Impact of Numerical Relativity information on effective-one-body waveform models

Abstract

We present a comprehensive comparison of the spin-aligned effective-one-body (EOB) waveform model of Nagar et al. [Phys. Rev. D93, 044046 (2016)], informed using 40 numerical-relativity (NR) datasets, against a set of 149, $\ell=m=2$, NR waveforms freely available through the Simulation Extreme Spacetime (SXS) catalog. We find that, without further calibration, these EOBNR waveforms have unfaithfulness (at design Advanced-LIGO sensitivity and evaluated with total mass $M$ varying as $10M_\odot\leq M \leq 200M_\odot$) always below $1\%$ against all NR waveforms except for three outliers, that still never exceed the $3\%$ level; with a minimal retuning of the (effective) next-to-next-to-next-to-leading-order spin-orbit coupling parameter for the non-equal-mass and non-equal-spin sector, that only needs three more NR waveforms, one is left with another two (though different) outliers, with maximal unfaithfulness of up to only $2\%$ for a total mass of $200M_\odot$. We show this is the effect of slight inaccuracies in the phenomenological description of the postmerger waveform of Del Pozzo and Nagar [arXiv:1606.03952] that was constructed by interpolating over only 40NR simulations. We argue that this is easily fixed by using either an alternative ringdown description (e.g., the superposition of quasi-normal-modes) or an improved version of the phenomenological representation. By analyzing a NR waveform with mass ratio $8$ and dimensionless spins $+0.85$ obtained with the BAM code, we conclude that the model would benefit from NR simulations specifically targeted at improving the postmerger-ringdown phenomenological fits for mass ratios $\gtrsim 8$ and spins $\gtrsim 0.8$.