Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.arcetri.astro.it/science/k20/papers/piii.ps Äàòà èçìåíåíèÿ: Tue Jun 11 17:17:36 2002 Äàòà èíäåêñèðîâàíèÿ: Fri Feb 28 06:57:12 2014 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: http www.eso.org public images list 171

Astronomy & Astrophysics manuscript no.
(will be inserted by hand later)
The K20 survey. III. Photometric and spectroscopic properties
of the sample ?
A. Cimatti 1 , M. Mignoli 2 , E. Daddi 3 , L. Pozzetti 2 , A. Fontana 4 , P. Saracco 5 , F. Poli 6 , A. Renzini 3 , G.
Zamorani 2 , T. Broadhurst 7 , S. Cristiani 8;9 , S. D'Odorico 3 , E. Giallongo 4 , R. Gilmozzi 3 , and N. Menci 4
1 Istituto Nazionale di Astrofisica, Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, I­50125 Firenze, Italy
2 Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Bologna, via Ranzani 1, I­40127, Bologna, Italy
3 European Southern Observatory, Karl­Schwarzschild­Str. 2, D­85748, Garching, Germany
4 Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Roma, via Dell'Osservatorio 2, Monteporzio, Italy
5 Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Brera, via E. Bianchi 46, Merate, Italy
6 Dipartimento di Astronomia, Universit`a ``La Sapienza'', Roma, Italy
7 Racah Institute for Physics, The Hebrew University, Jerusalem, 91904, Israel
8 ST, European Coordinating Facility, Karl­Schwarzschild­Str. 2, D­85748, Garching, Germany
9 Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Trieste, Via Tiepolo 11, I­34131, Trieste, Italy
Received ; accepted
Abstract. The K20 survey is an ESO VLT optical and near­infrared spectroscopic survey aimed at obtaining
spectral information and redshifts of a complete sample of about 550 objects to Ks Ÿ 20:0 over two independent
fields with a total area of 52 arcmin 2 . In this paper we discuss the scientific motivation of such a survey, we describe
the photometric and spectroscopic properties of the sample, and we release the Ks­band photometric catalog.
Extensive simulations showed that the sample is photometrically highly complete to Ks = 20. The observed galaxy
counts and the R \Gamma Ks color distribution are consistent with literature results. We observed spectroscopically 94%
of the sample, reaching a spectroscopic redshift identification completeness of 92% to Ks Ÿ 20:0 for the observed
targets, and of 87% for the whole sample (i.e. counting also the unobserved targets). Deep spectroscopy was
complemented with multi­band deep imaging in order to derive tested and reliable photometric redshifts for
the galaxies lacking spectroscopic redshifts. The results show a very good agreement between the spectroscopic
and the photometric redshifts with ! zspe \Gamma zphot ?= 0:01 and with a dispersion of oe \Deltaz =0.09. Using both the
spectroscopic and the photometric redshifts, we reached an overall redshift completeness of about 98%. The size
of the sample, the redshift completeness, the availability of high quality photometric redshifts and multicolor
spectral energy distributions make the K20 survey database one of the most complete samples available to date
for constraining the currently competing scenarios of galaxy formation and for a variety of other galaxy evolution
studies.
Key words. Galaxies: evolution; Galaxies: formation
1. Introduction
Contrary to surveys for high­z galaxies selected in the op­
tical, which are more sensitive to the star­formation ac­
tivity, the selection of galaxies in the K­band has the
important advantage of not being affected by strong K­
correction effects. This is due to the similarity of the dif­
ferent galaxy spectral types in the near­infrared. Cowie
et al. (1994) showed that over a broad range of redshifts
the K­correction amplitude in the K­band is much smaller
Send offprint requests to: Andrea Cimatti, e­mail:
cimatti@arcetri.astro.it
? Based on observations made at the European Southern
Observatory, Paranal, Chile (ESO LP 164.O­0560).
than those in I­ or in B­ bands, and very little dependent
on the galaxy spectral types (from elliptical to irregular
galaxies). Thus, the great advantage of a K­band selection
is that the resulting samples are almost free from strong
selection effects and do not critically depend on the galaxy
types as in optical samples.
