Detecting Gravitational Waves from Core-Collapse Supernovae using Deep Learning

All three classes of compact binary mergers - binary black hole (BBH), binary neutron star (BNS) and neutron star-black hole (NSBH) - have now been detected by the Laser Interferometer Gravitational-Wave Observatory (LIGO) and Virgo Collaborations. However, one primary source of gravitational waves (GWs) has yet to be discovered - Core-Collapse Supernovae (CCSNe). CCSNe are expected to emit stochastic GW signals and thus require the utilization of alternative detection methods to the standard matched filtering approach. We show that our convolutional neural network (CNN) can distinguish and classify CCSNe signals buried in periods of both stationary and non-stationary real detector noise using training sets constructed from publicly available three-dimensional (3D) CCSNe simulations. Next, we establish three additional gravitational wave search windows (GSWs) for CCSN Type II candidates that had exploded in past advanced observing runs (O1, O2 and O3) of LIGO. We present classification results and detection limits for a total of five CCSN candidates within their respective GSWs for a range of 3D CCSNe waveform models.