Context :

Aims : We want to derive bias free , accurate photometric redshifts for those fields of the Canada-France-Hawaii Telescope Legacy Survey ( CFHTLS ) Wide Data which are covered in the u ^ { * } ,g ^ { \prime } ,r ^ { \prime } ,i ^ { \prime } and z ^ { \prime } filters and are public on January 2008 .
These are 21 , 5 and 11 square degrees in the W1 , W3 and W4 fields with photometric data for 1.397.545 ( W1 ) , 366.190 ( W3 ) and 833.504 ( W4 ) galaxies i.e .
for a total of 2.597.239 galaxies .

Methods : We use the photometric redshift code PHOTO-z of Bender et al .
( 2001 ) .

Results : To study the reliability of the photometric redshifts for the CFHTLS broad band filter set we first derive redshifts for the CFHTLS-Deep field D1 , and compare the results to the spectroscopic and photometric redshifts presented in Ilbert et al .
( 2006 ) .
After that we compare our redshifts for the W1 , W3 and W4 fields to about 7500 spectroscopic redshifts from the VVDS therein .
For galaxies with 17.5 \leq i ^ { \prime } _ { AB } \leq 22.5 the accuracies and outlier rates become \sigma _ { \Delta z / ( 1 + z ) } = 0.033 , \eta \sim 2 % for W1 , \sigma _ { \Delta z / ( 1 + z ) } = 0.037 , \eta \sim 2 % for W3 and \sigma _ { \Delta z / ( 1 + z ) } = 0.035 , \eta = \sim 2.5 % outliers for W4 fields . Finally we consider the photometric redshifts of Erben et al .
( 2008 ) which were obtained with exactly the same photometric catalog using the BPZ-redshift code and compare them with our computed redshifts .
For the total galaxy sample with about 9000 spectroscopic redshifts from VVDS , DEEP2 or SDSS we obtain a \sigma _ { \Delta z / ( 1 + z ) } = 0.04 and \eta = 5.7 \% for the PHOTO-z redshifts .
We also merge the subsample with good photometric redshifts from PHOTO-z with that one from BPZ to obtain a sample which then contains ‘ secure ’ redshifts according to both the PHOTO-z and the BPZ codes .
This sample contains about 6100 spectra and the photometric redshift qualities become \sigma _ { \Delta z / ( 1 + z ) } = 0.037 and \eta = 1.0 \% for our PHOTO-z redshifts .

Conclusions : We conclude that this work provides a bias free , low dispersion photometric redshift catalog ( given the depth and filter set of the data ) , that we have criteria at hand to select a ‘ robust ’ subsample with fewer outliers .
Such a subsample is very useful to study the redshift dependent growth of the dark matter fluctuations with weak lensing cosmic shear analyses or to investigate the redshift dependent weak lensing signal behind clusters of galaxies in the framework of dark energy equation of state constraints . The PHOTO-z photometric redshift catalog is provided on request .
Send emails to fabrice @ usm.lmu.de