zCOSMOS Data Release

1 October 2008

• Overview and survey layout • Data format
• Release Content • Acknowledgements
• Release Notes

zCOSMOS (P.I. Simon Lilly) is a large redshift survey that is being undertaken in the COSMOS field using the VIMOS spectrograph mounted at the Melipal Unit Telescope of the VLT at ESO's Cerro Paranal Observatory, Chile. zCOSMOS (ESO Large Programme LP175.A-0839) has been awarded about 600 hours of Service Mode observing time on the ESO VLT.

The zCOSMOS redshift survey has been designed to efficiently utilize VIMOS by splitting the survey into two parts. The first, zCOSMOS-bright, aims to produce a redshift survey of approximately 20,000 I-band selected galaxies at redshifts z < 1.2. Covering the approximately 1.7 deg2 of the COSMOS field (essentially the full ACS-covered area), the transverse dimension at z ~ 1 is 75 Mpc. The second part, zCOSMOS-deep, will observe about 10,000 galaxies selected through well-defined colour selection criteria which mostly lie at 1.5 < z < 3.0. Simply to keep the required amount of telescope time manageable, the field of zCOSMOS-deep is restricted to the central 1 deg2 of the COSMOS field.

This release contains the results of the zCOSMOS-bright spectroscopic observations that were carried out in VLT Service Mode during the period April 2005 to June 2006. 83 masks were observed, yielding 10643 spectra. Future releases will be made as the sample builds up.

The full set of extracted and wavelength calibrated 1-dimensional spectra (FITS and TBL format) are being released. Furthermore, a catalogue is being provided where we give for each target the 1-D spectra filenames, the I-band magnitude used for the selection, as well as the measured redshift and confidence class. A full description of the survey can be found in the accompanying paper "zCOSMOS: A Large VLT/VIMOS Redshift Survey Covering 0 < z < 3 in the COSMOS field", Lilly et al. 2007, ApJS, 172, 70, which will soon be updated by a new data paper.

The data were reduced by the zCOSMOS team.

Content of this release

Summary of reduced VIMOS observations

A total of 10643 spectra could be extracted from the VIMOS observations and are presented in this release. Throughout the zCOSMOS-bright survey, 1 arcsec wide slits have been used with a wavelength range of approximately 5500 to 9700 sampled at roughly 2.5 /pixel.

The primary input catalogue for slit-mask design was generated using SExtractor (Bertin et al. 1996) applied to the COSMOS F814W HST/ACS images sampled at 0.03 arcsec/pixel (Koekemoer et al. 2007, Leauthaud et al. 2007) in a "hot and cold" two-pass process to first identify bright objects. This substantially reduced the tendency of the HST-based catalogue to "over-resolve" extended galaxies into multiple components. This initial SExtractor catalogue was then "cleaned" by carrying out a detailed comparison with one extracted from a stack of i* images obtained with MEGACAM on the 3.6m Canada-France-Hawaii telescope and processed at the TERAPIX data reduction center in Paris. This catalogue was also used to supplement the ACS catalogue for regions where the ACS images were unavailable or unusable. The zCOSMOS-bright target catalogue is intended to be simply defined as having an ACS/HST SExtractor "magauto" brightness in the range 15.00 < IAB(814) < 22.50.

Generally, objects to be inserted into the slit mask are chosen randomly from the target catalogue.   However, a few percent of targets (generally X-ray sources) are designated as "compulsory", indicated by 'i'=2 (column 7, selection flag in the catalog table) targets and inserted into the masks with first priority.   As a result, they are over-represented in the spectroscopic catalogue, by a factor which happens to be close to 2.0.   Objects strongly suspected of being stars on the basis of morphology and spectral energy distribution are not included in the masks as targets and are classificed as "forbidden" , indicated by 'i'=0 (column 7, selection flag in the catalog table). These are about 15% of the I < 22.5 sample.  However, the criteria for this exclusion are deliberately quite conservative and about 4% of the spectroscopic targets turn out to be stars.   

For each slit there is a primary target.   Naturally, sometimes other targets happen to fall in the same slit.   These "secondary targets" are indicated by preceding the redshift confidence class (see below) by a "2". Sometimes, these secondary objects were forbidden, but their spectra were anyway reduced and included in the catalogue.

Given the inevitable complexity of the sample, users who require statistically complete samples are strongly encouraged to contact the zCOSMOS team for guidance.

Release Notes

Data reduction method

The VIMOS observations were reduced using the v1.0 of the VIPGI software package (Scodeggio et al., 2005, Pub.Astr.Soc.Pac., 117, 1284). Determination of redshifts is a multi-step process and involves the use of different approaches tailored to the individual spectra. These include first a computer-aided determination based on cross-correlation with template spectra coupled to continuum fitting and principal component analysis, using the EZ software (Garilli et al., in preparation). This preliminary automated step is followed by a detailed visual examination of the one- and two-dimensional spectrograms of every object to critically assess the validity of the automated redshift. In those cases where the automatic procedure fails, a new redshift is computed based on the wavelengths of recognized features. Two fully independent reductions are carried out of each spectrum, yielding two independent redshift measurements. These are compared and "reconciled" (generally in a face-to-face meeting) to yield a final redshift and Confidence Class.

