The Cost of Circularity: Quantifying Eccentricity-Induced Biases in Binary Black Hole Inference

Abstract

Dynamically assembled binary black holes are expected to retain measurable orbital eccentricity in the LIGO-Virgo-KAGRA band, but most parameter estimation analyses still assume quasi-circular inspirals. This raises a critical question: how strongly does unmodeled eccentricity bias the inferred properties of BBH mergers? We address this by injecting eccentric signals generated with TEOBResumS-Dali and recovering them using the circular, precessing IMRPhenomXPHM waveform model. Across $20$-$80 \, M_\odot$ and eccentricities up to $e=0.5$, we find that circular waveform models remain reliable only for very small eccentricities. Above $e\sim0.2$ at 10 Hz, recovered masses, spins, inclination, and distances begin to show significant systematic offsets. Circular precessing templates mimic eccentric amplitude and phase modulations by introducing artificial precession, highlighting a major degeneracy between these effects. For high-mass, moderately eccentric mergers, circular models misestimate parameters at a level that would bias astrophysical interpretation and population studies. Our results establish the parameter-space boundaries where eccentric waveform models become essential for accurate inference in current and next-generation detectors.

Figures (11)

• Figure 1: Posterior distributions of detector-frame component masses $m_1$ versus $m_2$ from parameter estimation using a precessing-spin waveform model, for injections with varying initial eccentricities. The injected signals are for eccentric, non-spinning systems, while the recovery uses non-eccentric waveforms with non-aligned component spins, that lead to spin-precession. Each subplot corresponds to an injected eccentricity, ordered increasingly. Star markers indicate the injected $m_1$ and $m_2$ values. The 90%, 50%, and 10% credible regions are shown in the plots, and the injected detector-frame chirp mass $\mathcal{M}$ is listed in the legend.
• Figure 2: Posterior distributions of detector-frame component masses $m_1$ versus $m_2$ from parameter estimation using a non-spinning waveform model, for injections with varying initial eccentricities. The injected signals are eccentric and non-spinning, while the recovery uses non-eccentric waveforms with no component spins. Each subplot corresponds to an injected eccentricity, ordered increasingly. Star markers indicate the injected $m_1$ and $m_2$ values. The 90%, 50%, and 10% credible regions are shown, and the injected detector-frame chirp mass $\mathcal{M}$ is listed in the legend.
• Figure 3: Posterior distributions of the luminosity distance ($D_{\mathrm{L}}$) versus chirp mass ($\mathcal{M}$) from parameters estimation using a precessing-spin waveform model, for injections with varying initial eccentricities. The star markers indicate the injected chirp mass values ($\mathcal{M}$). The contours represent the 90%, 50%, and 10% credible intervals obtained from parameter estimation.
• Figure 4: Absolute differences in chirp mass, $|\Delta \mathcal{M}|$, as a function of the injected chirp mass, $\mathcal{M}$, for all injections. Here, $\Delta \mathcal{M}$ is defined as the difference between the injected chirp mass and the median of the recovered posterior distribution. The top, middle, and bottom panels correspond to the precessing-spin, aligned spin, and non-spinning configurations respectively. Different marker styles indicate the injected eccentricity. Larger differences are seen for higher eccentricity, while for a given eccentricity the differences increase with increasing injected chirp mass.
• Figure 5: Posterior distributions of the luminosity distance ($D_{\mathrm{L}}$) versus chirp mass ($\mathcal{M}$) from parameter estimation using a non-spinning waveform model, for injections with varying initial eccentricity. The star markers indicate the injected chirp mass values ($\mathcal{M}$). The contours represent the 90%, 50%, and 10% credible intervals obtained from parameter estimation.