CFHTLenS revisited: assessing concordance with Planck including astrophysical systematics

TL;DR

This work reevaluates CFHTLenS cosmic shear in the light of updated covariance from a large suite of N-body simulations and a new CosmoMC module that jointly models intrinsic alignments, baryonic feedback via hmcode, and photometric redshift uncertainties. It quantitatively assesses the concordance between CFHTLenS and Planck using Bayesian evidence and DIC-based tests, showing that the inferred level of agreement depends strongly on the treatment of systematics. A nonzero IA amplitude is favored by CFHTLenS alone and can increase tension with Planck, whereas more conservative joint-systematics scenarios can reduce or reverse this tension. The study provides a public, extensible pipeline for joint cosmic shear analyses including IA, baryons, and photo-z uncertainties, with extensions toward galaxy-galaxy lensing and clustering probes.

Abstract

We investigate the impact of astrophysical systematics on cosmic shear cosmological parameter constraints from the Canada-France-Hawaii Telescope Lensing Survey (CFHTLenS), and the concordance with cosmic microwave background measurements by Planck. We present updated CFHTLenS cosmic shear tomography measurements extended to degree scales using a covariance calibrated by a new suite of N-body simulations. We analyze these measurements with a new model fitting pipeline, accounting for key systematic uncertainties arising from intrinsic galaxy alignments, baryonic effects in the nonlinear matter power spectrum, and photometric redshift uncertainties. We examine the impact of the systematic degrees of freedom on the cosmological parameter constraints, both independently and jointly. When the systematic uncertainties are considered independently, the intrinsic alignment amplitude is the only degree of freedom that is substantially preferred by the data. When the systematic uncertainties are considered jointly, there is no consistently strong preference in favor of the more complex models. We quantify the level of concordance between the CFHTLenS and Planck datasets by employing two distinct data concordance tests, grounded in Bayesian evidence and information theory. We find that the two data concordance tests largely agree with one another, and that the level of concordance between the CFHTLenS and Planck datasets is sensitive to the exact details of the systematic uncertainties included in our analysis, ranging from decisive discordance to substantial concordance as the treatment of the systematic uncertainties becomes more conservative. The least conservative scenario is the one most favored by the cosmic shear data, but it is also the one that shows the greatest degree of discordance with Planck. The data and analysis code are public at https://github.com/sjoudaki/cfhtlens_revisited

Figures (12)

• Figure 1: The ratio of shear correlation functions for tomographic bin combinations {1,7} and {7,7}, taken with respect to hmcode with feedback amplitude $\log B = 0.496$, defined in equation (\ref{['eq:cm']}), including no systematic uncertainties (denoted as $\xi_\pm[{\rm fid}]$). For consistency, we fix the underlying cosmology to that of the best-fit cosmology of this 'fiducial' case. We allow for the Takahashi12 version of halofit (solid black), hmcode with $\log B = 0.3$ (dashed red), intrinsic alignments with $\{A, \eta, \beta\} = \{1,0,0\}$ (dot-dashed green), intrinsic alignments with $\{A, \eta, \beta\} = \{1,0,1\}$ (dotted blue), intrinsic alignments with $\{A, \eta, \beta\} = \{1,1,0\}$ (dot-dashed cyan), and photometric redshift uncertainties where all bins are positively perturbed by ${\Delta}z = 0.05$ (solid pink). The $\log B$ value of 0.496 corresponds to the DM-only case, while $\log B = 0.3$ agrees with the AGN case of the OWL simulations. The parameters $\{A, \eta, \beta\}$ refer to the intrinsic alignment amplitude, redshift dependence, and luminosity dependence, respectively, of the IA model defined in equation (\ref{['eqn:fz']}), while the photo-z shifts are defined in equation (\ref{['eqn:photoz']}). The IA model with $\{A, \eta, \beta\} = \{1,0,1\}$ lies along the unity line because the luminosity $L/L_0 < 1$ in each tomographic bin, such that a positive value of $\beta$ suppresses the IA signal (analogously, a negative value of $\eta$ for the redshift dependence would have a similar effect).
• Figure 2: Stacked bpz redshift probability distributions for CFHTLenS, weighted by the lensfit weights, in the seven tomographic photo-$z$ bins used in our analysis.
• Figure 3: Measurements of the cosmic shear statistics $\xi_+$ (upper triangle) and $\xi_-$ (lower triangle) against angular scale in arcminutes for all unique pairs of the 7 tomographic source bins, defined in Section \ref{['measlab']}. The error bars are determined using the mock catalogues described in Section \ref{['covlab']}. The grey regions correspond to angular scales that were removed from the cosmology analysis, due to low signal-to-noise or covariance under-estimation (discussed in Section \ref{['measlab']}). Open circles denote negative points. Fiducial theory lines have been included in red (solid) for comparison.
• Figure 4: Ratio of the cosmic shear error determined by jackknife sampling using athena, to that determined from the suite of mock catalogues, averaged across all tomographic bins as a function of angular scale for $\xi_+$ (black solid circles) and $\xi_-$ (red open circles).
• Figure 5: The correlation coefficient of the covariance matrix of the full data vector, plotted using a greyscale where white represents $r=0$ and black represents $r=1$.