However, deriving the complete redshift distribution of
a sample of faint K­band selected galaxies is a challenging
goal even for 8­10m class telescopes. Indeed, a fraction of
the galaxies are beyond the spectroscopic limits (e.g. very
red galaxies with R ? 26) and/or lie in a redshift range
where the spectra do not present prominent features in the
observer frame (e.g. the redshift ``desert'' at 1:4 ! z ! 2:0
for optical spectroscopy).

2 A. Cimatti et al.: The K20 survey sample
Songaila et al. (1994) carried out an optical spectro­
scopic survey of K­selected galaxies at different magnitude
depths over a wide range of field sizes (from ¸0.4 deg 2
down to K ! ¸ 15, to ¸5 arcmin 2 to K !
¸ 20) reaching
an identification completeness of ¸70% at K ¸ 19 \Gamma 20.
Cowie et al. (1996) performed with the Keck I 10m tele­
scope a deeper optical spectroscopic survey of galaxies
with K ! 20 over a field of 26.2 arcmin 2 , reaching a spec­
troscopic identification completeness of ¸60% at K ¸ 20.
Cohen et al. (1999a, 1999b) observed a sample of 195 ob­
jects over an area of 14.6 arcmin 2 reaching a spectroscopic
completeness of 83.6% to K s ! 20. A large spectroscopic
survey is currently in progress over four independent fields
covering a total area of about 100 arcmin 2 to K ! 20
(Stern et al. 2001).
In addition to purely spectroscopic surveys, a number
of more recent projects made use mostly of photometric
redshifts. For instance, Drory et al. (2001) selected ¸5000
galaxies to K ! 19 over 998 arcmin 2 , and estimated the
photometric redshifts for 94% of them using V RIJK pho­
tometry (spectroscopic redshifts were available for only
¸6% of the sample). Other surveys based largely on pho­
tometric redshifts range from that of Firth et al. (2002),
who covered a field of 744 arcmin 2 to H ! 20:0 \Gamma 20:5
and derived photometric redshifts for about 4000 galaxies
using UBV RIH photometry, to those of e.g. Fontana et
al. (1999,2000) and Rudnick et al. (2001) based on very
deep observations (typically to K s ! 21 \Gamma 23) of very small
fields (a few arcmin 2 ).
Surveys for faint galaxies selected in the K­band are
very important to investigate the formation and evolution
of galaxies.
One of the main questions of galaxy evolution is
whether massive galaxies (e.g. M stars
? ¸ 10 11 M fi ) are
the late product of merging of pre­existing disk galaxies
occurring mostly at z ! 1:5 \Gamma 2, as predicted by the cur­
rent CDM scenarios (e.g. Kauffmann et al. 1993, 1996;
Cole et al. 2000; Baugh et al. 1996; Somerville et al. 2001,
Baugh et al. 2002, and references therein), or whether they
formed at z ? 2 \Gamma 3 during a short­lived and single event
of vigorous star formation, followed by a passive evolution
(or pure luminosity evolution, PLE) of the stellar popula­
tion to nowadays (see e.g. Renzini 1999; Renzini & Cimatti
1999; Peebles 2002, for recent reviews).
Since the near­IR light is a good tracer of a galaxy
stellar mass (Gavazzi et al. 1996; Madau, Pozzetti &
Dickinson 1998; Kauffmann & Charlot 1998, KC98 here­
after), the above scenarios can in principle be tested by
selecting galaxies at 2.2¯m in the K­band and deriving
their redshift distribution and physical and evolutionary
properties (e.g. Broadhurst et al. 1992).
