[Paper Review] Supernova Legacy Survey (SNLS) : real time operations and photometric analysis

N. Palanque‐Delabrouille|arXiv (Cornell University)|Sep 15, 2005

Gamma-ray bursts and supernovae3 citations

TL;DR

This paper presents the real-time operations and photometric analysis pipeline of the Supernova Legacy Survey (SNLS), which uses image subtraction and multi-color light curve fitting to detect and classify type Ia supernovae with high efficiency. The offline analysis complements real-time detection by identifying additional candidates—especially at high redshift—reducing selection biases and enabling precise cosmological constraints with 10% accuracy on ΩM and ΩΛ and 0.1 precision on w.

ABSTRACT

Type Ia supernovae (SN Ia) have provided the first evidence for an accelerating universe and for the existence of an unknown ``dark energy'' driving this expansion. The 5-year Supernova Legacy Survey (SNLS) will deliver \~700 type Ia supernovae and as many type II supernovae with well-sampled light curves in 4 filters g', r', i' and z'. The current status of the project will be presented, along with the real time processing leading to the discovery and spectroscopic observation of the supernovae. We also present an offline selection of the SN candidates which aims at identifying and eliminating potential selection biases.

Motivation & Objective

• To improve cosmological constraints on dark energy by detecting and characterizing type Ia supernovae with high-precision light curves.
• To reduce selection biases in supernova detection by implementing an independent offline analysis pipeline.
• To enhance the efficiency of spectroscopic follow-up by estimating supernova maximum time and redshift from pre-maximum photometry.
• To increase the number of confirmed type Ia supernovae, especially at high redshift (z ≈ 0.6–1.2), through deeper, systematic offline processing.
• To enable redshift estimation from photometric light curves alone using a χ² map against SN Ia templates, supporting cosmological analysis of non-spectroscopic candidates.

Proposed method

• Real-time detection uses image subtraction via the Alard algorithm (France) and a non-parametric method (Canada) to identify transient candidates on nightly exposures.
• Candidates are filtered through automated artifact rejection (e.g., cosmic rays, saturated stars) and subjected to visual inspection to select ~5 per night for spectroscopy.
• Multi-color photometric light curves are fitted with SN Ia templates to estimate time of maximum brightness and redshift, optimizing spectroscopic exposure planning.
• Offline analysis constructs full two-year light curves from all subtraction images, using the SALT photometry code to ensure completeness and reduce bias.
• A three-step offline selection process excludes artifacts, stars/galaxies, and non-supernova-like light curves, focusing on significant, isolated, and template-fitting transients.
• Redshift is estimated via χ² minimization across redshifts (0–1.2) using 4-band light curves and the SALT SN Ia model, achieving 0.05–0.10 statistical uncertainty.

Experimental results

Research questions

• RQ1 How can real-time supernova detection be optimized to maximize the yield of type Ia supernovae for cosmological analysis?
• RQ2 To what extent do visual inspections in real-time pipelines introduce selection biases, and how can they be mitigated?
• RQ3 Can offline photometric analysis recover additional type Ia supernovae—especially at high redshift—beyond the real-time pipeline?
• RQ4 How accurately can redshift be estimated from multi-band photometry alone using SN Ia light curve templates?
• RQ5 What is the expected improvement in cosmological parameter constraints (ΩM, ΩΛ, w) from the full SNLS dataset?

Key findings

• As of January 2005, the SNLS had detected ~650 supernova candidates, with ~230 having spectroscopic data and ~150 confirmed as type Ia supernovae.
• The offline analysis pipeline is expected to recover ~40 additional supernovae per year, primarily at redshifts between 0.6 and 1.2, due to deeper magnitude reach (0.5 mag deeper than real-time).
• Photometric redshift estimation using the SALT model and χ² minimization achieves a statistical uncertainty of 0.05 for secure type Ia candidates and 0.10 for possible ones.
• The offline selection process effectively excludes variable stars, active galactic nuclei, and photometric artifacts by requiring isolated, significant, and template-fitting light curves.
• The survey’s design enables measurement of ΩM and ΩΛ to better than 10% and w to ~0.1, significantly improving constraints on dark energy.
• Light curves in the i′ filter show maximum brightness measurable with high accuracy, supporting robust cosmological distance measurements.