🔭 astrophysics

Evidence for mass-dependent spin subpopulations in GWTC-4

Analyzing GWTC-4 data with a novel regularized Gaussian mixture model, this study provides strong evidence for two distinct black hole spin subpopulations that transition from slowly spinning low-mass systems to rapidly spinning high-mass systems, offering new constraints for binary formation models.

Original authors: Asad Hussain, Maximiliano Isi, Aaron Zimmerman

Published 2026-05-26

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Asad Hussain, Maximiliano Isi, Aaron Zimmerman

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, cosmic dance floor where pairs of black holes spin around each other before colliding. For years, scientists have been listening to the "music" of these collisions using gravitational wave detectors (like LIGO, Virgo, and KAGRA). But a big mystery remained: Why do some black holes spin slowly, while others spin like tops?

This paper, written by Asad Hussain and colleagues, uses the latest "playlist" of 153 black hole collisions (called GWTC-4) to solve a piece of that mystery. They discovered that the speed at which these black holes spin isn't random; it depends entirely on how heavy they are.

Here is the story of their discovery, broken down into simple concepts:

1. The Two "Dance Styles" (Subpopulations)

The researchers found that black hole pairs don't all follow the same rules. Instead, there are two distinct "dance styles" or subpopulations:

The "Slow Dancers" (Low Mass): When the black holes are relatively light (roughly 35 times the mass of our Sun or less), they almost always spin very slowly. In fact, many of them might not be spinning at all. They are like heavy, slow-moving dancers who barely turn.
The "Fast Spinners" (High Mass): When the black holes are heavier (above that 35-sun mark), the rules change. These heavyweights often spin very fast. Some spin so fast they are like high-speed tops, and interestingly, they often come in pairs where both are spinning rapidly.

2. The "Transition Zone"

The most exciting finding is the transition. It's not a sudden switch like flipping a light on and off. Instead, imagine a ramp.

Below a certain weight (around 35 solar masses), the "Slow Dancers" rule the floor.
As you go up in weight, the "Fast Spinners" start to show up more and more.
By the time you reach the heaviest black holes (around 70 solar masses), the "Fast Spinners" are the main act, though the "Slow Dancers" are still there in the background.

3. Why Does This Matter? (The "Recipe" Analogy)

Think of black holes like cookies. If you find a tray of cookies, you might wonder: "Did they all come from the same recipe?"

The "Slow Dancers" suggest a recipe where the ingredients (stars) were mixed carefully, losing their spin energy before becoming black holes. This fits the story of two stars living together peacefully in isolation.
The "Fast Spinners" suggest a different recipe. The paper hints that these heavy, fast-spinning black holes might be "second-generation" cookies. This means they were formed from the remains of a previous black hole collision. When two black holes smash together, the new one inherits a lot of spin (about 0.7 on a scale of 0 to 1). This is like a "hierarchical" family tree where the parents were already black holes.

4. The "Blurry Photo" Problem

In the past, looking at these black holes was like trying to identify people in a blurry, black-and-white photo. Scientists used to look at a "reduced" number (called effective spin) that mixed the speed and the angle of the spin together. It was hard to tell if a fast spin was due to the speed or the angle.

This paper used a new, sharper camera. Instead of looking at the blurry mix, they looked at the individual spins of both black holes in the pair. They also used a new mathematical trick (called "Gaussian mixtures") to clean up the noise in the data. This allowed them to see the distinct "Slow" and "Fast" groups clearly, which previous methods missed.

5. What They Don't Know Yet

While they found the two groups, the paper admits it's still a bit of a puzzle.

They can't say for sure exactly which cosmic environment (like a crowded star cluster or a swirling disk around a galaxy) creates the "Fast Spinners," though they have good guesses.
They also found that the "Slow Dancers" might be even slower than they thought—some might have zero spin at all. This depends on how exactly the stars spun before they died.

The Bottom Line

The universe isn't using a single recipe for black holes. It's using at least two.

Lightweight black holes are mostly slow and quiet.
Heavyweight black holes are often fast and energetic, likely because they are the children of previous cosmic collisions.

This discovery gives scientists a new target: they need to build models of how stars live and die that can explain why light black holes are slow and heavy ones are fast. It's like realizing that small cars and big trucks are built in completely different factories.

Technical Summary: Evidence for Mass-Dependent Spin Subpopulations in GWTC-4

Problem Statement
The astrophysical origin of merging binary black hole (BBH) systems remains uncertain, with various formation channels expected to imprint distinct signatures on observed mass and spin distributions. While previous population studies have identified features in the mass distribution and effective spin ( $\chi_{\text{eff}}$ ), reduced observables like $\chi_{\text{eff}}$ can obscure physically distinct spin configurations by mixing them. Furthermore, inferring the distribution of component spin magnitudes ( $\chi_1, \chi_2$ ) is technically challenging, particularly for models supporting multiple narrow populations. Standard Monte Carlo (MC) likelihood estimators often suffer from high variance when dealing with narrow distributions or concentrated subpopulations, potentially masking subtle structures in the data. This work addresses these limitations by investigating the mass dependence of spin magnitudes in the fourth LIGO–Virgo–KAGRA gravitational wave transient catalog (GWTC-4), allowing for multiple spin subpopulations whose relative fractions vary with primary mass ( $m_1$ ).

Methodology
The authors employ hierarchical Bayesian inference to estimate the hyperparameters ( $\Lambda$ ) of a population model using 153 BBH candidate signals from GWTC-4 (False Alarm Rate $< 1 \text{ yr}^{-1}$ ).