KC98 estimated that ¸ 60% and ¸ 10% of the galax­
ies in a K ! 20 sample are expected to be at z ? 1, re­
spectively, in a PLE and in a standard CDM hierarchical
merging model (cf. their Fig. 4). Such a large difference
was in fact one of the main motivations of our original
project to undertake a complete redshift survey for all ob­
jects with K ! 20 in a small area of the sky. However,
more recent models consistently show that for z ? 1 the
difference between the predictions of different scenarios is
less extreme than in the KC98 realization (Menci et al.,
Pozzetti et al., in preparation; see also Firth et al. 2002).
Part of the effect is due to the now favored \LambdaCDM cos­
mology which pushes most of the merging activity in hier­
archical models at earlier times compared to ÜCDM and
SCDM models, and therefore get closer to the PLE case.
Moreover, a different tuning of the star­formation algo­
rithms (to accomodate for more star formation at high­z)
also reduces the differences between the two scenarios (e.g.
Somerville et al. 2001).
In order to address the question of the formation and
evolution of galaxies and to constrain the current models,
we performed a new spectroscopic survey of K­selected
galaxies with the ESO VLT. In this paper, we present the
main scientific motivations, we define the sample selection,
we discuss the photometry, the completeness, we present
the main spectroscopic and photometric redshift results,
and we release the K s ­band photometric catalog. The
spectral and multicolor catalogs will be presented else­
where together with the forthcoming papers on the scien­
tific analysis of the sample. Magnitudes are given in Vega
system. The widely accepted cosmology with H 0 = 70 km
s \Gamma1 Mpc \Gamma1
,\Omega m = 0:3
and\Omega \Lambda = 0:7 is adopted through­
out this paper.
2. The K20 Survey
In order to overcome the limitations of the previous sur­
veys (e.g. small samples and/or spectroscopic incomplete­
ness), we started in 1999 a project that was dubbed ``K20
survey''. To such project, 17 nights were allocated as an
ESO VLT Large Program distributed over a period of two
years (see also http://www. arcetri.astro.it/¸k20/
for more details).
The survey aims at obtaining spectroscopic redshifts
with the highest possible completeness for a sizeable sam­
ple of field galaxies with K s ! 20. The main scientific aim
of the K20 survey is to probe the evolution of galaxies to z
?
¸ 1, deriving their redshift distribution, luminosity func­
tion and stellar mass function. This body of observational
evidences is being and will be used for a comparison with
the different models of galaxy formation and evolution.
To this purpose, deep optical and near­infrared spec­
troscopy were complemented by deep multi­band imaging
in order to derive tested and reliable photometric redshifts
for the unavoidable fraction of galaxies with no spectro­
scopic redshifts.
Besides such a main goal, the K20 survey was also de­
signed to address other important issues such as: (1) the
nature of Extremely Red Objects (EROs), (2) the evo­
lution of elliptical galaxies, (3) the evolution of galaxy
clustering, (4) the spectral properties of a large num­
ber of galaxies and their evolution as a function of
redshift, (4) the evolution of the volume star forma­
tion density, (5) the fraction of AGN in K­selected

A. Cimatti et al.: The K20 survey sample 3
samples, (6) the optimization of the photometric red­
shift techniques based on ground­based imaging, and (7)
the brown dwarf population at high Galactic latitude.
The K20 database will also provide important informa­
tion to complement other multiwavelength surveys per­
formed in the same fields, such as the SIRTF GOODS
(http://www.stsci. edu/science/goods) and SWIRE
(http://www.ipac.caltech.edu/SWIRE/) projects.
Two papers based on the K20 survey have been al­
ready published showing an application of our database
to investigate the nature and the clustering of the EROs
(Cimatti et al. 2002; Daddi et al. 2002).
The sample was selected from two independent fields
in order to reduce the effects of the field­to­field variations.
