Gravitational Waves Reveal a Hidden Population of Merging Black Holes

The Universe Has Been Hiding Its Most Extreme Mergers

In 2015, when LIGO first heard the chirp of two black holes colliding a billion light years away, the sound was almost too clean. Two black holes, one 36 times the mass of the Sun, the other 29, spiraling together, ringing spacetime like a bell. It was exactly what theorists had predicted: a neat, almost boringly perfect signal from a binary system that looked like it had been designed in a textbook.

The universe, it turns out, was just warming up.

By the time the third LIGO-Virgo observing run ended in 2020, the detectors had catalogued 90 distinct gravitational wave events. Ninety separate collisions between dead stars and black holes. And when a team of 1,600 scientists led by Rich Abbott and Thomas Abbott sat down to analyze what they had found, the picture that emerged was not neat at all. It was messy, surprising, and full of holes in all the wrong places.

The universe is making black holes in ways we did not expect, merging them at rates that challenge our models, and leaving behind a population of objects that seem to defy the neat categories we invented for them.

What Exactly Did They Find?

The analysis, published in Physical Review X in 2023 as the definitive population study of the first three observing runs, is not about any single merger. It is about the census. The authors took every gravitational wave detection from the Gravitational Wave Transient Catalog 3 (GWTC-3) and asked: what does the whole population look like? Not just the loudest events, but the quiet ones too. The ones that barely registered above the noise. The ones that might have been missed.

They found three distinct classes of mergers: binary black holes, binary neutron stars, and neutron star-black hole pairs. Each class has its own story. But the black holes are where things get strange.

The Black Hole Mass Gap That Wasn't

For decades, astrophysicists believed in a clean separation between neutron stars and black holes. Neutron stars, the collapsed cores of dead stars, max out around 2.1 to 2.3 solar masses. Anything heavier should collapse into a black hole. But there was a catch: supernova simulations suggested that stars between roughly 2.5 and 5 solar masses might not form black holes at all. They might just... disappear. Explode completely, leaving nothing behind. This region became known as the "lower mass gap."

The LIGO-Virgo data says something different.

The authors found a neutron star mass distribution that extends from about 1.2 to 2.0 solar masses, with a sharp drop-off after the expected maximum neutron star mass (Abbott et al., 2023). But they could not confirm or rule out the existence of a lower mass gap. The data is still too sparse. The gap might exist, or it might be filled with black holes so small they blur the line between the two classes of objects.

This is not a small ambiguity. It means we do not actually know where neutron stars end and black holes begin. The neat boundary we drew in textbooks is a guess.

The Mass Distribution Has Peaks Where Theory Says There Shouldn't Be Peaks

If black holes formed randomly from collapsing stars, you would expect their masses to follow a smooth, power law distribution. More small ones, fewer big ones, no bumps.

The data shows bumps.

The authors found localized overdensities in the black hole mass distribution at two specific chirp masses: 8.3 solar masses and 27.9 solar masses (Abbott et al., 2023). A chirp mass is a particular combination of the two black hole masses that determines how the gravitational wave signal evolves. These peaks are statistically significant. They are not noise.

Something is preferentially producing black holes at these masses. The leading hypothesis is that these black holes formed in dense star clusters, where repeated mergers and dynamical interactions sculpt the mass distribution. But that is a hypothesis, not a conclusion. The peaks could also come from primordial black holes formed in the early universe, or from a population of black holes born from a specific type of massive star that no longer exists.

No Upper Mass Gap, But Something Else Is Going On

There is a theoretical prediction that black holes above about 60 solar masses should be rare. The physics of pair instability supernovae suggests that very massive stars should blow themselves apart rather than collapse into black holes. This creates an "upper mass gap" between roughly 60 and 120 solar masses.

The authors found no evidence of a strongly suppressed merger rate above 60 solar masses (Abbott et al., 2023). Black holes in that range exist and merge. The upper mass gap, if it exists, is not as clean as predicted.

But this does not mean the theory is wrong. It could mean that black holes above 60 solar masses form through mergers of smaller black holes, not through direct stellar collapse. A 70 solar mass black hole might be the child of two 35 solar mass parents, not a single dead star.

The Spin Problem: Why Are Black Holes So Slow?

Black holes spin. When they form from collapsing stars, they inherit the angular momentum of their parent star. You might expect them to spin fast.

The data says they do not.

The authors found that half of all observed black holes have spin magnitudes below about 0.25 (Abbott et al., 2023). That is slow. Really slow. If black holes formed in isolation from rapidly rotating massive stars, you would expect spins closer to 0.7 or 0.8.

There are two possible explanations. Either black holes are born spinning slowly, which would require a fundamental revision of how massive stars collapse. Or they are born spinning fast but then slow down through interactions with their environment before they merge. Both options are interesting. Both options imply that our models of stellar evolution are missing something.

The Alignment Problem: Black Holes Don't Line Up

When two black holes orbit each other, their spins should eventually align with the orbital plane. That is what happens in isolated binary systems. Tidal forces and mass transfer nudge the spins into alignment over time.

The authors found that while the majority of spins are preferentially aligned with the orbital angular momentum, there is clear evidence of antialigned spins among the binary population (Abbott et al., 2023). Some black holes are spinning in the opposite direction of their orbit.

