Modeling matter(s) in SEOBNRv5THM: Generating fast and accurate effective-one-body waveforms for spin-aligned binary neutron stars

TL;DR

SEOBNRv5THM delivers an accurate and fast gravitational-wave model for quasi-circular, spin-aligned binary neutron stars by augmenting the SEOBNRv5HM BBH baseline with matter effects, including spin-induced multipoles up to $l=4$, adiabatic and dynamical tides, and higher GW modes. It introduces a NR-informed pre-merger phenomenology for the dominant mode and calibrates the merger time against a large NR NR dataset, achieving speedups of $100$ to $1000\times$ over the previous BNS model for $M\ge 2\,M_\odot$ and enabling practical Bayesian parameter estimation, as demonstrated on GW170817 and GW190425. Validation against SACRA and BAM NR waveforms shows phase accuracy within NR uncertainties and competitive mismatches relative to TEOBResumS and NRTidalv3, with notable performance in non-spinning and moderately spinning regimes. The model also demonstrates feasible parameter estimation costs, suggesting it can be deployed for upcoming observing runs while remaining extensible to future additions such as spin precession and eccentricity. Overall, SEOBNRv5THM stands as a robust, scalable foundation for matter-including SEOBNR waveform models in current and next-generation GW observations.

Abstract

We present SEOBNRv5THM, an accurate and fast gravitational-waveform model for quasi-circular, spinning, non-precessing binary neutron stars (BNS) within the effective-one-body (EOB) formalism. It builds on the binary-black-hole approximant SEOBNRv5HM and, compared to its predecessor SEOBNRv4T, it i) incorporates recent high-order post-Newtonian results in the inspiral, including higher-order adiabatic tidal contributions, spin-induced multipoles and dynamical tides for spin-aligned neutron stars, ii) includes the gravitational modes $(\ell, |m|)=(2,2),(3,3),(2,1),(4,4),(5,5),(3,2),$ and $(4,3)$, iii) has a time of merger calibrated to BNS numerical-relativity (NR) simulations, iv) accurately models the pre-merger $(2,2)$ mode through a novel phenomenological ansatz, and v) is 100 to 1000 times faster than its predecessor model for BNS systems with total mass $M \geq 2\, M_\odot$. Thus, SEOBNRv5THM can be used in Bayesian parameter estimation, which we perform for two BNS events observed by the LIGO-Virgo Collaboration, GW170817 and GW190425. The model accurately reproduces BAM and SACRA NR waveforms with errors comparable to or lower than the intrinsic NR uncertainty. We validate the model against the other state-of-the-art BNS waveform models NRTidalv3 and TEOBResumS and find differences only for highly spinning and highly tidally deformable BNS, where there is no NR coverage and the models employ different spin prescriptions. Our model serves as a foundation for the development of subsequent SEOBNR waveform models with matter that incorporate further effects, such as spin-precession and eccentricity, to be employed for upcoming observing runs of the LIGO-Virgo-KAGRA Collaboration and future facilities on the ground.

Figures (12)

• Figure 1: The frequency-time (top) and amplitude-time evolution (bottom) of different pre-merger prescriptions compared to an exemplary SACRA simulation with EoS 15H for a $(1.25 + 1.25)\ M_{\odot}$ BNS. The NR waveform is shown as a solid, black line and the SEOBNRv5THM factorized waveform from Eq. \ref{['eq:hlmFactorized']} without application of any pre-merger model as a blue, solid line. The application of BNS-tuned NQCs on top of the factorized waveform is shown as a purple, dashed line and the phenomenological pre-merger and tapered model we use in SEOBNRv5THM as outlined in this section as a dot-dashed, red line. We further show TEOBResumS as a dotted, green line and indicate the time of merger of the NR simulation $t_{\rm peak}={t_\mathrm{match}}$ with a vertical line.
• Figure 2: Benchmarks for SEOBNRv5THM, our baseline SEOBNRv5HM and the predecessor approximant SEOBNRv4T for a BNS system $(q=1.5,\,\Lambda_1=700,\,\Lambda_2=800,\,\chi_1=0.1,\,\chi_2=0.2)$ starting from $20$ Hz where we only compute the $(2,2)$ mode. For the SEOBNRv5 models, we also show the benchmarks if the PA approximation is applied as solid lines. We call both tidal approximants at a sampling rate of $f_{\rm sample}=8192$ Hz and note that for low total-mass systems, SEOBNRv5HM, which is run with $\Lambda_1=\Lambda_2=0$, needs a higher sampling rate to resolve the ringdown, which is why SEOBNRv5THM can become faster than it's baseline. These benchmarks were run on an Apple M2 processor.
• Figure 3: The resulting $\Delta t_{22}$ for the equal-mass NR simulations prior to and after the extrapolation to a common resolution. We show the recovered $\Delta t_{22}$ that matches the merger of the NR simulation to $t_\mathrm{attach}$ from Eq. \ref{['eq:t_attach']} for all used equal-mass simulations, given their respective quadrupolar tidal coupling constant $\kappa_2^T$. We indicate with black dots the employed SACRA simulations and with stars the BAM simulations. The originally recovered $\Delta t_{22}$ values for the highest-resolution BAM simulation are shown in orange, and the $\Delta t_{22}$ extrapolated to SACRA resolution in black. We also indicate the resulting fit from Eq. \ref{['eq:t_fit']} as a blue line.
• Figure 4: Extrapolation of the merger time for some BAM simulations of different resolutions $1/n_{\rm grid}$. We show the difference between the time of merger $t_{\rm peak}$ for runs at different resolutions and the time of merger of the respective highest resolution run $t_{\rm peak}^{\text{(high-res)}}$ of a given system as stars of the same color. We also include the respective quadratic fit of the data as a solid line. The resolution employed by SACRA simulations is indicated by a black vertical line, from where we extract $\Delta t_{\rm peak}^{\rm (extrap)}$ as the value of the quadratic fit at the SACRA resolution, see Eq. \ref{['eq:tpeakNR']}.
• Figure 5: $\Delta t_{22}$ fit across the parameter space of the employed simulations. In the background of the image we show the result of our fit, while the data points show the respective mass-ratio $q$ and quadrupolar tidal coupling constant $\kappa_2^T$ of the employed non-spinning simulations. We indicate SACRA simulations as circles and BAM simulations as stars. Points with a black (white) interior have been used for fitting (validation).