The Twins Embedding of Type Ia Supernovae I: The Diversity of Spectra at Maximum Light

TL;DR

This study introduces the Twins Embedding, a nonlinear spectral parameterization of Type Ia supernovae at maximum light, revealing that their intrinsic spectral diversity is low and can be explained with fewer dimensions than previously thought.

Contribution

The paper presents a novel nonlinear embedding method for supernova spectra that captures intrinsic diversity more efficiently than linear models, improving understanding of supernova variability.

Findings

84.6% of spectral variance is common across SNe Ia

Intrinsic dispersions as low as ~0.02 mag in certain spectral regions

89.2% of diversity explained by a 3D model plus color

Abstract

We study the spectral diversity of Type Ia supernovae (SNe Ia) at maximum light using high signal-to-noise spectrophotometry of 173 SNe Ia from the Nearby Supernova Factory. We decompose the diversity of these spectra into different extrinsic and intrinsic components, and we construct a nonlinear parameterization of the intrinsic diversity of SNe Ia that preserves pairings of "twin" SNe Ia. We call this parameterization the "Twins Embedding". Our methodology naturally handles highly nonlinear variability in spectra, such as changes in the photosphere expansion velocity, and uses the full spectrum rather than being limited to specific spectral line strengths, ratios or velocities. We find that the time evolution of SNe Ia near maximum light is remarkably similar, with 84.6% of the variance in common to all SNe Ia. After correcting for brightness and color, the intrinsic variability of SNe Ia is mostly restricted to specific spectral lines, and we find intrinsic dispersions as low as ~0.02 mag between 6600 and 7200 A. With a nonlinear three-dimensional model plus one dimension for color, we can explain 89.2% of the intrinsic diversity in our sample of SNe Ia, which includes several different kinds of "peculiar" SNe Ia. A linear model requires seven dimensions to explain a comparable fraction of the intrinsic diversity. We show how a wide range of previously-established indicators of diversity in SNe Ia can be recovered from the Twins Embedding. In a companion article, we discuss how these results an be applied to standardization of SNe Ia for cosmology.