Neural Surrogate Model for Precessing Gravitational Waveforms from Binary Black Hole Mergers

Numerical relativity simulations are the most precise method for generating gravitational waveforms, but their extensive runtime prohibits their direct application in gravitational wave inference. Surrogate models have been developed to accelerate waveform generation without compromising accuracy. Studies have explored the use of artificial neural networks for their function modelling capabilities and computational efficiency to develop surrogate models for gravitational waveforms. Most prior research has focused on models from spin-aligned binary systems or on precessing binary systems where one black hole is arbitrarily spinning while the other is non-spinning, reducing the problem to four dimensions. In this study, we extend these efforts to model waveforms produced by generically precessing binaries over a broader parameter space than previously examined, while preserving the seven-dimensional nature of the system. We train our neural networks with data generated using NRSur7dq4. We also perform a comparative analysis to evaluate the speed, computational cost, and accuracy of our developed model against NRSur7dq4.