A New Method to Classify Type IIP/IIL Supernovae Based on their Spectra

TL;DR

This paper introduces a spectral classification method for Type IIP and IIL supernovae using principal component analysis and machine learning, achieving high accuracy with spectral features like Halpha.

Contribution

It presents a novel spectral classification approach combining FPCA, SVM, and ANN, with a focus on the Halpha line for distinguishing supernova types.

Findings

F1-Score of 0.881 for the classifier.

Halpha line alone can reach an F1-Score of 0.849.

Spectral profile of Halpha is key to classification.

Abstract

Type IIP and type IIL supernovae (SNe) are defined on their light curves, but the spectrum criteria in distinguishing these two type SNe remains unclear. We propose a new classification method. Firstly, we subtract the principal components of different wavelength bands in the spectra based on Functional Principal Components Analysis (FPCA) method. Then, we use Support Vector Machine (SVM) and Artificial Neural Network (ANN) to classify these two types of SNe. The best F1-Score of our classifier is 0.881, and we found that solely using Halpha line at 6150-6800 A for classification can reach a F1-Score up to 0.849. Our result indicates that the profile of the Halpha is the key to distinguish the two type SNe.