Parameter estimation for the GWTC-4.0 catalog with phenomenological waveform models that include orbital eccentricity and an updated description of spin precession

TL;DR

The study extends GWTC-4.0 parameter estimation to include orbital eccentricity and improved spin precession using three IMRPhenom waveform families (XPNR, TPHM, TEHM). It employs Bayesian inference with Bilby across 84 BBH events, contrasts quasi-circular and eccentric/precessing hypotheses via Bayes factors, and analyzes waveform-systematic differences through several diagnostics. The results show broad agreement across models for most events, with seven eccentric candidates and a subset of high-mass/high-spin events exhibiting model-driven discrepancies, often tied to mode content or data-quality issues. The work highlights the need for joint precessing–eccentric waveform models with extended NR calibration to unambiguously identify dynamical formation channels in future observing runs, and provides a publicly available extended catalog and automation framework for reproducible large-scale analyses.

Abstract

The GWTC-4.0 catalog of transient gravitational wave signals describes observations made in the first part of the fourth observing run of the LIGO-Virgo-KAGRA (LVK) gravitational wave detector network. Here we extend the LVK's GWTC-4.0 analysis to elliptic orbits, and an improved description of spin precession in the frequency domain. For this study we use state-of-the-art waveforms from the IMRPhenom family (specifically XPNR, TPHM, and TEHM), and we consider the 84 confidently detected events that are consistent with binary-black-hole mergers. We present an extended catalog of updated posterior samples, quantify how incorporation of these waveform effects alters inferred source properties relative to previous analyses, and discuss waveform systematics.

Figures (13)

• Figure 1: Scalar diagnostics used to quantify differences between posterior distributions from the GWTC-4.0 analyses and our new analyses with IMRPhenomXPNR and IMRPhenomTPHM. Each panel shows, from top to bottom, the Jensen-Shannon divergence, the normalized median shift $\Delta$, and the complement of the credible-interval overlap fraction $1-f_{\rm overlap}$, evaluated for the source-frame total mass $M$, mass ratio $q$, and effective spin parameters $\chi_{\rm eff}$ and $\chi_{\rm p}$. Horizontal dashed lines mark the adopted empirical thresholds ($\mathrm{JS}\!\ge\!0.05$, $\Delta\!\ge\!0.3$, $1-f_{\rm overlap}\!\ge\!0.2$), above which waveform-model differences are considered significant. Points are colored by parameter, and events exceeding any threshold are highlighted in red along the $x$-axis labels. The metrics collectively indicate that only a small subset of events exhibit measurable waveform-model systematics.
• Figure 2: Dependence of waveform-model systematics on source parameters. Each point corresponds to a GWTC-4.0 event, positioned according to the median posterior values of the redshifted total mass $M$, the combined spin magnitude $\sqrt{\chi_{\rm eff}^{2} + \chi_{\rm p}^{2}}$, and the network matched-filter signal-to-noise ratio $\rho_{\rm mf}^{N}$ obtained from the reference GWTC analysis (using the combined samples). Points are colored by the combined JS divergence (maximized over all model pairs and parameters), shown in logarithmic scale. Events with the largest JS values are annotated. The figure shows that only a small subset of high-mass, high-spin, or high-SNR systems exhibit significant waveform-model differences, while most of the population occupies the region where posteriors are consistent across models.
• Figure 3: Comparison of posterior distributions for the seven events flagged as potentially affected by waveform–model systematics. Each column corresponds to one intrinsic parameter ($M$, $q$, $\chi_{\rm eff}$, and $\chi_{\rm p}$), and each row to a different event. Posteriors are shown for IMRPhenomXPNR, IMRPhenomTPHM, and the models used in GWTC–4. The majority of these events only show mild discrepancies in the inferred parameters, while the most massive ones (GW231028_153006 and GW231123_135430) exhibit visible larger discrepancies.
• Figure 4: Inferred luminosity distance and source inclination for GW231123_135430 using different waveform models. The discrepant results of IMRPhenomXPHM_SpinTaylor are not reproduced by IMRPhenomXPNR nor IMRPhenomTPHM, which show improved consistency in these parameters with the other models.
• Figure 5: Posterior comparison for GW230814_230901 using different waveform models. The differences between IMRPhenomXPNR and IMRPhenomTPHM arise from the absence of the $(3,2)$ mode in IMRPhenomTPHM. When the $(3,2)$ mode is removed from IMRPhenomXPNR, the resulting posteriors match those of IMRPhenomTPHM, confirming the mode–content origin of the discrepancy.