Redshift accuracy and reliability

About 600 objects have been observed more than once, either from repeat observations of whole masks, repeated inclusion in different masks as primary targets, or because an object is a primary target in one mask and a secondary in another.  From these repeat observations, the typical redshift uncertainty in zCOSMOS-bright is estimated to be (1 sigma) 110 km/s.

Each redshift is assigned to a Confidence Class which captures this crucial information. zCOSMOS Confidence Classes have been developed from those adopted in the CFRS and VVDS surveys. It is important to note that they are based on the confidence in the final redshift and not on the quality of the spectrum per se. The basic confidence scheme ranges from Class 0 (no redshift obtained) to Class 4 (very secure redshift). In practice there is little real difference between Classes 3 and 4 and they may be safely combined for most purposes. Two additional classes with special meaning are then added. Class 9 signifies a one-line redshift where the line is undoubtedly real, as well as being sufficiently strong and isolated that we can be confident that the line is either H-alpha or [OII] 3727 - yielding two possible redshifts. A Class 8 is also a one line redshift, but for Broad Line AGN. For these two cases, we use photo-z to check the line identification (see below) - finding that we guessed right in over 85% of cases - and modifying the redshift to the alternate one if that is consistent with the photo-z.

The basic confidence scheme is then modified with possible preceding digit(s) as follows: An additional "1" digit before the Class (i.e. adding "10" to the Class) signifies a broad line AGN: e.g. Class 13 is a very secure BL AGN redshift. A "2" digit before the Class signifies that the object was a secondary target because it was detected in a slit positioned on another target: e.g. Class 24 is a very secure serendipitous object while Class 213 is a secure serendipitous broad line AGN redshift. Note that all Class 0 and 20 objects (i.e. redshift failures of primary and secondary targets) have been retained in the catalogue.  For the Class 0 objects, these objects were given a "fair trial".  However, the Class 20 failures are not fair since the light could have been substantially reduced through misalignment of the secondary target with the slit.  Indeed the distribution of Confidence Classes for the secondary targets is noticably skewed towards lower confidence redshifts for the same reason.

The statistical reliability of the spectroscopic redshifts in the various Confidence Classes is assessed by both the agreement or otherwise of redshifts independently derived from repeat observations of 600 galaxies (see above), and by the consistency or otherwise with photometric redshifts derived from the COSMOS multi-band photometric data using photo-z for AGN (Salvato et al. 2008), stars and galaxies (Ilbert et al. 2008) and where these are unavailable, from ZEBRA measurements (Feldmann et al 2008).

Confidence Class Description Spectroscopic verification Photo-z consistency within Δz=0.08(1+z)
Class 4 Very secure redshift, beautiful redshift >99.5% 97%
Class 3 Secure redshift >99.5% 97%
Class 9 One line redshift (best guess) 95% after correction 84%-95% after correction with photo-z
Class 2 Probable redshift 92% 93%
Class 1 Insecure redshift 70% 72%

These two approaches show very good agreement, apart from a "ceiling" in the photo-z reliability of about 97% due to difficulties in the photometry (merged objects etc).  This suggests that consistency or otherwise with the photometric redshift can be used to indicate which of the less reliable spectroscopic redshifts are probably right and which are likely to be wrong. We therefore add a decimal place to the Confidence Class to reflect this additional information. The integer part is based solely on the spectrum itself, followed by a decimal place which contains (a posteriori) information on the consistency of the spectroscopic redshift with the photometric redshift as given in the following table.

Confidence class spectroscopic/photometric consistency
.5 Spectroscopic redshift consistent within 0.08(1+z) of the photometric redshift, both for galaxies, stars and AGN.
.4 No photometric redshift available available, includes all spectroscopic AGN and stars
.3 Special case for Class 18 and 9: Consistent with photo-z only after the redshift changed to the alternate redshift, a switch which is then applied
.1 Spectroscopic and photometric redshifts are not consistent at the level of 0.08(1+z)

There is an obvious trade-off between reliability and completeness: the two most reliable Classes 3 and 4, with spectroscopic reliability > 99.5% in the above table, currently comprise about 61% of the zCOSMOS-bright sample. However, many users will likely prefer to utilize the photo-z information and use the extended set of Classes (All 4's, all 3's,  9.5 + 9.4 + 9.3 + 2.5 + 2.4 + 1.5) which is still 99 % reliable and comprises over 80% of the sample.

Data format

For each target the following data files are being released: the 1-dimensional spectrum in FITS and TBL format.

The following naming convention has been adopted for the individual files:

sc_aa_bb.fits 1-d FITS spectrum
sc_aa_bb.tbl 1-d TBL spectrum

where <aa> is the object ID; <bb> indicates the mask ID of the observation (e.g., ZCMRa35_M1), quadrant, slit and object number.

The target catalog zCOSMOS_BRIGHT_v3.5_spec.tbl, containing object ID, target position, redshift, confidence class, target magnitude (F814W), and filenames of 1-d spectra, is being released in TBL format.


When using data products provided in this release, acknowledgement should be given in the text to the zCOSMOS project, referring to this release and the publication Lilly et al., 2007, ApJS, 172, 70 or later version. Please make sure that the following statement appears in your acknowledgements when using these data:

Based on zCOSMOS observations carried out using the Very Large Telescope at the ESO Paranal Observatory under Programme ID: LP175.A-0839.