High-accuracy inference of binary black hole source properties using numerical relativity surrogate models for gravitational waveform

Characterization of gravitational wave (GW) signals from the merger of binary black holes (BBHs) heavily rely on accurate gravitational waveform models. The most accurate way to compute waveforms is by solving the Einstein equation directly. This can only be done numerically (e.g., numerical relativity) — which is computationally expensive, taking weeks to months for a single simulation. Such a huge computational cost has paved the way for the development of approximate models that are now widely used in data analysis studies. However, these models can compromise accuracy in favor of speed. Recently, data-driven surrogate models have been shown to be capable of generating waveforms that are almost indistinguishable from NR in a fraction of a second. In this talk, I will summarize state-of-the-art NR surrogate models and discuss the application of these models to high-accuracy inference. We find that for many GW events, NR surrogates provide both more accurate and better-constrained parameter estimates in many of the key parameters including sky localization, mass ratio, and spin quantities. Furthermore, using numerical relativity surrogate models, we are able to constrain the properties of the remnant black hole and kick velocity. This information will help understand the environment and evolutionary pathways of black holes.