🔭 astrophysics

Characterizing Binary Black Hole Subpopulations in GWTC-4 with Binned Gaussian Processes: On the Origins of the $35M_{\odot}$ Peak

Using a binned Gaussian process framework on GWTC-4 data, this study identifies three distinct binary black hole subpopulations and concludes that the observed $35M_{\odot}$ mass peak likely originates from dynamically assembled systems in globular clusters with specific birth spin constraints.

Original authors: Omkar Sridhar, Anarya Ray, Vicky Kalogera

Published 2026-07-10

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Omkar Sridhar, Anarya Ray, Vicky Kalogera

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, noisy dance floor where black holes are the dancers. For years, scientists have been trying to figure out who is dancing with whom, how fast they are spinning, and where they came from. In this new study, the authors looked at a massive guest list of 153 binary black hole pairs detected by the LIGO-Virgo-KAGRA network (a catalog called GWTC-4). Instead of guessing the dance moves with a rigid script, they used a super-flexible, data-driven tool called "binned Gaussian processes." Think of this tool as a smart, shape-shifting net that can catch the exact shape of the crowd without forcing them into a pre-made mold.

When they sorted the dancers, they didn't just see one big mixed group. They found three distinct subpopulations, like three different dance crews with their own unique styles, moves, and origins.

The Three Dance Crews

The Low-Mass Crew (5 to 31.6 solar masses): These are the lighter dancers. They have a sharp peak in their numbers around 10 solar masses. Their style is characterized by partners who are somewhat different sizes (mass ratios between 0.6 and 0.8) and spins that are mostly aligned with their orbit, pointing in the same direction. The authors suggest this crew likely formed from isolated binary evolution, where two stars were born together and danced their way to a merger without outside help.
The Middle-Mass Crew (31.6 to 44.2 solar masses): This is the star of the show. This group is responsible for the famous "35 solar mass peak" that scientists have been debating. What makes them special? They are almost exclusively equal-mass pairs (partners of the exact same size) and their spins are isotropic, meaning they point in random directions, like a crowd spinning in all different ways. The authors argue that this specific combination of features is a fingerprint of dynamical formation in globular clusters. These are dense star clusters where black holes bump into each other and pair up by chance, rather than being born together.
The High-Mass Crew (Above 44.2 solar masses): These are the heavyweights. They show a broad mix of sizes and spins, with a hint of a bump around 60–70 solar masses. Because there are fewer of them, the authors note that their exact shape is partly influenced by their assumptions (priors), but they likely come from a mix of different formation channels, including things like hierarchical mergers (dancers who have merged before) or exotic growth in dense environments.

Cracking the 35 Solar Mass Mystery

The big question the paper tackles is: Where does that 35 solar mass peak come from?

Some scientists thought it was just a pile-up of stars dying in a specific way (pair-instability supernovae). However, the authors' analysis suggests that only the Middle-Mass Crew shows this peak. The Low-Mass and High-Mass crews actually show a steady drop-off in numbers in that 30–40 solar mass range.

The Middle-Mass Crew's "random spin" and "equal partner" style is a strong clue. The authors compared their findings to computer simulations of black holes in globular clusters. They found that only simulations where black holes are born with a specific, moderate spin (between 0.1 and 0.2) match the Middle-Mass Crew's data perfectly. If the birth spins were higher or lower, the simulation wouldn't look like the real data.

What the Paper Rules Out

The authors are careful to say what this data doesn't support for the 35 solar mass peak:

It is unlikely that the peak comes from isolated binary evolution (stars born together), because those systems usually have aligned spins, not the random ones seen in the Middle-Mass Crew.
It is unlikely that active galactic nuclei (supermassive black hole disks) are the main source, as those environments should produce a mix of equal and unequal mass pairs, not the strict equal-mass preference seen here.
It is unlikely that repeated hierarchical mergers (black holes merging multiple times) are the main cause, because those would create very high spins, whereas the Middle-Mass Crew has small, random spins.
They also found no significant evidence for a correlation between the mass ratio and the spin direction across the whole population. In other words, knowing the size of the partners doesn't tell you how they are spinning.

How Sure Are They?

The authors are confident that the Middle-Mass Crew is distinct from the others, with a significance of more than 90% in the 30–40 solar mass range. They estimate the merger rate for these globular cluster systems to be 0.69 +0.23 −0.33 Gpc⁻³ yr⁻¹. This number fits well with most theoretical predictions, which range from 0.2 to 57 Gpc⁻³ yr⁻¹.

However, they don't claim this is a solved mystery. They explicitly state that while the evidence points strongly to globular clusters being the origin of the 35 solar mass peak, more data is needed to be absolutely certain. They also note that their flexible model is great for finding patterns but might not be the perfect tool to isolate every single detail. They suggest that future studies with even more detections and targeted models will be necessary to rigorously lock down this interpretation.

In short, the paper suggests that the universe's "35 solar mass peak" is likely a dance floor in a crowded globular cluster, where black holes bump into each other, pair up with identical partners, and spin in random directions—provided they were born with just the right amount of spin. But until the dance floor gets even more crowded with new data, the story remains a very strong hypothesis rather than a final fact.

