Common Horizon Formation in Simulations of Binary Black Hole Mergers

Gravitational wave astronomy is an emergent field, offering a relatively less explored method for studying the universe. One source of gravitational waves being observed today are merging black hole binaries. Intrinsic parameters of binary systems are derived from observed gravitational waveforms through a comparison against models of general relativity waveforms. Numerical relativity simulations are crucial to the construction and calibration of such waveform models. To simulate a binary black hole merger and calibrate waveform models, an accurate estimate of the time at which black holes merge is required. Researchers currently use a post-Newtonian framework to estimate the merger time for a potential simulation with a desired set of initial parameters. Within the theoretical framework of quasi-local horizons, I present data-driven alternatives for predicting the merger times of simulations as the time at which a common horizon forms between merging black holes. For my first model, I begin with predictions for the time of common horizon formation from a simple gravitational wave model and use multiple iterations of symbolic regression to correct that model for generic astrophysical binary black hole systems. For my second model, I train a simple neural network on simulation metadata to predict the time of common horizon formation. Both models perform on par with the standard post-Newtonian approach used by researchers today, demonstrating the promising predictive power of data-driven models.