The targets were extracted from a sub­area of the Chandra
Deep Field South (CDFS; Giacconi et al. 2001) and from
a field centered around the QSO 0055­269 at z=3.656 (see
Tab. 1 for details). The total sample includes 546 objects
over 52 arcmin 2 selected on the basis of the single crite­
rion of K s Ÿ 20:0. Adopting the K­correction effects de­
rived from the Bruzual & Charlot (1993) spectral synthesis
models (GISSEL version 2000, hereafter BC2000), an ap­
parent magnitude of K s = 20:0 corresponds to rest­frame
typical absolute magnitudes of MKs ¸­21.7 and MKs ¸
­23.4 for z = 0:5 and z = 1:0 respectively, with a weak
dependence on the galaxy type because of the small K­
correction effects. Such absolute magnitudes correspond
to ¸ 0:1L \Lambda
K and ¸ 0:4L \Lambda
K respectively (adopting the L \Lambda
K
of the Cole et al. 2001 near­IR local luminosity function
of galaxies). Since the stellar mass­to­infrared luminosity
ratio (M stars =LK ) is relatively insensitive to the star for­
mation history, the limiting K s ­band luminosities of our
survey can be converted into the corresponding limiting
stellar masses. As the stellar population ages, M stars =LK
(in solar units) remains very close to unity, independent
of the galaxy color and Hubble type, and has a weak de­
pendence on z.
According to the mean stellar mass­to­light ratio and
representative stellar mass M \Lambda
stars in the local universe
(Cole et al. 2001), and adopting the predictions of the
BC2000 spectral synthesis models, the limiting stellar
masses corresponding to K s =20, for the Salpeter IMF,
are about M stars ? 10 10 M fi ' 0:07M \Lambda
stars and M stars ?
4 \Theta 10 10 M fi ' 0:3M \Lambda
stars for z = 0:5 and z = 1 respec­
tively.
3. Near­infrared imaging
For both fields, the K s ­band images were obtained with
the ESO NTT equipped with SOFI (Moorwood et al.
1998) with a pixel size of 0.29 00 and under photometric
conditions (see also Rengelink et al. 1998 for more details
on observations and database for the CDFS).
Since the ESO Imaging Survey (EIS;
http://www.eso.org/science/eis/) public JK s
images of the CDFS available at the time of the K20
sample selection were affected by a loss of the flux up
to 0.3 magnitudes at the faint limit of our survey, we
performed a new independent reduction and calibration
of the CDFS K s ­band image.
The K s ­band data of the CDFS and the JK s data of
the 0055­2659 field were reduced in a standard manner us­
ing the IRAF 1 software package DIMSUM 2 . Raw frames
were first corrected for bias and dark current by subtract­
ing a median dark frame, and flat­fielded using an average
differential sky light flat­field image. After deriving a mas­
ter mask frame with DIMSUM, a sky background image
for each frame was obtained by averaging a set (from 6 to
10) of time adjacent frames where the sources were masked
out. The sky­subtracted frames were then inspected in or­
der to reject or to correct the frames with bad sky residu­
als, and the flux of bright objects visible in the individual
frames was monitored to verify the photometric conditions
and to reject discrepant frames if necessary. The final sky­
subtracted frames were rescaled to the same airmass and
photometric zero­point, shifted and coadded together to
generate the final image. The photometric calibration was
achieved through the observation of standard stars from
the Persson et al. (1998) sample. The Galactic extinction
was estimated using the maps of Schlegel et al. (1998) (see
Tab. 1).
4. Photometry and completeness of the sample
4.1. Source extraction and photometry
The sample was extracted from the K s ­band images using
the SExtractor package (Bertin & Arnouts 1996). After
convolving the images with a Gaussian function match­
ing the measured seeing FWHM, the objects exceeding a
S=N ? 2:2 over the background noise in a minimum area
of 10 pixels (i.e. 0.7oe/pixel) were extracted.
The total flux of the selected objects was then mea­
sured using the SExtractor BEST magnitude. Such a mag­
nitude is defined to be either a Kron magnitude (Kron
1980), which is measured in an elliptical aperture whose
size is determined by the profile of the object, or an isopho­
tal magnitude for objects blended or in crowded fields (see
Bertin & Arnouts 1996 for more details).
