PDF《Source Redshift Distributions》

6 Oct

2017 2 DES Collaboration galaxies from DES photometry in the griz bands. These pho- tometric redshifts ( photo-z ) are used twice: ?rst to assign galaxies to tomographic bins, and then to determine the nor- malised redshift distribution ni(z) of galaxies in the i-th bin. This paper describes the calibration of the source galaxy redshift distributions used in the DES Y1 Key Project by looking at their cross-correlations with high-?delity lens galaxy photometric redshifts. In analogy with photomet- ric redshifts, we shall refer to these estimates of the red- shift distribution as clustering-z . The redshift distribu- tions are crucial for the prediction of the observable cosmo- logical signals, and their uncertainties must be propagated into our cosmological parameter inference pipeline. For our measurement precision, the most important parameter of a source galaxy redshift distribution is its mean redshift. We focus here on the calibration of that parameter using angu- lar cross-correlations with redMaGiC galaxies. Gatti et al. (2017, henceforth G17) describe how we use simulations to estimate the systematic uncertainties in this method. Hoyle et al. (2017) describe the binning and redshift determina- tion of source galaxies as well as their validation with 30- band COSMOS redshifts (Laigle et al. 2016). The assignment and validation of lens galaxy redshifts are described in Rozo et al. (2016), Elvin-Poole et al. (2017), and Cawthon et al. (2017). It has now been almost a decade since Newman (2008) ?rst demonstrated on simulations the use of angular cross- correlations with ?nely-binned, high-?delity-redshift galax- ies to determine redshift distributions, and over a decade since Schneider et al. (2006) proposed using galaxy angular two-point correlation functions to determine redshift distri- butions. Since then, the method has been applied to both simulation and real data (M? enard et al. 2013;

Schmidt et al. 2013;

McQuinn &

White 2013;

Rahman et al. 2015). Of par- ticular relevance is the recent work in Hildebrandt et al. (2017), Johnson et al. (2017), and Morrison et al. (2017), where clustering-z methods were applied to Kilo-Degree Sur- vey photometric data by cross-correlating with spectroscopic redshifts from the Galaxy And Mass Assembly (GAMA) sur- vey and the Sloan Digital Sky Survey (SDSS). These papers used clustering-z distributions as alternative redshift distri- butions to those from photometric techniques, and demon- strated the viability C and the potential C of using clustering- z methods to determine redshift distributions. In our present work, we make two signi?cant modi?cations. First, instead of spectroscopic redshifts over a minimal area in the sky (of- ten only 10-100 deg2), we use the high-?delity photometric redshifts determined by the redMaGiC algorithm to mea- sure clustering-z over our entire

1321 deg2 footprint. This is also di?erent from our previous work on DES Science Ver- i?cation data in Davis et al. (2017), where we instead used redMaPPer galaxy clusters (Ryko? et al. 2014, 2016). red- MaGiC redshifts are more than su?ciently accurate for our purposes, and we have many more redMaGiC galaxies than spectroscopic galaxies in our footprint. Second, because the limited redshift range of our redMaGiC photo-z'

s means we can only measure part of the source redshift distribution directly with clustering-z, we use the clustering-z instead to calibrate shifts to the redshift distributions measured by photo-z. This calibration procedure is combined with COS- MOS calibrations described in Hoyle et al. (2017) to obtain yet tighter constraints on the mean redshift of each source bin. This paper is organised as follows. In Section