1. Cosmology with Photometric redsfhits
Filipe Batoni Abdalla
M. Banerji, S. Bridle, E. Cypriano, O. Lahav, J Tang, J Weller (UCL), A. Amara (Saclay), P. Capak, J. Rhodes (Caltech/JPL), H. Lin (Chicago)

2. Outline: Quick pass over photo-z & weak lensing
The DUNE mock catalogues
Results from the Fisher analysis on the mocks
More problems: Intrinsic alignements
Sensitivity of weak lensing to w(z)

3. Photometric Redshifts
Photometric redshifts (photo-z’s) are determined from the fluxes of galaxies through a set of filters
May be thought of as low-resolution spectroscopy
Photo-z signal comes primarily from strong galaxy spectral features, like the 4000 Å break, as they redshift through the filter bandpasses
All key projects depend crucially on photo-z’s
Photo-z calibrations will be
optimized using both simulated catalogs and images.
Galaxy spectrum at 3 different redshifts, overlaid on griz and IR bandpasses

4. Template Fitting methods Training Set Methods
Use a set of standard SED’s - templates (CWW80, etc.)
Calculate fluxes in filters of redshifted templates.
Match object’s fluxes (2 minimization)
Outputs type and redshift
Bayesian Photo-z
Determine functional relation
Examples
Nearest Neighbors
(Csabai et al. 2003)
Polynomial Nearest Neighbors
(Cunha et al.
in prep. 2005)
Polynomial
(Connolly et al. 1995)
Hyper-z (Bolzonella et al. 2000)
BPZ (Benitez 2000)
Neural Network
(Firth, Lahav & Somerville 2003;
Collister & Lahav 2004)
Cross correlations (Newman)

5. Statistical measure of shear pattern, ~1% distortion
Background sources
Background sources
Background sources
Background sources
Dark matter halos
Dark matter halos
Dark matter halos
Dark matter halos
Dark matter halos
Observer
Observer
Cosmic Web – bottom up scenario – clusters then filaments then walls (membranes).
Note that filaments not well traced by galaxies & too ephemeral to emit x-rays, so lensing only way to detect.
Statistical measure of shear pattern, ~1% distortion
Radial distances depend on geometry of Universe
Foreground mass distribution depends on growth of structure

6. Statistical measure of shear pattern, ~1% distortion
Background sources
Background sources
Background sources
Background sources
Dark matter halos
Dark matter halos
Dark matter halos
Dark matter halos
Dark matter halos
Observer
Observer
Cosmic Web – bottom up scenario – clusters then filaments then walls (membranes).
Note that filaments not well traced by galaxies & too ephemeral to emit x-rays, so lensing only way to detect.
Statistical measure of shear pattern, ~1% distortion
Radial distances depend on geometry of Universe
Foreground mass distribution depends on growth of structure

7. DUNE: Dark UNiverse Explorer
Mission baseline:
1.2m telescope
FOV 0.5 deg2
PSF FWHM 0.23’’
Pixels 0.11’’
GEO (or HEO) orbit
Surveys (3-year initial programme):
WL survey: 20,000 deg2 in 1 red broad band, 35 galaxies/amin2 with median z ~ 1, ground based complement for photo-z’s
Near-IR survey (Y,J,H). Deeper than possible from ground. Secures z > 1 photo-z’s
Changes are currently being discussed at ESA: i.e. merging of DUNE and SPACE
(we will hear more about this in Talks thurs Rassat/Guzzo), inclusing of a small spectrograph on the near-IR plane
4/14/2019

8. Surveys considered: galaxies with RIZ<25 considered

9. JPL Simulated catalogueAv
Type z

10. Know the requirements:
Catastrophic
outliers
Biases
Uninformative
region
Abdalla et al. astro-ph:
A case study: the DUNE satellite
I have performed analysis within the DES framework as well: VDES

12. Mock dependence: comparison to DES mocks.
DES (grizY)
DES+VISTA(JHK)
M. Banerji, F. B. Abdalla, O. Lahav, H. Lin et al.
In regions of interest photo-z
are worst by 30%

13. FOM: Results & Number of spectra needed
FOM prop 1/ dw x dw’
IR improves error on DE parameters by a factor of depending on optical data available
If u band data is available improvement is minimal
Number of spectra needed to calibrate these photo-z for wl is around 10^5 in each of the 5 redshift bins
Fisher matrix analysis marginalizing over errors in photo-z.

14. Intrinsic alignements.
Additional contributions
What we
measure
Cosmic shear

15. Intrinsic-shear correlation (GI)
Galaxy at z1 is tidally sheared
Hirata & Seljak
Dark matter at z1
Net anti-correlation between galaxy ellipticities with no preferred scale
High z galaxy gravitationally
sheared tangentially

16. Removing intrinsic alignments:
Finding a weighting function insensitive of shape-shear correlations. (P. Schneider)
- Is all the information still there?
Modelling of the intrinsic effects (Bridle & King.)
- FOM definitely will decreased as need to constrain other parameters in GI correlations.
Using galaxy-shear correlation function.
In any case there will be the need of a given photometric redshift accuracy.

17. Different Cl contributions:
Bridle & King

18. Are photo-zs good enough?
The FOM is a slow function of the photo-z quality if we consider only the shear-shear term.
If we consider modelling the shape-shear correlations this is not the case anymore.
This does not include the galaxy-shear correlation function so “reality” is most likely in between this “pessimistic” result and the optimistic result of neglecting GI
Cypriano, Lahav & Rhodes
Abdalla, Amara, Capak
High demand on photo-z
for intrinsic alignement calibration
Bridle & King

19. PCA and Fisher Information Matrix
Fisher Information Matrix is an efficient method to measure the covariance of the random variables
Fisher information matrix F is defined as
To combine different experiments F=F_1+F_2
To marginalize over parameters
We include a parameter set combined with cosmological parameters, w and other nuisance parameters
In the e-vector basis, w is reconstructed as
For more details see posted by Tang,
Where she reproduced
all the DETF report work using
w binning + e-modes formalism

20. Redshift information in e-modes:

21. Conclusions
Today dw=1/10 prospect: dwxdw’=1/160 but there is a big demand on photometric redshifts, specially for future surveys such as DUNE alone.
Need of around 10^5 spectra in ~5 redshift bins
Removing poor photo-z is possible, removes systematic effects and does not hit the statistical limits of certain surveys.
IR data can significantly improve FOM form 1.3 to 1.7
Importance of the u band filter, potentially being as important as the IR.
It is possible to measure intrinsic alignments with spectroscopic redshift surveys, need to assess it that is possible with photo-z.
Map the redshift sensitivity to w for future wl surveys.