Generating Higher Order Modes from Binary Black Hole Mergers with Machine Learning

We introduce a machine learning model designed to rapidly and accuratelypredict the time domain gravitational wave emission of non-precessing binaryblack hole coalescences, incorporating the effects of higher order modes of themultipole expansion of the waveform. Expanding on our prior work, we decomposeeach mode by amplitude and phase and reduce dimensionality using principalcomponent analysis. An ensemble of artificial neural networks is trained tolearn the relationship between orbital parameters and the low-dimensionalrepresentation of each mode. We train our model on ∼ 10^5 signals withmass ratio q ∈ [1,10] and dimensionless spins χ_i ∈ [-0.9, 0.9],generated with the state-of-the-art approximant SEOBNRv4HM. We find that itachieves a median faithfulness of 10^-4 averaged across the parameterspace. We show that our model generates a single waveform two orders ofmagnitude faster than the training model, with the speed up increasing whenwaveforms are generated in batches. This framework is entirely general and canbe applied to any other time domain approximant capable of generating waveformsfrom aligned spin circular binaries, possibly incorporating higher order modes.