In both fields, the images are considerably deeper than
the K s = 20 threshold of the spectroscopic sample, thus
minimizing both selections effects and biases in the esti­
mated magnitudes. Typically, galaxies at the faint limit
of our sample have S=N ' 10 \Gamma 15, according to the
SExtractor noise estimate. The cases of blended objects
were analysed in more detail in order to correct the total
magnitudes for the effects of blending.
1 IRAF is distributed by the National Optical Astronomy
Observatories, which are operated by the Association of
Universities for Research in Astronomy, Inc., under cooper­
ative agreement with the National Science Foundation.
2 Deep Infrared Mosaicing Software, a package written
by Eisenhardt, Dickinson, Stanford and Ward, available at
ftp://iraf.noao.edu/contrib/dimsumV2.

4 A. Cimatti et al.: The K20 survey sample
Table 1. The Ks­band images of the target fields
Field Center Coordinates Field Size Seeing Int. time Gal. Coord. AV AK
(J2000) (arcmin 2 ) (FWHM) (hours) (mag) (mag)
QSO 0055­269 00 h 57 m 58.91 s ­26 ffi 43 0 12.1 00 19.8 0.9 00 3.93 197.7 ffi ­88.5 ffi 0.065 0.007
CDFS 03 h 32 m 22.52 s ­27 ffi 46 0 23.5 00 32.2 0.8 00 3.0 223.5 ffi ­54.5 ffi 0.028 0.003
4.2. Photometric completeness
Since the goal of photometric surveys for faint galaxies is
to measure their total integrated flux, the completeness
is usually defined only by the total apparent magnitude.
However, in practice it is well known that a fraction of
the flux is lost when measuring magnitudes with aper­
ture or isophotal photometry because of the finite size of
the photometric apertures and a minimum area for object
detection (e.g. Impey & Bothun 1997; Dalcanton 1998).
The fraction of the lost light at the survey limiting flux
depends on the size and on the surface brightness profile
of the galaxies. Moreover, since the surface brightness of
galaxies rapidly becomes fainter with increasing redshift
as (1 + z) \Gamma4 , this cosmological dimming could make many
high­z galaxies undetectable below the adopted surface
brightness threshold.
Thus, it is important to evaluate such potential se­
lection effects in order to make a meaningful comparison
between the observed properties of the sample and the
model predictions of galaxy formation and evolution. We
have therefore explored in our survey the effects due to
the detection limits, i.e. surface brightness limit (¯K ¸ 23
mag/arcsec 2 ) and minimum area (10 pixels), and to the
procedure adopted for the photometric measurements, us­
ing both real and artificial galaxies.
4.2.1. Limiting surface brightness
Following Lilly et al. (1995), we show in Figure 1 the
surface brightness at the minimum detection radius, as
a function of magnitude for all the objects in our survey.
The vertical line marks the magnitude limit, K s ! 20,
while the horizontal lines show the surface brightness lim­
its used in the objects detection for the two fields (i.e. a
source is detected only if its detection surface brightness
is above the surface brightness limit). Most of the objects
have surface brightness well above the nominal limit.
Figure 1 also shows the tracks expected for L \Lambda and
10L \Lambda elliptical and spiral galaxies. These tracks have been
obtained using the observed relations between luminos­
ity, effective radius and surface brightness for local spi­
rals (Impey et al. 1996) and elliptical (Bender et al. 1992,
Pahre 1999) galaxies, and taking into account the the
effect of the seeing on the intrinsic galaxy profile (see
Angeretti, Pozzetti & Zamorani 2002 for details). The
adopted galaxy profile parameters are indicated in the
Figure caption. Note that at the faint limit of the sam­
ple, the L \Lambda tracks are essentially parallel to the star locus
because for these galaxies the observed profiles are domi­
nated by the seeing. Fig. 1 shows that the surface bright­
ness selection effects are negligible for L ! 10L \Lambda galaxies.
