Impact of eccentricity on the population properties of neutron star - black hole mergers

Abstract

We revisit the population properties of neutron star-black hole (NSBH) mergers using low-mass compact binary coalescences reported through GWTC-4. Employing pyEFPE, an inspiral-only waveform model that captures both orbital eccentricity and spin-induced precession, we reanalyse all binary neutron star (BNS) and NSBH events observed via gravitational waves. The BNS systems GW170817 and GW190425 are fully consistent with quasi-circular inspirals, while GW200105 stands out among the NSBH binaries as the only system exhibiting significant residual eccentricity at 20 Hz, strengthening evidence for dynamically driven formation pathways. The remaining NSBH events show no measurable eccentricity and appear broadly compatible with low-spin binaries formed through isolated stellar evolution. Using hierarchical Bayesian inference, we obtain the first joint constraints on the mass, spin, and eccentricity distributions of NSBH binaries. Our results also yield the first simultaneous constraints on spin precession and orbital eccentricity in NSBH mergers, while the inferred merger rates remain fully consistent with previous LVK measurements. Treating all NSBH systems as a single population yields results compatible with formation in hierarchical triples, whereas the quasi-circular population remains broadly consistent with isolated evolution. Our results highlight the emerging role of eccentricity as a key discriminator between formation channels. As the number of NSBH detections grows, joint constraints on masses, spins, and orbital eccentricity will enable increasingly sharp tests of dynamical versus isolated binary evolution, establishing NSBH systems as powerful probes of compact-object astrophysics.

Figures (6)

• Figure 1: Marginal posterior distributions for selected source parameters of the two BNS events. Shown are the source-frame component masses ($m_1^\mathrm{source}$ and $m_2^\mathrm{source}$), eccentricity at 20Hz ($e_{20\mathrm{Hz}}$), the effective inspiral spin ($\chi_\mathrm{eff}$), the effective precession spin ($\chi_p$), and the luminosity distance ($D_L$). In each violin plot, the upper half corresponds to results from the eccentric pyEFPE analysis, while the lower half shows posteriors from the quasi-circular IMRPhenomXP (solid) and IMRPhenomXPHM (dashed) analyses.
• Figure 2: Marginal posterior distributions for selected source parameters of the selected NSBH events. Shown are the source-frame component masses ($m_1^\mathrm{source}$ and $m_2^\mathrm{source}$), eccentricity at 20Hz ($e_{20\mathrm{Hz}}$), the effective inspiral spin ($\chi_\mathrm{eff}$), the effective precession spin ($\chi_p$), and the luminosity distance ($D_L$). In each violin plot, the upper half corresponds to results from the eccentric pyEFPE analysis, while the lower half shows posteriors from the quasi-circular IMRPhenomXP (solid) and IMRPhenomXPHM (dashed) analyses.
• Figure 3: The median and 90% credible intervals of the population predictive distributions for the black hole mass, the mass ratio, the black hole spin, and the neutron star mass. The blue curves analyses all NSBH events using pyEFPE, the red curve only the quasi-circular NSBH events, and the grey curve all NSBH events but using IMRPhenomXPHM (i.e. neglecting eccentricity).
• Figure 4: Population predictive distributions for the truncated Normal population model for eccentricity. The top panel shows the cosine of the black hole spin tilt angle $\cos \theta_{\rm BH}$ and the lower panel the eccentricity at $20$Hz.
• Figure 5: Select hyperposteriors for the truncated Normal mixture model, representing an eccentric (dynamical) channel and a quasi-circular (isolated) channel. The upper panel shows the population mean eccentricity of the dynamical component $\mu_{\rm dyn}$, exhibiting a mild peak near the value inferred for GW200105 Morras:2025xfu, denoted by the vertical dashed line. The lower panel displays the mixing fraction $\lambda_{\rm qc}$ between the two channels. This parameter is only weakly constrained, but does disfavor a purely dynamical population, consistent with event-level parameter estimation.