Measuring Eccentricity and Addressing Waveform Systematics in GW231123

TL;DR

This paper reanalyzes the GW231123 gravitational-wave event using a comprehensive model that includes spin precession and eccentricity, finding no strong evidence for eccentricity and highlighting waveform model discrepancies affecting parameter estimates.

Contribution

It demonstrates that waveform model disagreements can cause parameter biases and clarifies that GW231123 shows no significant eccentricity when using advanced models.

Findings

GW231123 shows no strong evidence for eccentricity.

Discrepancies in waveform models cause parameter estimation biases.

Bayesian analysis favors a zero-eccentricity, spin-precessing model.

Abstract

The gravitational-wave event GW231123_135430 is the heaviest binary black hole system observed by the LIGO--Virgo--KAGRA Collaboration to date, with the initial analysis indicating the individual black hole masses lie within or above the theorized pair-instability mass gap of roughly $60$--$130\,M_\odot$. The inference further suggests that both black holes possess high spins, measured to be $0.90^{+0.10}_{-0.19}$ and $0.80^{+0.20}_{-0.51}$. Therefore, the observation of this event suggests the formation of black holes from channels beyond the standard stellar collapse. However, different waveform models yield significantly different parameter estimates, possibly due to missing physics in the models used in inference. In this work, we carry out a reanalysis of GW231123 using a physically complete model, accounting for both spin precession and eccentricity. Our analysis shows that this event does not exhibit strong evidence for eccentricity and the exclusion of eccentricity has minimal impact on inference. Furthermore, for GW231123-like systems, even eccentricities as large as $0.15$ at $10$ Hz do not yield a confident nonzero eccentricity measurement. Through a zero-noise injection recovery study, we show that the observed discrepancies in the parameter estimates can be explained by disagreement in the waveform models at strong spin precession, with the degree of parameter bias in the zero-noise runs being comparable to that observed for the real signal. We also show that inference performed with an eccentric, aligned-spin waveform model can yield a confident nonzero eccentricity measurement due to the degeneracy between eccentricity and spin precession. Bayesian model selection, however, rules out this interpretation in favor of the eccentric, spin precessing hypothesis, which supports zero eccentricity -- a conclusion we confirm with additional zero-noise injection-recovery tests.