This is hard to explain if the black holes formed together and stayed together. It is easy to explain if they met later in life, in a dense cluster, captured into orbit by gravitational interactions. An antialigned spin is a calling card of a dynamical origin.

The Rate Is Higher Than Expected, And It's Growing

How often do black holes merge? The authors calculated the binary black hole merger rate at a fiducial redshift of 0.2 (roughly 2.5 billion years ago) to be between 17.9 and 44 events per cubic gigaparsec per year (Abbott et al., 2023). That is a lot. For comparison, the Milky Way is about 0.03 cubic megaparsecs. Scale that up and you get roughly one merger every few minutes somewhere in the observable universe.

But here is the kicker: the rate increases with redshift. The farther back in time you look, the more mergers there are. The authors found that the merger rate scales as (1+z)^kappa, with kappa = 2.9, give or take 1.7 (Abbott et al., 2023). For redshift less than 1, that means the merger rate was about 2.9 times higher at a given lookback time than it is today.

This makes sense if black hole formation was more common in the early universe, when star formation rates were higher and galaxies were more active. But it also means that the current merger rate is not the whole story. The universe was a much more violent place a few billion years ago.

The Neutron Star Black Hole Problem

Neutron star black hole mergers are the rarest class of events in the catalog. The authors inferred a merger rate between 7.8 and 140 events per cubic gigaparsec per year (Abbott et al., 2023). That is a wide range, reflecting the small number of detections.

But the interesting thing is not the rate. It is the masses. The neutron star mass distribution, inferred from both binary neutron star and neutron star black hole systems, is broad and relatively flat, extending from 1.2 to 2.0 solar masses (Abbott et al., 2023). That flatness is surprising. It suggests that neutron stars do not have a preferred mass. They come in all sizes, up to the maximum allowed by physics.

This matters because neutron star mass tells you about the equation of state of nuclear matter. The densest stuff in the universe, compressed to the limit of what physics allows, and it turns out it can take many forms.

How Did They Actually Do This?

The methodology is worth understanding because it is not straightforward. You cannot just count events. Gravitational wave detectors are not equally sensitive to all mergers. A 10 solar mass black hole merger at redshift 1 produces a much louder signal than a 50 solar mass merger at the same distance. The detectors are also more sensitive to certain orientations and sky positions.

The authors used a technique called hierarchical Bayesian inference. They built a model of the population that includes the mass distribution, spin distribution, redshift distribution, and merger rate. Then they asked: given the observed events, what population parameters are most likely? They accounted for selection effects, the fact that the detectors miss most events, and the fact that the noise in the detectors can mimic signals.

This is not a simple counting exercise. It is a statistical inversion problem, and the authors were careful to report uncertainties. The 90 percent credible intervals on every parameter are wide. The data is still sparse enough that the population model is not uniquely determined.

What This Does Not Prove

The paper is careful about what it claims, and the limitations are instructive.

First, the authors cannot confirm or rule out the existence of a lower mass gap between neutron stars and black holes (Abbott et al., 2023). The data is consistent with a gap, but it is also consistent with no gap. More events are needed.

Second, the evidence for antialigned spins is present but not overwhelming. The authors report that they "infer evidence of antialigned spins among the binary population" (Abbott et al., 2023). That is not the same as a detection. It is a statistical preference that could change with more data.

Third, the redshift evolution of the merger rate is constrained only for redshift less than about 1. Beyond that, the data is too sparse to say anything definitive. The universe at high redshift remains a mystery.

Fourth, the mass peaks at 8.3 and 27.9 solar masses are real in the data, but their physical origin is unknown. They could come from multiple formation channels, or they could be a statistical fluctuation that will disappear with more events.

What This Actually Means

The paper changes how we think about black holes in several concrete ways.

• ▸The neat categories we use for dead stars are wrong. There is no clean boundary between neutron stars and black holes, and the upper mass gap is not as clean as predicted. The universe makes compact objects on a continuum, and our classification schemes are human inventions that do not match reality.

• ▸Most black holes form in dense environments, not in isolation. The evidence from spins and alignment points strongly toward dynamical formation in star clusters or galactic nuclei. This means that the standard model of binary evolution, where two stars are born together and stay together, is not the dominant channel for black hole mergers.

• ▸The merger rate is higher than expected, and it was even higher in the past. This has implications for the stochastic gravitational wave background, the total mass in black holes in the universe, and the rate at which black holes grow through mergers.

• ▸Black holes are born spinning slowly, or they spin down quickly. Either way, something is suppressing black hole spins relative to what stellar evolution models predict. This is a clue about the physics of core collapse and the role of angular momentum transport in massive stars.

• ▸The mass distribution of merging black holes is not smooth. The peaks at 8.3 and 27.9 solar masses are real features that demand an explanation. They could point to a specific formation channel, a preferred mass scale for black hole formation, or a population of black holes that formed in the early universe.

• ▸We need more data, and we are getting it. The fourth LIGO-Virgo-KAGRA observing run is already underway, with improved sensitivity. Each new event adds statistical power to the population analysis. The current catalog of 90 events is a beginning, not an end.

The universe is not hiding its secrets. It is broadcasting them in gravitational waves, waiting for us to build better ears. And every time we listen, we learn that the cosmos is stranger and more interesting than we imagined.