Incorporating waveform calibration error in gravitational-wave modeling and inference for SEOBNRv4

Abstract

As gravitational wave (GW) detector networks continue to improve in sensitivity, the demand on the accuracy of waveform models which predict the GW signals from compact binary coalescences is becoming more stringent. At high signal-to-noise ratios (SNRs) discrepancies between waveform models and the true solutions of Einstein's equations can introduce significant systematic biases in parameter estimation (PE). These biases affect the inferred astrophysical properties, including matter effects, and can also lead to erroneous claims of deviations from general relativity, impacting the interpretation of astrophysical populations and cosmological parameters. While efforts to address these biases have focused on developing more precise models, we explore an alternative strategy to account for uncertainties in waveform models, particularly from calibrating an effective-one-body (EOB) model against numerical relativity (NR) data. We introduce an efficient method for modeling and marginalizing over waveform uncertainty in the SEOBNRv4 model, which captures the dominant $(2,2)$ mode for non-precessing quasi-circular binary black holes (BBHs). Our approach uses Gaussian process regression (GPR) to model amplitude and phase deviations in the Fourier domain. This method mitigates systematic biases in PE and increases posterior variance by incorporating a broader distribution of waveforms, consistent with previous findings. This study emphasizes the importance of incorporating waveform uncertainties in GW data analysis and presents a novel, practical framework to include these uncertainties in Bayesian PE for EOB models, with broad applicability.