Analysis of GWTC-3 with fully precessing numerical relativity surrogate models

Abstract

The third Gravitational-Wave Transient Catalog (GWTC-3) contains 90 binary coalescence candidates detected by the LIGO-Virgo-KAGRA Collaboration (LVK). We provide a re-analysis of binary black hole (BBH) events using a recently developed numerical relativity (NR) waveform surrogate model, NRSur7dq4, that includes all $\ell \leq 4$ spin-weighted spherical harmonic modes as well as the complete physical effects of precession. Properties of the remnant black holes' (BH's) mass, spin vector, and kick vector are found using an associated remnant surrogate model NRSur7dq4Remnant. Both NRSur7dq4 and NRSur7dq4Remnant models have errors comparable to numerical relativity simulations and allow for high-accuracy parameter estimates. We restrict our analysis to 47 BBH events that fall within the regime of validity of NRSur7dq4 (mass ratios greater than 1/6 and total masses greater than $60 M_{\odot}$). While for most of these events our results match the LVK analyses that were obtained using the semi-analytical models such as IMRPhenomXPHM and SEOBNRv4PHM, we find that for more than 20\% of events the NRSur7dq4 model recovers noticeably different measurements of black hole properties like the masses and spins, as well as extrinsic properties like the binary inclination and distance. For instance, GW150914_095045 exhibits noticeable differences in spin precession and spin magnitude measurements. Other notable findings include one event (GW191109_010717) that constrains the effective spin $χ_{eff}$ to be negative at a 99.3\% credible level and two events (GW191109_010717 and GW200129_065458) with well-constrained kick velocities. Furthermore, compared to the models used in the LVK analyses, NRSur7dq4 recovers a larger signal-to-noise ratio and/or Bayes factors for several events.

Figures (19)

• Figure 1: Posteriors for the source-frame total mass $M$, mass ratio $q$, source-frame component masses $m_1$, $m_2$, luminosity distance $D_{\rm L}$ and the cosine of the inclination angle $\theta_{\rm JN}$ for all 47 events analyzed with NRSur7dq4 model (blue). For comparison, we also show the public LVK posteriors LIGOScientific:2021usbLIGOScientific:2021djpGWTC2.1_PEGWTC3_PE obtained using IMRPhenomXPHM (orange) and SEOBNRv4PHM (green) models. For some events, the SEOBNRv4PHM posteriors are missing from the LVK release (see Tab. \ref{['Tab:special_events']}); the SEOBNRv4PHM results are, therefore, also absent in this and all following figures for that subset of events. The grey dashed line represents the mass ratio cut of $q=1/6$ used for NRSur7dq4. Posteriors are reported in the wave frame (see Sec. \ref{['subsec:frame_choice']}) at $f_{\rm ref}=20$ Hz. Further details are given in Sec. \ref{['sec:results']}.
• Figure 2: Posteriors for the spin magnitudes $\chi_1$ and $\chi_2$, spin angles $\theta_{1}$ and $\theta_{2}$, effective inspiral spin parameter $\chi_{\mathrm{eff}}$, and spin precession parameter $\chi_p$ for all 47 events analyzed with NRSur7dq4 model (blue). For comparison, we also show the public LVK posteriors LIGOScientific:2021usbLIGOScientific:2021djpGWTC2.1_PEGWTC3_PE obtained using IMRPhenomXPHM in orange and SEOBNRv4PHM (where available) in green. Posteriors are reported in the wave frame (see Sec. \ref{['subsec:frame_choice']}) at $f_{\rm ref}=20$ Hz. We provide 3D visualizations of the full spin posteriors at Ref. NRSurCatalog. Further details are given in Sec. \ref{['sec:results']}.
• Figure 3: Jensen-Shannon divergence (JSD) values between the one-dimensional marginalized posteriors of the source-frame total mass $M$, mass ratio $q$, source-frame component masses $m_1$, $m_2$, luminosity distance $D_{\rm L}$ and the inclination angle $\theta_{\rm JN}$ obtained using NRSur7dq4 and the public LVK posteriors LIGOScientific:2021usbLIGOScientific:2021djpGWTC2.1_PEGWTC3_PE obtained using IMRPhenomXPHM (blue circles) and SEOBNRv4PHM (green squares, where available). Dashed red lines correspond to a JS divergence of 0.02, indicating significant differences between these posteriors. Further details are discussed in Sec. \ref{['sec:results']}.
• Figure 4: Jensen-Shannon divergence (JSD) values between the one-dimensional marginalized posteriors of the spin magnitudes $\chi_1$ and $\chi_2$, spin angles $\theta_{1}$ and $\theta_{2}$, effective inspiral spin parameter $\chi_{\mathrm{eff}}$ and spin precession parameter $\chi_p$ obtained using NRSur7dq4 and the public LVK posterior samples LIGOScientific:2021usbLIGOScientific:2021djpGWTC2.1_PEGWTC3_PE obtained using IMRPhenomXPHM (blue circles) and SEOBNRv4PHM (green squares, where available). Dashed red lines correspond to a JS divergence of 0.02, indicating significant differences between these posteriors. Further details are discussed in Sec. \ref{['sec:results']}.
• Figure 5: Posteriors for the source-frame total mass $M$, mass ratio $q$, spin magnitudes $\chi_1$ and $\chi_2$, spin angles $\theta_{1}$ and $\theta_{2}$, effective inspiral spin parameter $\chi_{\mathrm{eff}}$, and spin precession parameter $\chi_p$ for a list of seven events for which we infer the most significant differences between results obtained using NRSur7dq4 (blue histogram), IMRPhenomXPHM (orange histogram) and SEOBNRv4PHM (green histogram, where available). The under-sampled posteriors for SEOBNRv4PHM is a consequence of an inefficient post-processing procedure used in the RIFT code whenever calibration uncertainties are accounted for Payne:2019wmyPayne:2020myg. Posteriors are reported in the wave frame (see Sec. \ref{['subsec:frame_choice']}) at $f_{\rm ref}=20$ Hz. Further details are given in Sec. \ref{['sec:special_events']}.