Speaker
Description
Core-collapse supernova (CCSN) events remain one of the most compelling yet still uncovered sources of gravitational wave (GW) signals. Their detection would open a direct window into the physics unfolding deep within a collapsing stellar core. In this work, we evaluate deep-learning (DL) approaches for analyzing GW signals from rotating CCSN, addressing both the classification of these signals in detector noise and parameter estimation of three key bounce-signal parameters: the maximum (peak) frequency, $(f_\textrm{peak})$, the GW strain amplitude ($\Delta h$) scaled by luminosity distance ($D$), $(D \cdot \Delta h)$, and the rotation of the core. Training data are constructed from a set of numerically simulated CCSN waveforms, which we inject into real detector noise drawn from the second part of the fourth observing run (O4b) of the Advanced LIGO and Advanced Virgo instruments.
Our classification network reliably distinguishes Galactic CCSN signals from noise once the signal-to-noise ratio (SNR) of the network exceeds $8$, achieving an overall accuracy of $0.99$. For parameter estimation, we find that all three parameters achieve an $R^2$ above $0.85$, with the rotation rate achieving the best results ($R^2 =0.91$). These results demonstrate that DL techniques offer a promising and computationally efficient approach for extracting physical information from a future Galactic CCSN GW detection.
| Which topic best fits your talk? | Astrophysics and Cosmology |
|---|---|
| Affiliations | Centro de Física das Universidades do Minho e do Porto, Universidade de Valência |