WL and BAO Surveys and Photometric Redshifts

WL and BAO Surveys and Photometric Redshifts Lloyd Knox University of California, Davis Yong-Seon Song (U Chicago) Tony Tyson (UC Davis) and Hu Zhan (UC Davis) Also: Chris Fassnacht, Vera Margoniner and David Wittman

Outline How WL Probes Cosmology Photometric Redshift Challenge Strategies for Meeting the Challenge Combined WL + BAO Forecasts

Distance (and growth) reconstructed from LSST WL survey + Planck Knox, Song & Tyson (2005) With the parameters of the highz Universe pinned down by Planck, only thing left to measure is g(z) and D A (z) (here called r(z)) in the dark energy-dominated era *. They can both be reconstructed from tomographic cosmic shear data. D.E. constraints come almost entirely from D A (z) constraints (Simpson & Bridle 04, KST05). Sensitivity to D A comes from matter-radiation equality feature at k = H EQ -1. * exceptions: e.g. neutrinos

Need Redshifts WL shear power spectra do not depend on any redshifts (neither sources nor lenses) only on D os, D ol and D ls. But we want to constrain cosmology via D(z)! Without source z s we are stuck with weak constraints that would come from D(g) (Zhan & Knox 05). The need for z (and for photo-z) exposes otherwise very clean WL probe to risks of astrophysical modeling.

Bernstein and Jain 2003, Ma et al. 2005, Huterer et al. 2005 Redshift Error Tolerance Can tolerate large errors on z for individual objects, but need to know P(z) very well. For D A (z) / z want σ(<z>)/<z> < 0.3 σ(d)/d (in order to be subdominant) If σ(d)/d = 0.01 need σ(<z>) < 3 10-3 <z> <z> is not exactly desired quantity, but some weighted average of z. This means need to also know σ(z) just as well as we know <z>. Training set size: σ(σ(z)) = σ(z) (1/(2N spec )) 1/2 and σ(<z>) = σ(z)/(1/n spec ) 1/2 N ~ 1000 per z bin or ~10 4 total (for 1% distance errors, assuming σ(z) = 0.1). N larger due to non-gauss.

Spectroscopy of Faint Objects is Very Hard Steidel et al. (2003) get redshifts down to m=25.5 on Keck (but these are pre-selected to have strong emission lines fair sample concerns) LSST high source densities achieved by going to m=26.5 Due to night sky lines, we need Subaru + WFMOS in space (to solve the problem directly)!

Strategies for Meeting the Photoz Challenge Indirectly Use sub samples of galaxies as the source redshift population (to address fair sample concerns). Use galaxy clustering to control errors.

Strategies for Meeting the Photoz Challenge Indirectly Use cross-correlation cosmography (Jain & Taylor 2003, Bernstein & Jain 2004, Hu & Jain 2004, Song & LK 2004) since highly sensitive to z errors. Obtain subsample with spectroscopy and a subsample with ~15 bands stretching into IR to test extrapolation of ~ 5-band observations into magnitudes with no spectra. Plan shallower surveys.

Galaxy power spectra alone do not self-calibrate the photo-z s Forecasts for LSST photo-z BAO Zhan & Knox 2005 Contours of constant error on w 0 Contours of constant error on w a

Consistency Test Using Galaxy Using galaxy power spectra from 7 redshift bins, with a prior on the photo-z rms, the data can then constrain the mean redshift of each of the 7 galaxy groupings. Results for four of them are shown. Power Spectra Zhan & Knox 05

Effect on Cross-Correlation w (θ) W (θ) M. Quilici & H. Zhan With redshift errors Without redshift errors Angular separation Cross-correlation between two bins: z=.2-0.8 and z=0.8-1.4 based on Virgo Hubble Volume simulation θ Photo-z errors lead to correlations across photo-z bins that would otherwise be much smaller. P(z) assume Gaussian with σ(z)=0.1(1+z). Have not yet quantified utility of this signature.

Subsampling Strategies (and Curvature Comment) Tomographic WL 2-pt function for 40 gal/arcmin^2, half-sky, <z>=1.5 ( Deep ) bao survey high-z n reducer σ(w piv ) σ(w a ) σ(ω k ) none 1 0.023 0.18 ------ none 1 0.024 0.20 0.0010 none 0.3 0.025 0.25 0.0012 none 0.1 0.026 0.33 0.0013 none 0.01 0.027 0.40 0.0015 Note on importance of high z decreases d.e. model dependence of Ω k determination. See Knox 2005.

Subsampling Strategies (and Curvature Comment) Tomographic WL 2-pt function for 40 gal/arcmin^2, half-sky, <z>=1.5 ( Deep ) bao survey high-z n reducer * σ(w piv ) σ(w a ) σ(ω k ) none 1 0.023 0.18 ------ none 1 0.024 0.20 0.0010 none 0.3 0.025 0.25 0.0012 none 0.1 0.026 0.33 0.0013 none 0.01 0.027 0.40 0.0015 20,000sq. deg.pessimistic 0.01 0.026 0.37 0.0010 20,000sq. deg.optimistic 0.01 0.026 0.33 0.0007 *reduce n at z > 1.6 by this factor bao surveys at z>1.6 only

Shallower Surveys Deep : Tomographic WL 2-pt function for 40 gal/arcmin^2, half-sky, <z>=1.5) Shallow : Tomographic WL 2-pt function for 12 gal/arcmin^2, half-sky, <z>=0.5) WL survey bao survey σ(w piv ) σ(w a ) σ(ω k ) Deep none 0.024 0.20 0.0010 Shallow none 0.050 0.69 0.0032 Shallow 20,000sq.deg. pessimistic 0.049 0.59 0.0016 Shallow 20,000sq.deg.optimistic 0.044 0.46 0.0010 Dramatic degradation of WL survey quality factor of 2 to 3 worse errors Shallow survey significantly improved by high-z bao survey

2000 sq. deg WL Survey Deep : Tomographic WL 2-pt function for 40 gal/arcmin^2, half-sky, <z>=1.5) Shallow : Tomographic WL 2-pt function for 12 gal/arcmin^2, half-sky, <z>=0.5) 2000 : Tomographic WL 2-pt function for 12 gal/arcmin^2, 1/20 th -sky, <z>=0.5) WL survey bao survey σ(w piv ) σ(w a ) σ(ω k ) Deep none 0.024 0.20 0.0010 Shallow none 0.050 0.69 0.0032 2000 none 0.11 1.20 0.0060 2000 1000sq.deg spec 0.068 0.84 0.0015 2000 phot4000o 0.079 0.93 0.0017 2000 phot4000p 0.11 1.08 0.0025

Summary WL is a distance-redshift probe (like bao and SNe) Mean distances to sources can be measured very well need mean redshifts to be measured very well also. Understanding photo-z redshift errors sufficently well is a great challenge for photometric BAO and WL. We are pursuing a number of strategies.