Waveform and remnant surrogate models for precessing binaries including early inspiral, higher mass ratios, and memory

Accurate gravitational wave predictions are necessary for extracting astrophysical source properties from binary black hole observations while predictions of the remnant properties are necessary for fundamental tests of general relativity and astrophysical population modeling. Numerical relativity (NR) surrogate models are the current most accurate waveform and remnant models for precessing binaries, but have only been trained on mass ratios less than 4 and the last ~20 orbits before merger. We construct a bank of ~1000 new NR simulations of generically precessing binaries with mass ratios up to 8. We use these simulations to train two data-driven surrogate models, one for the gravitational waveform and one for the mass, spin and kick velocity of the final black hole. These models capture the full physics of the simulations, including precession, higher-modes, and gravitational memory. Finally, we hybridize the waveform model in real-time using post-Newtonian waveforms to include the early inspiral, thereby lifting the main limitation of previous precessing surrogate models.