds that LIGO’s heavyweights arise near the cores of colossal galaxies, where violently incandescent disks of gas whirl around central supermassive black holes. Thanks to that omnipresent gas, these so-called active galactic nuclei (AGNs) could be factories for building big black holes out of smaller ones. If so, gravitational-wave detectors such as LIGO should be able to tease out signs of the hierarchical assembly of these swollen giants. Although the overall number of detections remains too low for conclusive classifications, some researchers have recently pointed to two hefty mergers as tantalizing hints of what AGN-facilitated black hole fusions might look like—a step toward using gravitational waves to study not only black holes but also the stars and galaxies that birthed them.

Figuring out how black hole pairs form “tells us a lot about stars,” says Maya Fishbach, a LIGO member, who was not involved in the recent research. “Stars are the building blocks of galaxies. They’re the atoms of astronomy.”

Run-of-the-mill black holes, born from the remnants of an exploded star, would typically start off in orbits skewed against the plane of an AGN’s disk of gas. Each time they would dip into that disk, however, friction would slow them and tip their paths in line with the disk. Once embedded, uneven pressures may shepherd those black holes from their initially scattered locations into special rings around the galaxy’s central black hole—a trapping process analogous to the one that forms the seeds of planets in dusty disks around a single star but with black holes instead of microscopic piles of dust.

Imre Bartos, a LIGO collaboration member at the University of Florida, estimates that these galactic “migration traps” can quickly collect tens of thousands of black holes, many of which will get close enough to pair off. Then friction from the lingering gas would drive them to collide 1,000 times earlier than they otherwise would in empty space. “They will be forced to merge together,” he says. It’s like “a black hole assembly line, where we are adding black holes one after another.”

Most of the universe’s black hole mergers are thought to be one-off finales between stellar binaries—star couplets that were born, lived and died together—of moderate mass. But if AGN disks really are cranking out large black holes made from small ones, that population should eventually stand out in two ways from gravitational-wave observations.

Now in its third observing run, LIGO announced dozens of preliminary gravitational-wave candidates for astronomical observation this year. But only 10 black hole mergers appear in the published catalogue from its first two observing runs, and nine of them seem to have come from pairs that spun slowly or not at all. A twirling crash—as would happen in an AGN disk—would, however, spin merging black holes up, typically causing subsequent generational mergers to spin even faster. Specifically, two black holes of even mass should spin at 70 percent of a theoretical top speed after colliding, so Bartos and his colleagues are on the lookout for collisions between already whirling dervishes.

They are also watching for mammoth mergers. Stars above a certain size are thought to undergo supernovae so savage that they blow their core to smithereens, preventing them from collapsing to form black holes. Theorists are unsure where the limit lies, but many expect the mass of stellar black holes to top out around that of 50 suns. “If you see a single event with 80 solar masses,” says Davide Gerosa, an astrophysicist at the University of Birmingham in England, “that’s a strong signature of some exotic formation channel.” Black hole nurseries would be one explanation for heavy outliers.

Recent data, however, may complicate that simple picture. Last week astronomers announced the discovery of what seems to be a huge black hole born together with a partner star. If the current controversial estimate of the former’s mass—roughly 70 suns—stands, then the 50-sun limit may be a less clear-cut line for generational mergers. “I don’t think we’ve been hitting this problem hard enough,” Gerosa says.

Globular clusters, small clumps of stars within a galaxy, are another possible cosmic construction site for abnormally heavy, quickly spinning black holes. In these star-rich regions, black holes presumably could form dense crowds in which they would occasionally bump into one another. But recent research by Gerosa and one of his colleagues found that the recoil from such collisions would likely eject most pairs from the globular cluster, preventing them from finding future companions to merge with. Larger groups of stars, such as those within AGN disks, are more likely locations for strings of mergers, Gerosa says, because black holes there require much more recoil to escape.

After years of theoretical speculation, some researchers are starting to see hints of what may be an extra heavy population beginning to reveal itself. LIGO’s heaviest catalogued merger, GW170729, is exactly the kind an AGN disk would produce, Bartos and his colleagues proposed in Physical Review Letters in November. In that event, one of the black holes weighed roughly 50 suns, and a measure of the pair’s collective spin clocked them turning at about 40 percent of top speed before the merger—a hint that that an earlier collision could have spun them up.

Another candidate AGN-driven event appeared in October, when researchers at the Institute for Advanced Study posted a preprint paper announcing two possible mergers from LIGO’s data that, while not meeting LIGO’s criteria for publication, may well be genuine. Called GW170817A, and with a mass of roughly 56 suns and a combined premerger spin of 50 percent the maximal value, this candidate merger matches predictions for an AGN collision even more closely than GW170729, according to a not yet peer-reviewed preprint study posted on arXiv.org in late November. “This is exactly the same kind of event,” Bartos says.

Neither candidate is a smoking gun for an AGN black hole assembly line, however. GW170817A only registered in one of LIGO’s two detectors—a potential sign that it was a false alarm arising from contaminating noise on Earth rather than some far-off celestial cataclysm. Moreover because only a small fraction of the universe’s stars reside in AGNs, Bartos’s group concluded that the suggestive properties of these two mergers are just as likely to reflect normal binary black holes that just happened to be extra heavy and to spin extra fast as they are AGN black holes.

Other researchers agree that AGN disks could occasionally smack black holes together but stress that the community will need more data, as well as better predictions, to conclusively prove the reality of this rare collision type or others. “I don’t think there’s anyone who would be able to pick one side, because they’ll know that if they’re proven wrong, it will be in, like, a year,” Fishbach says.

Regardless the ability to distinguish one-off stellar binaries from AGN assembly lines and other putative production mechanisms for black hole mergers is coming. As LIGO’s catalogue swells, categories based on spin and mass should become much clearer. Bartos suggests that traditional astronomy based on light rather than gravitational waves could help, too. Gravitational crashes that align in the sky with known AGNs will supply further hints. And if astronomers can use their telescopes to rapidly observe gravitational-wave sources with AGN signatures as they are detected, a recent publication in the Astrophysical Journal Letters proposes, they may glimpse flashes of light hypothesized to come from postcollision shock waves in the gas.

In the midst of building this new black hole taxonomy, astrophysicists are already brainstorming what they will be able to do with it. Light reveals what a galaxy is made of, says Katelyn Breivik, an astrophysicist at the Canadian Institute for Theoretical Astrophysics, but gravitational waves may unmask its more subtle dynamics. “If you have black holes that are embedded in these disks,” she says, “they are like literal gravitational probes into the shape of these disks,” revealing mass and motion.

While those probes have yet to materialize, LIGO’s big, spinning black holes encourage Bartos that they are not far off. “I was used to predicting the far future,” he says. “Coming from there to having these black holes that basically reproduce what you’re predicting is super exciting.”