photometric redshifts: some considerations for the ctio dark energy camera survey

Photometric Redshifts: Some Considerations

for the CTIO Dark Energy Camera Survey

Huan Lin

Experimental Astrophysics Group

Fermilab

● Photometric Redshifts for Red (Cluster) Galaxies– Illustration using the SDSS luminous red galaxy (LRG) sample– Monte Carlo simulations based on model spectral energy

distributions (SEDs)

● Photometric Redshifts for the General Galaxy Population– Illustrative example from CFH12K imaging of a CNOC2

redshift survey field

Outline

My focus is on effects of photometric errors on photo-z measurement

A sample of SDSS luminous red galaxies (LRGs)

photo-z offset per 0.1 mag color offset

redshift

● Model SEDs– Non-evolving CWW elliptical; normalized to apparent

magnitude z=23 at redshift=1– Evolving Pegase-2 model from Jim Annis; a 0.5 L* galaxy for

redshifts 0-1– Flat, ΩΛ= 0.7 cosmology

● Photometric Errors– Filter g r i z– Exposure 900 900 1700 900 sec– 10σ mag 25.0 24.5 24.3 23.3– Simply scale S/N assuming constant noise from sky

background, plus a calibration error floor of 3%

Red Galaxy Monte Carlo Simulations

color

redshift

Results of template and polynomial photo-z fits

photometric redshift error σ(z) ~ 0.03

Photo-z bias from using the wrong template

bias is about the same as the statistical uncertainty σ(z) ~ 0.03

Photo-z bias from 0.03 mag offset in given filter

bias looks negligible

Photo-z bias from 0.1 mag offset in given filter

bias is about the same as the statistical uncertainty σ(z) ~ 0.03

● CFH12K Images (CNOC2 0223B4 field)– 3.6m CFHT– Exposure times comparable: B 1320s, V 900s, R 900s, I 900s– But seeing is better than typical CTIO: 0.5-0.9 arcsec– Not shown today, but z-band data, plus other fields, are also

available for photo-z tests

● CNOC2 Redshift Survey – General field galaxy population– Caveat: calibrators mainly limited to z < 0.7 and R < 22

CFH12K/CNOC2 BVRI Photometric Redshifts

BVRI photo-z results: CNOC2 0223B4 field from CFH12K

photometric redshift error σ(z) ~ 0.05

Photo-z bias from 0.03 mag offset in given filter

bias relative to original solution looks negligible

Photo-z bias from 0.1 mag offset in given filter

bias relative to original solution is about the same as the statistical uncertainty σ(z) ~ 0.05

Conclusions and Outlook

● Simple Monte Carlo simulations indicate photo-z error σ(z) ~ 0.03 for red (cluster) galaxies, but with some redshift dependence and some degeneracy at z > 0.8

● Should acquire spectroscopic cluster calibration sample to accurately determine cluster red galaxy SED

● General field galaxy sample from example CFH12K/CNOC2 field indicates σ(z) ~ 0.05, but with color/magnitude dependence, and redshift-dependent systematics

● Will also need to test using real/simulated galaxies at both fainter magnitudes and higher redshifts

Conclusions and Outlook (cont’d)

● Photometric calibration errors of 0.03 mag introduce negligible bias, while errors of 0.1 mag introduce biases comparable to photo-z statistical error σ(z)

● Work is in progress on improving photo-z fits and reducing systematics, using CFH12K/CNOC2 and SDSS samples

● Not yet addressed question of how many calibrators are needed to sufficiently characterize photo-z error distributions for cluster counts, weak lensing, etc.