# Extracting Progenitor Parameters of Rotating CCSNe via Pattern Recognition and Machine Learning

Document #:
LIGO-P1400188-v2
Document type:
P - Publications
Other Versions:
Abstract:
Core-collapse supernovae (CCSNe) are among the most energetic events in the universe, releasing up to 10^{53} erg = 100 B of gravitational potential energy. Based on theoretical predictions, they are also expected to emit bursts of gravitational waves (GWs) that will be detectable by second-generation laser interferometer GW observatories such as Advanced LIGO (aLIGO), Advanced Virgo, and KAGRA. In a novel pattern-recognition approach, we investigate the inference of progenitor parameters from numerical GW signals produced by state-of-the-art rotating core-collapse simulations. After associating physical processes with characteristic spectrogram features, we develop several machine-learning (ML) algorithms that can accurately (often within average ±20% relative error) and precisely determine progenitor parameters from optimally-oriented CCSN signals located 5 kpc away from Earth. In particular, our ±2σ prediction intervals for β_{ic,b} are ~ 0.03 wide on average at 5 kpc. At a source distance of 10 kpc, we still achieve average relative errors within ±20% for our mean predictions, though our ±2σ prediction intervals for β_{ic,b} become ~ 0.05 wide. In addition to our hand-picked physical'' feature vector (FV) approach, we also investigate FV constructions with principal component analysis (PCA) and the scale-invariant feature transform (SIFT). In the future, our analysis could be implemented in the aLIGO data analysis pipelines to help determine the inner core dynamics of the next galactic CCSN; this information would otherwise be inaccessible via electromagnetic radiation.
Files in Document:
Topics:
Author Groups:
Notes and Changes:
Made some corrections before submitting a final version to SFP.
Related Documents:
Referenced by:

DCC Version 3.4.1, contact Document Database Administrators