Analyzing type Ia supernovae near-infrared light curves with Principal Component Analysis

TL;DR

This study analyzes Type Ia supernovae in the near-infrared (YJH) with Principal Component Analysis to identify dominant variability modes in rest-frame light curves. By preprocessing observed data, applying SNooPy for K-corrections and PISCOLA for interpolation, and performing PCA, the authors extract three principal components that explain ~90% of the variance, linking the first two to the timing and brightness of the secondary NIR peak. They interpret these components through Kasen2006 physics, associating them with $M_{ m Ni}$, Ni-mixing, Fe-group elements, and metallicity, and explore correlations with light-curve shape parameters ($s_{BV}$) and optical peaks, as well as color curves and host properties. Importantly, incorporating PCA coefficients as corrections in Y-band standardization reduces the intrinsic scatter of SN distances, highlighting a potential path toward improved cosmological use of SNe Ia in the NIR, albeit limited by sample size.

Abstract

Type Ia supernovae (SNeIa), the thermonuclear explosions of C/O white dwarf stars in binary systems, are phenomena that remain poorly understood. The complexity of their progenitor systems, explosion physics and intrinsic diversity poses not only challenges for their understanding as astrophysical objects, but also for their standardization and use as cosmological probes. Near-infrared (NIR) observations offer a promising avenue for studying the physics of SNeIa and for reducing systematic uncertainties in distance estimations, as they exhibit lower dust extinction and smaller dispersion in peak luminosity than optical bands. Here, Principal Component Analysis (PCA) is applied to a sample of SNeIa with well-sampled NIR (YJH-band) light curves to identify the dominant components of their variability and constrain physical underlying properties. The theoretical models of Kasen2006 are used for the physical interpretation of the PCA components, where we found the 56Ni mass to describe the dominant variability. Other factors, such as mixing and metallicity, were found to contribute significantly as well. However, some differences are found between the components of the NIR bands which may be attributed to differences in the explosion aspects they each trace. Additionally, the PCA components are compared to various light-curve parameters, identifying strong correlations between some components and peak brightness in both the NIR and optical bands, particularly in the Y band. When applying PCA to NIR color curves, we found interesting correlations with the host-galaxy mass, where SNeIa with redder NIR colors are predominantly found in less massive galaxies. We also investigate the potential for improved standardization in the Y band by incorporating PCA coefficients as correction parameters, leading to a reduction in the scatter of the intrinsic luminosity of SNeIa.

Figures (17)

• Figure 1: PISCOLA light-curve (YJH-bands) fits of three SNe Ia: 2005df (Literature; top), 2008gp (CSP; middle), and 2020jjf (DEHVILS; bottom). The photometric uncertainties propagates into the fits themselves.
• Figure 2: SNe Ia rest-frame light curves after the preprocessing steps of Sect. \ref{['subsec:preprocessing']}. The phase range covers from $-8$ to $+34$ days with respect to optical peak. The light curves are color-coded by $s_{BV}$ value, obtained with SNooPy fits using the max_model model, limiting the color range between $0.8 \leq s_{BV} \leq 1.2$ for visualization purposes, as the bulk of the sample falls within this range.
• Figure 3: PCA decomposition for the Y- (top row), J- (middle row), and H-band (bottom row) light curves. Color-coded are the contributions of each of the coefficients with a 2$\sigma$ range: $p_0$ (left column), $p_1$ (middle column), and $p_2$ (right column). The sample mean is shown as a solid red line. The values in parentheses are the percentages of the explained variance. Three components explain $\sim90\%$ of the variance for each of the bands.
• Figure 4: Comparison of the first ($p_0$; top row) and second ($p_1$; bottom row) PCA coefficients vs. the peak absolute magnitude for each NIR band. The Pearson correlation coefficient ($\rho$) and $p$-value for each of the comparisons, with their respective 1$\sigma$ uncertainty estimated by Monte-Carlo sampling, are shown with the respective linear relations (red lines). The null-hypothesis is that there is no correlation (zero slope). The panels have the same x- and y-axis ranges for visualization purposes.
• Figure 5: Same as Fig. \ref{['fig:Mmax_vs_p01']}, but for $s_{BV}$, obtained with SNooPy using the max_model.