Identifying Correlations in Precessing Gravitational-Wave Signals with Machine Learning

Binary black hole (BBH) spins provide unique insights into the formation environments, evolutionary history, and dynamics of these objects. For highly massive, precessing BBH systems with merger-dominated signals, spin parameters, traditionally derived from the inspiral phase, cannot be characterized analytically and are poorly understood. The degeneracies in waveforms, where dissimilar parameters yield similar waveforms, complicate source identification. Using a neural network trained on a numerical relativity surrogate waveform model including higher harmonics and precession, we study waveform degeneracies between BBH systems with different intrinsic parameters. Based on mismatch measurements generated by the network, we present a methodology of analyzing phenomenologically how spin degrees of freedom impact the resulting waveform for highly massive, precessing BBHs and report the recovered correlations.