Not All Who Wander Are Lost: Early Excess Demographics in the Volume-limited ZTF DR2 SN Ia Sample

Technical Summary: Not All Who Wander Are Lost: Early Excess Demographics in the Volume-limited ZTF DR2 SN Ia Sample

Problem and Motivation
Early-time flux excesses in Type Ia supernovae (SNe Ia) offer critical insights into progenitor systems and explosion mechanisms, potentially distinguishing between scenarios such as ejecta-companion interaction, shallow radioactive heating, circumstellar material interaction, or double detonations. While individual events and smaller searches have been reported, demographic studies have historically been limited by sample size and heterogeneity. Distinguishing physical scenarios in single events is challenging due to overlapping signatures and viewing-angle dependencies. Consequently, there is a need for large, homogeneous samples to quantify the incidence of early-time excesses and their correlations with SN and host-galaxy parameters.

Methodology
This study presents a systematic search for early-time excess emission using the Zwicky Transient Facility (ZTF) Data Release 2 (DR2). The analysis focuses on a volume-limited sample ($z < 0.06$) of 1,547 spectroscopically classified SNe Ia to minimize Malmquist bias and selection effects.

The methodology involves a rigorous multi-stage pipeline:

1. Data Selection: From the initial ZTF DR2 sample, the authors applied redshift cuts ($z < 0.06$), subtype filters (normal or 1991T-like), and strict quality cuts on SALT2 parameters (stretch $x_1$, color $c$, and time of peak $t_0$) and Milky Way extinction.
2. Photometric Cleaning: Light curves were cleaned according to ZTF quality control guidelines, removing points with unphysical errors, poor PSF fits, or cloud contamination. Only $g$- and $r$-band data were utilized due to the lower cadence of the $i$-band.
3. Model Fitting: For each SN, joint two-band fits were performed on the early-time light curves (from $t_0 - 40$ days to $t_0 - N$ days, where $N$ ranges from 7 to 13). Two models were compared: a simple power-law (PL) rise and a power-law plus Gaussian (PL+G) model to account for potential bumps.
4. Detection Criteria:
   • Coverage Gates: Strict requirements were imposed to ensure model reliability, including pre-explosion baselines, immediate post-explosion sampling, and late-time coverage.
   • Model Selection: The Bayesian Information Criterion (BIC) was used to select the preferred model. A PL+G model was favored if $\Delta \text{BIC} > 2$.
   • Excess Flagging: Candidates were flagged as having a "bump" if residuals relative to the baseline model exceeded $3\sigma$ within the window $[t_i, t_i + 5]$ days.
   • Robustness: A "Consecutive-N" test was applied to bump candidates, requiring the detection to persist across multiple $N$ values to avoid false positives driven by single data points.
   • Visual Inspection: Final catalogs were vetted manually to remove objects with galaxy contamination, single-epoch spikes, or inconsistent multi-band behavior.

Key Contributions and Results
The study produces the largest homogeneous catalog of early-excess SNe Ia to date, derived from a coherent analysis of the ZTF DR2 volume-limited sample.

• Final Catalogs: From an initial 1,547 SNe, the final sample consists of 42 robust early-excess candidates ("bump") and 110 "no-bump" candidates.
• Demographic Differences: Using Welch's t-tests and Kolmogorov-Smirnov tests, the authors identified significant statistical differences between the two populations:
  • Light-Curve Properties: The most significant differences are found in the SALT2 stretch parameter $x_1$ ($7.91\sigma$) and the integrated flux under the $r$-band secondary maximum, $F_{r2}$ ($6.25\sigma$). The bump population exhibits systematically larger $x_1$ (broader light curves) and higher $F_{r2}$.
  • Host Environment: The bump population is found in bluer local environments ($(g-z)_{\text{local}}$, $3.41\sigma$) and lower local stellar mass regions ($\log_{10}(M_*/M_\odot)_{\text{local}}$, $2.73\sigma$). Differences in global host mass and SALT2 color $c$ are weak or negligible.
  • Redshift Independence: The significant differences in $x_1$ and $F_{r2}$ persist across four redshift bins, suggesting these are intrinsic population differences rather than artifacts of redshift-dependent selection effects.
• Comparison with Prior Work:
  • Ye et al. (2024): The authors recover 8 of the 11 bump candidates identified by Ye et al. (at $z < 0.06$) after applying their stricter robustness cuts. Unlike Ye et al., who found no significant difference in $x_1$, this study finds a highly significant shift to larger $x_1$ in the bump population. The authors attribute this discrepancy to the larger, more homogeneous sample and stricter coverage/robustness criteria used here.
  • Deckers et al. (2022): The results are consistent with Deckers et al., who noted a preference for high-stretch values and lower-mass hosts in their smaller sample of six excess events.

Significance and Claims
The paper claims to provide the first systematic, volume-limited demographic study of early excesses in SNe Ia using ZTF DR2. The primary significance lies in the robust statistical evidence linking early-time excesses to specific light-curve shapes (larger stretch) and secondary maximum properties, rather than just color or host mass.

The authors interpret these trends as suggesting that early excesses may be associated with a specific progenitor channel or explosion physics that produces broader light curves and distinct secondary maxima, potentially linked to higher $^{56}\text{Ni}$ production or specific ejecta structures. The preference for bluer, lower-mass local environments supports the idea that local star-formation conditions influence the probability of an early excess. The authors emphasize that while these demographic trends constrain the origin of early excesses, they do not definitively favor a single physical model (e.g., companion shock vs. double detonation) due to the overlap in predicted morphologies and viewing-angle dependencies. The work serves as a foundation for future analyses with larger samples and physically motivated modeling to connect these demographic correlations to specific progenitor channels.