By the fifth observing run, the gravitational-wave observatories LIGO, Virgo, and KAGRA had amassed a hundred events. Like an archaeologist sifting through cultural strata, astrophysicists are beginning to discern distinct populations within this sample. Previously they saw only peaks and dips in the mass distribution; now the outlines of entire “epochs” of binary black hole birth are emerging. Figuring out how many such epochs exist and what sets them apart means reconstructing the evolutionary paths of massive stars and the dynamics of clusters. The new analysis tackles exactly this question by marrying non-parametric structure detection with precise parametric modeling.
First, using the –π (pi-dash) maximum-likelihood method, which imposes no pre-set distribution shape, the authors reconstructed the density on the plane of component masses (m1, m2). The pi method represents the population as a set of delta functions; their clumps, like clusters of artifacts at a dig site, clearly outlined four regions. Then each cluster was fitted with its own parametric model—with its own laws for masses, mass ratio, and gravitational waves. The findings rest on 259 events from GWTC-5, with careful accounting of the detectors’ numerical simulation.
The four-subpopulation model decisively outperformed the standard one: the log Bayes factor reached 19.2. The first group, stellar evolution, is the richest layer, accounting for roughly 70% of all mergers. Its black holes cluster around 10 M⊙, the two components are nearly equal, and the gravitational waves are small and aligned with the orbit. Immediately beyond it, a dip appears at ~14 M⊙, where finds are 26 times scarcer than expected for a smooth distribution. Next come two intermediate layers. The “horizontal” layer resembles unequal pairs: the secondary mass stalls near 10 M⊙ while the primary reaches up to 40 M⊙; the gravitational waves here are mixed, and the rate drops steeply with mass. The “diagonal” layer consists of almost equal partners peaking at 31 M⊙; their effective gravitational waves are small, but the orbital tilts may be chaotic. Finally, the top layer, high-mass, starts above 47 M⊙: a broad mass-ratio distribution and large, randomly oriented gravitational waves betray globular cluster mergers—products of encounters in dense environments.
This layering means that the population of merging black holes is not monolithic but a mosaic of different evolutionary pathways. The stellar evolution layer was likely born from isolated evolution and peculiar supernova explosions that leave behind compact cores. The globular cluster layer may have been forged in globular clusters, where close encounters create massive pairs with misaligned spins. The binary star branch remains a puzzle: it requires either a rare capture of a companion from the low-mass-peak era or an exotic mass-transfer stage. And the top layer is almost irrefutable evidence of repeated mergers. This entire stratigraphy turns each observation into an archaeological fact, letting us trace the life-and-death history of massive stars.
Future observing seasons of LIGO, Virgo, and KAGRA—and eventually next-generation gravitational waves telescopes like the Einstein Telescope—will multiply the number of detections by tens. Then it will be possible to test whether the identified layers persist and, perhaps, to split them even finer—for instance, separating the contribution of globular clusters from isolated evolution or tracking population changes with redshift. Joint analysis with electromagnetic and neutrino signals will open an era of multi-messenger black hole.
The results directly impact the theory of massive star formation evolution, the dynamics of globular cluster clusters, the cosmological interpretation of the gravitational waves background, and population-analysis methods.
Pin down the physical nature of the horizontal component—its origin is the most uncertain. Verify the detected structures with alternative non-parametric approaches and binary-disruption simulations.
The work touches on the problem of supernova mechanisms and the transition from successful explosion to direct collapse, as well as the hunt for black hole mergers—an important step toward understanding the formation of intermediate-mass black holes.
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The authors reconstructed the density on the mass component plane (m1, m2) using the nonparametric method π, identified four clusters, and fitted a parametric model for each with its own mass, mass ratio, and spin distributions. The analysis relied on 259 GWTC-5 events, accounting for selection effects, and showed that the population of merging black holes splits into four subpopulations with distinct evolutionary pathways.
Where its strength is
Combining nonparametric structure detection with subsequent parametric modeling of clusters allowed not just identifying peaks but characterizing each subpopulation with a full set of parameters. This provides a basis for direct physical interpretation: the low-mass component is linked to isolated evolution, the diagonal one to cluster dynamics, and the high-mass one to hierarchical mergers.
Where it could go next
Worth trying
Why this may matter
If eccentricity or dynamical sorting indeed turn out to be markers of specific subpopulations, this would turn morphological classification into a physical map of formation channels. Then, from a single event, one could indicate with high probability where and how a black hole pair was born.
Nearby works in our archive
The closest work in the archive sits at 0.69 by meaning — the usual level for our corpus.
Four prototype events — GW170608, GW190412, GW150914, and GW170729—were already recorded in the first two observing catalogs. Only now, with more data, has it become clear that each likely represents a distinct evolutionary era.
Hierarchical mergers, in which black holes grow generation after generation, are reminiscent of the “hyperspatial” structures in Greg Egan’s novel “Schild’s Ladder,” where objects are built through successive amalgamations in boundless space.