Marcus Haberland, Alessandra Buonanno

Gravitational-wave inference for long signals, like those from binary neutron-star (BNS) systems, requires waveform models that are both physically faithful and computationally efficient, otherwise, one risks drawing incorrect conclusions about nuclear matter from observations. To address this challenge, we present a frequency-domain implementation of the accurate SEOBNRv5THM waveform model for quasi-circular, spin-aligned BNS systems within the effective-one-body framework. Our approach combines the stationary-phase approximation (SPA) for the early inspiral with a fast Fourier transform treatment of the late- and post-inspiral regime, applied mode-by-mode. Our hybrid approach retains the efficiency of the SPA without affecting the waveform accuracy close to merger, where matter effects are most significant. The resulting waveform's generation speed can be further decreased using modern parameter-estimation techniques, such as multibanding and relative binning. We demonstrate excellent agreement with the baseline SEOBNRv5THM model in both mismatches and when analyzing real and synthetic data, and show how waveform systematics could affect BNS detections in upcoming observational runs and new facilities on the ground. We find that our method significantly reduces computational costs, enabling faithful parameter estimation for BNS signals within practical runtimes of order days. Our procedure can be readily extended to coalescing binary black hole systems.