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

TL;DR

This work tackles the origin of the 35 $M_\odot$ feature in the binary black hole population by unbiasedly dissecting GWTC-4 with a flexible, non-parametric approach. The authors model the BBH merger-rate density as a 3D, bin-based function in $(m_1,q,\chi_{\rm eff})$ using a binned Gaussian Process prior, enabling simultaneous inference of mass, mass-ratio, and spin distributions under a hierarchical Bayesian framework. They identify three subpopulations and show that only Subpopulation 2—characterized by near-equal masses and a symmetric, near-zero effective spin—produces the 35 $M_\odot$ peak, consistent with a dynamical origin in globular clusters if birth spins lie in $0.1-0.2$. By comparing to GC simulations, they place a lower bound on GC BBH merger rates of $0.69^{+0.23}_{-0.33}\rm{Gpc^{-3}\,yr^{-1}}$ and argue that GC formation plausibly dominates in the $30-40M_\odot$ range, while other channels contribute to the other mass ranges. The analysis remains robust to binning choices and highlights the need for more data and targeted models to rigorously establish the GC origin for the $35M_\odot$ feature.

Abstract

Understanding the astrophysical origins of binary black holes requires accurate and flexible modeling of multi-dimensional population properties. In this paper, using a data-driven framework based on binned Gaussian processes, we characterize the joint distribution of BBH primary masses, mass ratios, and effective inspiral spins. We identify three distinct subpopulations in the GWTC-4 sample of observations and investigate their astrophysical origins. We find that only one of the three subpopulations exhibits the $35M_{\odot}$ peak, which is characterized by a strong preference for equal mass systems and isotropic spin orientations. Our inferred distributions are consistent with a predominantly dynamical origin of this feature. By comparing with theoretical simulations, we further show that the subpopulation that exhibits the $35M_{\sun}$ peak can exclusively comprise dynamically assembled systems in globular clusters, specifically if black hole birth spins are in the range~$(0.1-0.2)$, whereas the other two subpopulations require substantial contributions from alternative formation channels. We constrain the \textit{lower bound} on the merger rate of BBHs in globular clusters to be $0.69^{+0.23}_{-0.33} \rm{Gpc}^{-3}\rm{yr}^{-1}$, which is consistent with theoretical predictions. We conclude that dynamical formation in globular clusters remains a strong candidate for the origin of this excess near $30-40M_{\odot}$ and that more data and targeted parametric models are necessary to rigorously establish this interpretation.

Figures (8)

• Figure 1: Marginal and conditional distributions of primary mass, mass-ratio and effective spin inferred from GWTC-4 for the bin choice described in Table \ref{['tab: Bin choices']}. The top panels show the marginal primary mass distribution on the left and the mass ratio distribution conditioned on three primary mass ranges to the right. The bottom panels show the effective spin distribution for three different primary mass and mass-ratio ranges to the left and the primary mass distribution conditioned on three different mass ratio and effective spin ranges on the right.
• Figure 2: The Pearson correlation coefficient posteriors between effective spin and mass-ratio for three different mass ranges, namely $m_1 \in (5.0 M_{\odot}, 31.6 M_{\odot})$, $m_1 \in (31.6 M_{\odot}, 44.2 M_{\odot})$ and $m_1 \in (44.2 M_{\odot}, 200 M_{\odot})$, along with the same marginalized across the entire primary mass space and for the prior distribution.
• Figure 3: Comparison of conditional mass-ratio and effective spin distributions for the three subpopulations with the 1G+1G component of CMC simulations (which do not include hierarchical mergers) Rodriguez:2019huv. The faintest to brightest lines correspond to birth spins of 0, 0.1, 0.2 and 0.5 respectively.
• Figure 4: The mass-ratio distribution (top) and fraction of events with $\chi_{\rm eff}<0$ (bottom) for the low-mass $(m_1<30M_{\odot})$ subpopulation.
• Figure 5: Fraction of events with $q>0.6$ (top) and $\chi_{\rm eff}<0$ (bottom) for the high-mass $(m_1>40M_{\odot})$ subpopulation. The orange line represents all hierarchical mergers from the simulations of Rodriguez:2019huv, with the models corresponding to each BH birth spin combined with equal weightage.