Principal components of dark energy with SNLS supernovae: the effects of systematic errors

TL;DR

This study quantifies how current Type Ia supernova (SN Ia) systematics influence dark energy constraints obtained by combining SNLS SN Ia data with BAO and CMB measurements. It adopts a general description of dark energy evolution through principal components (PCs) of the equation of state $w(z)$, alongside traditional parametrizations like constant $w$ and the $w_0$–$w_a$ form, and assesses constraints with and without SN systematics. The authors show that SN systematics shrink the generalized FoM by roughly a factor of 2–3, though constraints on more than five PCs remain robust, with the first two PCs being the most affected. They also demonstrate that finite BAO detection significance has only a modest effect on BAO-only constraints and is negligible for the combined SN+BAO+CMB result, underscoring the resilience of current cosmological inferences to plausible data limitations. Overall, while controlling systematics remains essential for future improvements, the present data provide strong evidence for dark energy and constrain its possible evolution across cosmic time.

Abstract

We study the effects of current systematic errors in Type Ia supernova (SN Ia) measurements on dark energy (DE) constraints using current data from the Supernova Legacy Survey (SNLS). We consider how SN systematic errors affect constraints from combined SN Ia, baryon acoustic oscillations (BAO), and cosmic microwave background (CMB) data, given that SNe Ia still provide the strongest constraints on DE but are arguably subject to more significant systematics than the latter two probes. We focus our attention on the temporal evolution of DE described in terms of principal components (PCs) of the equation of state, though we examine a few of the more common, simpler parametrizations as well. We find that the SN Ia systematics degrade the total generalized figure of merit (FoM), which characterizes constraints in multi-dimensional DE parameter space, by a factor of two to three. Nevertheless, overall constraints obtained on more than five PCs are very good even with current data and systematics. We further show that current constraints are robust to allowing for the finite detection significance of the BAO feature in galaxy surveys.

Figures (11)

• Figure 1: Hubble diagram for the compilation of all SN Ia data used in this paper, labeling SNe from each survey separately and showing the (diagonal-only) magnitude uncertainties. The solid black line represents the best fit to the data.
• Figure 2: Left panel: correlation matrix obtained from the complete covariance matrix $\mathbf{C}^\mathrm{full}$, sorted first by survey and then by redshift within each survey. Right panel: same, but using only the systematic covariance matrix $\mathbf{C}^\mathrm{sys}$. In both cases we assume $\alpha_s = 1.43$ and $\beta_c = 3.26$, the best-fit values for the flat $w = \mathrm{const}$ model. The right panel is similar to Fig. 12 from Conley2011, but we repeat it here and show the full covariance (left panel) for completeness.
• Figure 3: Measured values of $A(z)$ and their (diagonal-only) uncertainties for each effective redshift. The black curve shows $A(z)$ for a model that fits the data points well, and the parameters for this model are given in the legend.
• Figure 4: 68.3%, 95.4%, and 99.7% likelihood constraints on $\Omega_M$ and $w$, assuming a constant value for $w$ and a flat universe. We use only SN data and marginalize over the nuisance parameters. We compare the case of diagonal statistical errors only (shaded blue) with the full systematic covariance matrix (red).
• Figure 5: 68.3%, 95.4%, and 99.7% likelihood constraints on $\alpha_s$ and $\beta_c$, assuming a constant value for $w$ and a flat universe. We use only SN data and marginalize over $\mathcal{M}$, $\Omega_M$, and $w$. We compare the case of diagonal statistical errors only (shaded blue) with the full systematic covariance matrix (red).