Binary black hole population inference combining confident and marginal events from the $\tt{IAS\text{-}HM}$ search pipeline

TL;DR

This paper develops a Bayesian method to include both confident and marginal binary black hole merger events in population analyses, revealing insights into merger rates, redshift evolution, and mass ratios from the LVK O3 data.

Contribution

It introduces a Bayesian framework that incorporates marginal events into population inference, extending beyond traditional threshold-based analyses.

Findings

Inclusion of marginal events increases inferred merger rate density.

Results suggest a stronger redshift evolution in merger rates.

Mass ratio distribution is consistent with being flat.

Abstract

We present the population properties of binary black hole mergers identified by the $\tt{IAS\text{-}HM}$ pipeline (which incorporates higher-order modes in the search templates) during the third observing run (O3) of the LIGO, Virgo, and KAGRA (LVK) detectors. In our population inference analysis, instead of only using events above a sharp cut based on a particular detection threshold (e.g., false alarm rate), we use a Bayesian framework to consistently include both marginal and confident events. We find that our inference based solely on highly significant events ($p_{\mathrm{astro}} \sim 1$) is broadly consistent with the GWTC-3 population analysis performed by the LVK collaboration. However, incorporating marginal events into the analysis leads to a preference for stronger redshift evolution in the merger rate and an increased density of asymmetric mass-ratio mergers relative to the GWTC-3 analysis, while remaining within its allowed parameter ranges. Using simple parametric models to describe the binary black hole population, we estimate a merger rate density of $32.4^{+18.5}_{-12.2}\ \mathrm{Gpc}^{-3}\,\mathrm{yr}^{-1}$ at redshift $z = 0.2$, and a redshift evolution parameter of $\kappa = 4.4^{+1.9}_{-2.0}$. Assuming a power-law form for the mass ratio distribution ($\propto q^{\beta}$), we infer $\beta = 0.1^{+1.9}_{-1.4}$, indicating a relatively flat distribution. These results highlight the potential impact of marginal events on population inferences and motivate future analyses with data from upcoming observing runs.