学术文献库

We present a surrogate model of \texttt{SEOBNRv4PHM}, a fully precessing time-domain effective-one-body waveform model including subdominant modes. We follow an approach similar to that used to build recent numerical relativity surrogate models. Our surrogate is 5000M in duration, covers mass-ratios up to 1:20 and dimensionless spin magnitudes up to 0.8. Validating the surrogate against an independent test set we find that the bulk of the surrogate errors is less than $\sim 1\%$ in mismatch, which is similar to the modelling error of \texttt{SEOBNRv4PHM} itself. At high total mass a few percent of configurations can exceed this threshold if they are highly precessing and they exceed a mass-ratio of 1:4. This surrogate is nearly two orders of magnitude faster than the underlying time-domain \texttt{SEOBNRv4PHM} model and can be evaluated in $\sim 50$ ms. Bayesian inference analyses with \texttt{SEOBNRv4PHM} are typically very computationally demanding and can take from weeks to months to complete. The two order of magnitude speedup attained by our surrogate model enables practical parameter estimation analyses with this waveform family. This is \emph{crucial} because Bayesian inference allows us to recover the masses and spins of binary black hole mergers given a model of the emitted gravitational waveform along with a description of the noise.