Population Model: The baseline mass and redshift model follows the fiducial strongly modeled BBH population from Abac et al. (2025a). The core innovation lies in the spin magnitude model, which assumes a mixture of two correlated, truncated 2D normal distributions (subpopulations $a$ and $b$ ) over the unit square $[0, 1] \times [0, 1]$ .
Mass-Dependent Mixing: Unlike previous constant-mixing models, this work introduces a mass-dependent mixing fraction $\eta(m_1)$ . This fraction transitions linearly from a low-mass value $\eta_s$ to a high-mass value $\eta_e$ across a transition region defined by $m_s$ and $m_e$ . The model is constrained such that subpopulation $a$ dominates at low masses ( $\eta_s \geq 0.5$ ).
Spin Tilt Models: The analysis considers two spin-tilt configurations: a "free" model where the peak of the truncated normal distribution for spin tilt is a free parameter, and a "fixed" model where the peak is constrained to $\cos \theta_i = 1$ (aligned spins). The fiducial analysis uses the free tilt model.
Likelihood Estimation (TGMM): To mitigate the variance issues inherent in standard MC estimators for narrow populations, the authors utilize Truncated Gaussian Mixture Models (TGMMs). They fit TGMMs to the underlying event posteriors and recovered injections. This allows for semi-analytic integration over the spin sector and MC integration over the mass/redshift sector, significantly reducing variance even when subpopulations are narrow or concentrated in parameter space corners.
Inference: Hyper-posteriors are sampled using the No-U-Turn Sampler (NUTS) Hamiltonian MC algorithm implemented in numpyro.

Key Contributions

Mass-Dependent Mixing: The paper introduces and constrains a model where the fraction of slowly vs. rapidly spinning BBHs varies as a function of primary mass, rather than assuming a constant global mixture.
Regularized Likelihood: The application of TGMMs to regularize the population likelihood provides a robust method for detecting narrow subpopulations that standard estimators might miss due to variance.
Component-Spin Resolution: The analysis operates directly in component-spin space ( $\chi_1, \chi_2$ ) rather than relying solely on reduced observables, allowing for the detection of specific configurations (e.g., asymmetric spins) that inform formation channels.

Results

Evidence for Two Subpopulations: The data provide strong evidence for two distinct spin-magnitude subpopulations, with a log Bayes factor of $\log_{10} B = 4^{+6}_{-2}$ favoring the two-population model over a single population.
Mass-Dependent Transition: There is strong evidence for a mass-dependent mixing fraction. The transition begins at $m_s \approx 36^{+19}_{-21} M_\odot$ $m_{s} \approx 3 6_{- 21}^{+ 19} M_{⊙}$ and ends at $m_e \approx 68^{+65}_{-31} M_\odot$ $m_{e} \approx 6 8_{- 31}^{+ 65} M_{⊙}$ .
- Low-Mass Regime ( $m_1 \lesssim 35 M_\odot$ ): The population is dominated by a slowly spinning subpopulation ( $\eta_s \approx 0.95$ ). The spins are tightly concentrated near zero, with mean values $\mu_a \approx (0.07, 0.09)$ and widths $\sigma_a \approx (0.08, 0.12)$ . The data do not exclude the hypothesis of identically vanishing spins ( $\chi_1 = \chi_2 = 0$ ).
- High-Mass Regime ( $m_1 \gtrsim 70 M_\odot$ ): The rapidly spinning subpopulation becomes significant ( $\eta_e \approx 0.17$ ). This component is broader, peaking at $\mu_b \approx (0.75, 0.69)$ with large widths ( $\sigma_b \approx (0.43, 0.49)$ ). It supports configurations where both black holes have high spins ( $\chi \sim 0.7$ ) but also retains a non-negligible tail toward asymmetric spins (e.g., $\chi_1 \sim 0.7, \chi_2 \sim 0$ ).
Correlation: There is weak evidence for anti-correlation between spins in the slowly spinning subpopulation, while the correlation in the rapidly spinning subpopulation remains consistent with the prior.
Robustness: The results are robust to leave-one-out analyses excluding the most extreme events (GW231123 and GW190517). While the constant mixing fraction model is disfavored relative to the mass-dependent model ( $\log_{10} B \approx 1.6$ ), the inferred properties of the subpopulations remain consistent across models.
Tilt Model Sensitivity: The inference of the slowly spinning population's properties is sensitive to the spin-tilt model. The "free" tilt model allows for vanishing spins, whereas the "fixed" tilt model shifts the slowly spinning peak to $\chi \approx 0.15$ .

Significance and Claims
The authors claim that their results provide a new, empirical target for BBH formation models. The findings suggest that distinct formation channels operate across the BBH mass spectrum:

Low Mass: Consistent with isolated binary evolution where efficient angular momentum transport leaves black holes with negligible natal spins.
High Mass: The presence of a rapidly spinning component with $\chi \sim 0.7$ is suggestive of hierarchical mergers (second-generation black holes) or chemically homogeneous evolution. The data support a picture where high-mass systems are dominated by this broader, rapidly spinning component, potentially filling the pair-instability mass gap.

The paper emphasizes that the mass-dependent structure of individual component spins ( $p(\chi_1, \chi_2 | m_1)$ ) offers a sharper diagnostic for distinguishing formation channels than effective spin alone. The authors note that while their results are consistent with hierarchical mergers in dense environments (globular clusters, AGN disks) and chemically homogeneous evolution, the current data do not uniquely discriminate between these specific scenarios. They conclude that future catalogs with increased event counts will be necessary to further sharpen these constraints and resolve the uncertain origins of these explosive events.