Comparing the light curves of simulated Type Ia Supernovae with observations using data-driven models

TL;DR

This paper introduces a quantitative, data-driven approach to compare simulated Type Ia Supernova light curves with observations, using a statistical framework to evaluate model accuracy and consistency with real data.

Contribution

It develops a robust method combining light curve fitting and population modeling to assess the agreement of explosion models with observed supernovae.

Findings

The method provides a reliable figure of merit for model evaluation.

It can distinguish between different explosion models based on observational consistency.

The approach accounts for systematic uncertainties and outliers in supernova data.

Abstract

We propose a robust, quantitative method to compare the synthetic light curves of a Type Ia Supernova (SNIa) explosion model with a large set of observed SNeIa, and derive a figure of merit for the explosion model's agreement with observations. The synthetic light curves are fit with the data-driven model SALT2 which returns values for stretch, color, and magnitude at peak brightness, as well as a goodness-of-fit parameter. Each fit is performed multiple times with different choices of filter bands and epoch range in order to quantify the systematic uncertainty on the fitted parameters. We use a parametric population model for the distribution of observed SNIa parameters from large surveys, and extend it to represent red, dim, and bright outliers found in a low-redshift SNIa data set. We discuss the potential uncertainties of this population model and find it to be reliable given the current uncertainties on cosmological parameters. Using our population model, we assign each set of fitted parameters a likelihood of being observed in nature, and a figure of merit based on this likelihood. We define a second figure of merit based on the quality of the light curve fit, and combine the two measures into an overall figure of merit for each explosion model. We compute figures of merit for a variety of 1D, 2D and 3D explosion models and show that our evaluation method allows meaningful inferences across a wide range of light curve quality and fitted parameters.