Measuring Type Ia Supernova Populations of Stretch and Color and Predicting Distance Biases

TL;DR

This paper improves supernova simulations by accurately modeling intrinsic color and stretch populations, revealing biases in luminosity measurements, and applying corrections that refine cosmological parameter estimates.

Contribution

It introduces a robust method to evaluate and correct biases in supernova color and stretch distributions, enhancing the accuracy of cosmological inferences.

Findings

Detected significant biases in Hubble residuals related to color and stretch.

Correcting these biases reduces the intrinsic scatter of supernova brightness measurements.

Refined population models lead to a smaller bias in dark energy equation-of-state parameter w.

Abstract

Simulations of Type Ia Supernovae (SNIa) surveys are a critical tool for correcting biases in the analysis of SNIa to infer cosmological parameters. Large scale Monte Carlo simulations include a thorough treatment of observation history, measurement noise, intrinsic scatter models and selection effects. In this paper, we improve simulations with a robust technique to evaluate the underlying populations of SNIa color and stretch that correlate with luminosity. In typical analyses, the standardized SNIa brightness is determined from linear `Tripp' relations between the light curve color and luminosity and between stretch and luminosity. However, this solution produces Hubble residual biases because intrinsic scatter and measurement noise result in measured color and stretch values that do not follow the Tripp relation. We find a $10\sigma$ bias (up to 0.3 mag) in Hubble residuals versus color and $5\sigma$ bias (up to 0.2 mag) in Hubble residuals versus stretch in a joint sample of 920 spectroscopically confirmed SNIa from PS1, SNLS, SDSS and several low-z surveys. After we determine the underlying color and stretch distributions, we use simulations to predict and correct the biases in the data. We show that removing these biases has a small impact on the low-z sample, but reduces the intrinsic scatter $\sigma_{\textrm{int}}$ from $0.101$ to $0.083$ in the combined PS1, SNLS and SDSS sample. Past estimates of the underlying populations were too broad, leading to a small bias in the equation-of-state of dark energy $w$ of $\Delta w=0.005$.