Improved numerical relativity surrogates using domain decomposition

Gravitational inspiral-merger-ringdown (IMR) waveform modeling of binary black hole systems (BBHs) is key to our ability to detect gravitational waves and to extract physics from these detections. Numerical relativity surrogate models are powerful IMR waveform models that leverage the accuracy of numerical relativity while extending parameter space coverage. Robust, accurate surrogate models, such as NRSur7dq4, cover a subspace of the precessing quasi-circular parameter space. However, there is evidence to suggest that, while NRSur7dq4 is accurate over the time domain as a whole, it may be possible to further increase accuracy in the ringdown, which is important for performing various tests of GR and consistency checks on the properties of the remnant black hole. We present a domain-decomposed methodology for surrogate construction that allows for more targeted modeling choices on the individual subdomains while preserving waveform smoothness. We employ this methodology to construct NRSur7dq4v2, an inspiral-ringdown domain-decomposed surrogate that improves upon NRSur7dq4 in ringdown accuracy. Specifically, NRSur7dq4v2 is three times as accurate as its predecessor in extracting remnant black hole parameters from its ringdown signal. We conclude by discussing the performance of NRSur7dq4v2 under other such tests, as well as the applications of our improved waveform model to problems such as parameter estimation of mergers of heavy binary black holes.