ZTF Early Observations of Type Ia Supernovae III: Early-Time Colors as a Test for Explosion Models and Multiple Populations

Mattia Bulla, Adam A. Miller, Yuhan Yao, Luc Dessart, Suhail Dhawan, Semeli Papadogiannakis, Rahul Biswas, Ariel Goobar, S. R. Kulkarni, Jakob Nordin, Peter Nugent, Abigail Polin, Jesper Sollerman, Eric C. Bellm, Michael W. Coughlin, Richard Dekany, V. Zach Golkhou, Matthew J. Graham, Mansi M. Kasliwal, Thomas Kupfer, Russ R. Laher, Frank J. Masci, Michael Porter, Ben Rusholme, David L. Shupe

Published 2020-01-02Version 1

Colors of Type Ia supernovae in the first few days after explosion provide a potential discriminant between different models. In this paper, we present $g-r$ colors of 38 Type Ia supernovae discovered within 3 days from first light by the Zwicky Transient Facility in 2018, a sample which is twice as large as that in the literature. We find that $g-r$ colors are intrinsically rather homogeneous at early phases, with about half of the dispersion in the first 5 days attributable to photometric uncertainties ($\sigma_\mathrm{noise}\sim\sigma_\mathrm{int}\sim0.25$ mag). Colors are nearly constant starting from 5 days after first light ($g-r\sim-0.15$ mag), while the time evolution at earlier epochs is characterized by a continuous range of slopes, from events rapidly transitioning from redder to bluer colors (slope of $\sim-0.25$ mag day$^{-1}$) to events with a flatter evolution. The continuum in the slope distribution is in good agreement both with models requiring some amount of $^{56}$Ni mixed in the outermost regions of the ejecta and with "double-detonation" models having thin helium layers ($M_\mathrm{He}=0.01\,M_\odot$) and varying carbon-oxygen core masses. At the same time, four events show evidence for a distinctive "red bump" signature predicted by "double-detonation" models with larger helium masses. We finally identify a significant correlation between the early-time $g-r$ slopes and supernova brightness, with brighter events associated to flatter color evolution (p-value=0.001). The distribution of slopes, however, is consistent with being drawn from a single population, with no evidence for two components as claimed in the literature.