LIGO Core-Collapse Supernova Detection Using Convolutional Neural Networks.

Area of Science:

• Astronomy and Astrophysics
• Gravitational Wave Detection
• Computational Astrophysics

Background:

• Core-collapse supernovae (CCSNe) are key targets for gravitational wave (GW) astronomy.
• Understanding CCSNe explosion mechanisms requires sensitive GW detection methods.
• Current detection techniques face challenges with low signal-to-noise ratio (SNR) events.

Purpose of the Study:

• To investigate the efficacy of convolutional neural networks (CNNs) for enhancing CCSNe GW signal detection.
• To compare the performance of CNNs trained on different time-frequency representations.
• To improve the identification of faint CCSNe GW signals.

Main Methods:

• Simulated CCSNe GW signals and Advanced LIGO noise were used for training data.
• Two time-frequency analysis techniques, Short-Time Fourier Transform (STFT) and Q-transform (QT), were employed to generate spectrograms.
• Two independent CNNs were trained on STFT and QT spectrograms, respectively.

Main Results:

• Both CNNs achieved near 100% true positive rates for CCSNe GW events with SNR > 0.5.
• The CNN trained on STFT spectrograms outperformed the QT-based CNN for SNRs below 0.5.
• CNNs effectively detect CCSNe signals via their time-frequency representations.

Conclusions:

• CNNs offer a promising approach to enhance CCSNe GW detection.
• STFT-based CNNs demonstrate superior sensitivity for weaker GW signals.
• These findings contribute to advancing the search for and analysis of CCSNe GWs.