Evidence for Three Subpopulations of Merging Binary Black Holes at Different Primary Masses

Problem Statement
With the release of the fourth LIGO–Virgo–KAGRA gravitational-wave catalog (GWTC-4.0), the number of confidently detected binary black hole (BBH) mergers has surpassed 150. While early expectations suggested a small number of detections might suffice to identify a single primary formation channel, the data now reveals a complex population that cannot be easily explained by a single mechanism. Previous attempts to identify subpopulations using spin tilts have faced challenges due to measurement difficulties and theoretical uncertainties. Furthermore, while global correlations between parameters exist, their astrophysical interpretation remains non-trivial. The central problem addressed is whether distinct subpopulations of merging BBHs exist, characterized by clear transition boundaries in primary mass ($m_1$) and differing distributions of mass ratios ($q$) and spin magnitudes ($\chi$).

Methodology
The authors adopt a phenomenological approach to model the BBH population using the cumulative GWTC-4.0 data (excluding GW190814 due to secondary ambiguity and GW231123 due to concerns regarding waveform systematics and extreme spins). They employ hierarchical Bayesian inference to compare models with varying levels of complexity.

Key methodological features include:

• Parametrization: The primary mass distribution ($p(m_1)$) is modeled as a broken power law with two Gaussian peaks. Spin magnitudes ($p(\chi)$) and tilts ($p(\cos \theta)$) are modeled using truncated normals and mixtures of isotropic/aligned components, respectively. Redshift ($p(z)$) follows a power law.
• Transition Masses: The core innovation is allowing the distributions of $q$, $\chi$, and $\cos \theta$ to change shape at up to three transition masses, effectively creating up to four subpopulations. The functional form of the distribution may remain the same across a transition, but the hyperparameters controlling the shape are allowed to differ.
• Model Selection: The authors utilize Bayes factors to perform model selection, applying an Occam penalty to avoid overfitting. They compare a "multipop" model (allowing transitions) against a fiducial strongly-parameterized model (gwtc4-fiducial) and variants where specific features are removed.

Key Contributions and Results
The analysis provides strong evidence for at least three distinct subpopulations of merging binary black holes, separated by two transition masses: $m_t^{AB} = 27.7^{+4.1}_{-3.4} M_\odot$ and $m_t^{BC} = 40.2^{+4.7}_{-3.2} M_\odot$. The "multipop" model is preferred over the fiducial model with a natural log Bayes factor of 12.3.

The three subpopulations are characterized as follows:

1. Subpopulation A ($m_1 \lesssim 27.7 M_\odot$):

   • Mass Ratio: Exhibits a nearly flat distribution ($\beta_0 \approx -0.8$).
   • Spin: Characterized by small spin magnitudes, with support primarily for $\chi \leq 0.5^{+0.1}_{-0.1}$.
   • Interpretation: Likely includes contributions from multiple formation channels, potentially including isolated binaries with efficient angular momentum transport.

2. Subpopulation B ($27.7 M_\odot \lesssim m_1 \lesssim 40.2 M_\odot$):

   • Mass Ratio: Shows a much sharper preference for equal-mass pairing ($q \approx 1$), with $\beta_1 \approx 3.6$. This distinct pairing distribution resolves a previously reported peak in the primary mass distribution at $\sim 35 M_\odot$, suggesting that peak was an artifact of model misspecification.
   • Spin: Spin magnitude and tilt distributions are statistically indistinguishable from Subpopulation A.
   • Interpretation: The authors suggest Chemically Homogeneous Evolution (CHE) or Population III stars as potential origins, though they note challenges in reconciling the lack of spin differences with standard CHE predictions. Dynamical formation is considered less likely given the lack of distinct spin tilt distributions.

3. Subpopulation C ($m_1 \gtrsim 40.2 M_\odot$):

   • Spin: Characterized by a significantly broader spin magnitude distribution with substantial support for $\chi > 0.5$ (peaking around $\chi \sim 0.7$).
   • Mass Ratio: Shows tentative support for a preference toward $q \approx 0.5$.
   • Interpretation: The properties are largely consistent with hierarchical mergers (mergers of previous-generation black holes), particularly the broad spin distribution and the $q \approx 0.5$ preference. However, the presence of small spins within this group and the lack of a transition in spin tilts suggest that hierarchical mergers in dense clusters may not fully explain this subpopulation.

Significance and Claims
The paper claims that the data strongly prefers the interpretation that these three subpopulations exist in conjunction, rather than as a single continuous population. The identification of sharp transition boundaries in mass ratio and spin magnitude distributions implies that the assumption of separability in the population distribution of BBH parameters breaks down above $\sim 25 M_\odot$.

The authors emphasize that while their results rely on strongly-parametrized models and are subject to potential model misspecification, the statistical strength of the signal and the independent confirmation of these transition masses in other literature analyses suggest the existence of these subpopulations is real. They conclude that different formation pathways likely dominate different mass ranges, with Subpopulation B potentially linked to CHE or Population III stars, and Subpopulation C linked to hierarchical mergers, though alternative scenarios (such as accretion in AGN disks) remain possible. The study explicitly states it finds no evidence for transitions around the $10 M_\odot$ peak, leaving that substructure for future investigation.