Reduction of supernova light curves by vector Gaussian processes

TL;DR

This paper introduces vector Gaussian processes to accurately approximate and reconstruct supernova bolometric light curves from multicolor photometric data, considering correlations across different passbands.

Contribution

The study presents a novel application of vector Gaussian processes for supernova light curve reduction, accommodating inhomogeneous data and inter-band correlations, with implementation in an accessible Python library.

Findings

Successfully applied to 29 superluminous supernovae

Enabled reconstruction of bolometric light curves under black-body assumptions

Provides an efficient, open-source tool for astrophysical light curve analysis

Abstract

Bolometric light curves play an important role in understanding the underlying physics of various astrophysical phenomena, as they allow for a comprehensive modeling of the event and enable comparison between different objects. However, constructing these curves often requires the approximation and extrapolation from multicolor photometric observations. In this study, we introduce vector Gaussian processes as a new method for reduction of supernova light curves. This method enables us to approximate vector functions, even with inhomogeneous time-series data, while considering the correlation between light curves in different passbands. We applied this methodology to a sample of 29 superluminous supernovae (SLSNe) assembled using the Open Supernova Catalog. Their multicolor light curves were approximated using vector Gaussian processes. Subsequently, under the black-body assumption for the SLSN spectra at each moment of time, we reconstructed the bolometric light curves. The vector Gaussian processes developed in this work are accessible via the Python library gp-multistate-kernel on GitHub. Our approach provides an efficient tool for analyzing light curve data, opening new possibilities for astrophysical research.