Technical Summary: Characterizing Binary Black Hole Subpopulations in GWTC-4 with Binned Gaussian Processes

Problem Statement
Understanding the astrophysical origins of binary black holes (BBHs) requires accurate modeling of multi-dimensional population properties, specifically the correlations between primary masses ( $m_1$ ), mass ratios ( $q$ ), and effective inspiral spins ( $\chi_{\text{eff}}$ ). While the fourth gravitational wave transient catalog (GWTC-4) has confirmed the existence of a sharp peak near $10 M_\odot$ and an over-abundance of systems near $35 M_\odot$ , the physical origin of the $35 M_\odot$ feature remains debated. Previous hypotheses range from pair-instability supernovae (PISNe) pile-ups to dynamical formation in globular clusters (GCs). However, existing studies have often relied on strongly parametrized models or failed to simultaneously constrain the joint distribution of masses, mass ratios, and spins with sufficient flexibility to distinguish between formation channels. Specifically, it remains unclear whether the $35 M_\odot$ feature arises exclusively from a dynamically formed subpopulation or if it is a composite of multiple channels.

Methodology
The authors employ a data-driven, non-parametric framework based on Binned Gaussian Processes (BGPs) to characterize the joint merger rate density of BBHs as a function of $m_1$ , $q$ , and $\chi_{\text{eff}}$ .

Model Construction: The merger rate density is modeled as a piecewise binned function in three dimensions, treated as a hyperparameter to be inferred in each bin. The joint distribution is constructed without assuming specific astrophysically motivated functional forms, allowing for the discovery of arbitrary correlations up to the chosen bin resolution.
Inference: Bayesian hierarchical inference is used, modeling BBH events as an inhomogeneous Poisson process. The logarithmic rate density is regularized using a Gaussian Process prior with an exponential quadratic kernel to handle sparse data regions.
Data and Validation: The analysis utilizes 153 confidently identified BBH events from GWTC-4. The authors validate their inference by varying bin resolutions and analyzing a large simulated catalog to ensure results are not prior-driven. They also compute Jensen-Shannon (JS) divergences between conditional distributions to quantitatively assess the distinctness of subpopulations.
Comparison: Inferred distributions are compared against theoretical simulations of dynamical BBH formation in globular clusters (using Cluster Monte Carlo code) and other formation scenarios (isolated binaries, AGN disks, hierarchical mergers).

Key Results
The analysis identifies three distinct subpopulations in the GWTC-4 sample, defined by their primary mass ranges and unique conditional distributions of mass ratio and effective spin:

Subpopulation 1 ( $5 < m_1 < 31.6 M_\odot$ ):
- Dominates the low-mass spectrum with a sharp peak near $10 M_\odot$ .
- Characterized by a mass-ratio distribution peaking at $q \in (0.6, 0.8)$ .
- Exhibits a positively skewed effective spin distribution (peaked away from $\chi_{\text{eff}} = 0$ ), indicating preferential spin alignment and non-negligible spin magnitudes.
- Does not exhibit the $35 M_\odot$ peak.
Subpopulation 2 ( $31.6 < m_1 < 44.2 M_\odot$ ):
- Exclusively contributes to the $35 M_\odot$ peak.
- Characterized by a strong preference for equal-mass systems ( $q$ distribution railing against $q=1$ ).
- Displays a symmetric effective spin distribution peaked narrowly at $\chi_{\text{eff}} \approx 0$ , indicative of isotropic spin orientations.
- The mass spectrum of this subpopulation differs from the others by more than 90% significance in the $30-40 M_\odot$ range.
Subpopulation 3 ( $m_1 > 44.2 M_\odot$ ):
- Dominates the high-mass end ( $>45 M_\odot$ ).
- Shows a broad, nearly flat mass-ratio distribution and a broad, positively skewed effective spin distribution.
- Displays a hint of a bump near $60-70 M_\odot$ and a fall-off above $70 M_\odot$ .
- In this high-mass regime, the inference is partially prior-driven due to data scarcity.

Correlations and Constraints

Mass-Ratio vs. Spin: The study finds no significant evidence for an intrinsic correlation between mass ratio ( $q$ ) and effective spin ( $\chi_{\text{eff}}$ ) across the entire population or within specific mass ranges. This suggests that the observed mass-spin correlations are not artifacts of marginalizing over an underlying $q-\chi_{\text{eff}}$ correlation.
Globular Cluster Constraints: By comparing Subpopulation 2 with theoretical simulations, the authors find that this subpopulation is fully consistent with dynamically assembled systems in globular clusters, provided black hole birth spins are in the range 0.1–0.2. Simulations with birth spins outside this range produce effective spin distributions inconsistent with the data.
Merger Rate: The merger rate for BBHs in globular clusters is constrained to a lower bound of $0.69^{+0.23}_{-0.33} \text{ Gpc}^{-3} \text{ yr}^{-1}$ , consistent with most theoretical predictions.

Significance and Claims
The paper claims that the $35 M_\odot$ feature in the BBH mass spectrum likely originates from a dynamically formed subpopulation of BBHs in globular clusters.

The authors argue that the specific combination of equal mass preference and isotropic spins in Subpopulation 2 is inconsistent with isolated binary evolution (which predicts aligned spins), AGN disk formation (which supports unequal masses), and hierarchical mergers (which predict high spins).
They conclude that while dynamical formation in GCs likely dominates the $30-40 M_\odot$ range, the other two subpopulations require substantial contributions from alternative formation channels.
The authors maintain a modest stance, noting that while the evidence for a dynamical origin is strong, the BGP framework alone cannot fully isolate subpopulations or provide Bayesian evidences for model comparison. They emphasize that targeted parametric models and future data from the LVK are necessary to rigorously establish this interpretation and constrain higher-dimensional correlations (e.g., precessing spins).

More like this