Fig. 1. Surface brightness at the minimum detection radius as
a function of magnitude for all objects in the sample. The verti­
cal dotted line represents the sample magnitude limit, Ks ! 20,
while the horizontal dotted lines indicate the limiting detec­
tion surface brightness in the two fields (¯K ¸ 22:64; 23:11
mag/arcsec 2 in the CDFS and 0055­269 field respectively). The
tracks indicate the expected behavior of elliptical (thick lines)
and spiral (thin lines) galaxies for two representative luminosi­
ties (Kochanek et al. 2001): L \Lambda (solid lines: M \Lambda
K = \Gamma24:3 +5log
h70 , re = 3:1 h \Gamma1
70 kpc for ellipticals and M \Lambda
K = \Gamma23:8 + 5log
h70 , re = 6:1 h \Gamma1
70 kpc for spirals) and 10 L \Lambda (dashed lines:
M \Lambda
K = \Gamma26:8 + 5log h70 , re = 16:3 h \Gamma1
70 for ellipticals and
M \Lambda
K = \Gamma26:3 + 5logh70 , re = 13:0 h \Gamma1
70 for spirals). Star sym­
bols represent the stars identified in our survey on the basis of
their spectra.

A. Cimatti et al.: The K20 survey sample 5
Fig. 2. The results of a simulation in the 0055­269 field
made with artificial galaxies with de Vaucouleurs profiles,
b=a=0.7, four values of re and for 6 input magnitudes
(Ks=18.0,18.5,19.0, 19.5,20.0,20.5) (see text for more details).
The top and the bottom panels show respectively the aver­
age difference between the measured (Ks) and the input mag­
nitudes, and the fraction of the number of recovered objects,
N(r), to the number of input artificial objects, N(i). The points
are shifted along the abscissa around each of the 6 magnitudes
to improve their visibility. Similar results are obtained for the
CDFS.
4.2.2. Adding artificial galaxies
In order to better assess the effects for different galaxy
types, we applied a method consisting in the addition of
artificial galaxies to the real images of our survey fields.
About 100 artificial galaxies were generated using the
artdata.mkobjects software in IRAF. We used three
morphological types (one de Vaucouleurs profile with axial
ratio b=a=0.7, and two exponential profiles with b=a=0.4
and b=a=0.8 respectively), and four intrinsic half­light
radii for each morphology type (r e =0.3 00 ,0.6 00 ,1.0 00 ,1.5 00 ).
The choice of such values for r e was motivated by the
quantitative morphological studies of faint galaxies which
showed that objects with r e
?
¸ 1:0 00 are extremely rare
(e.g. Marleau & Simard 1998).
Such 12 artificial galaxy models were then con­
volved with a PSF with a Moffat profile (as derived
from field stars) and created for different magnitudes
(K s = 18:0; 18:5; 19:0; 19:5; 20:0; 20:5). In both fields, the
results indicate that at K s = 20:0 the completeness
(N recovered =N input ) is 100% for r e Ÿ 1:0 00 , whereas it de­
creases to ¸90% for r e = 1:5 00 . We also verified that the
results do not significantly change for other values of b=a.
Fig. 2 and Fig. 3 show examples of the results of such
simulation. The recovered magnitudes at K s Ÿ 20:0 are
Fig. 3. Same as Fig. 2, but for a simulation in the 0055­269 field
made with exponential profiles and b=a=0.4 Similar results are
obtained for the CDFS.
generally consistent with the input ones for the exponen­
tial profile galaxies, but they turned out to be systemat­
ically underestimated for galaxies with a de Vaucouleurs
profile (typically 0.1­0.3 magnitudes for 0:3 00 Ÿ r e Ÿ 1:0 00 ).
Such an effect was already discussed for example by
Fasano et al. (1998), Dalcanton (1998) and Martini (2001).
The origin of such an effect is due to the magnitudes be­
ing limited to an aperture which exclude the outermost
regions of the galaxy surface brightness profiles, and it
does not depend on the software used for the automatic
photometry.
While this is less critical for rapidly declining exponen­
tial profiles, its influence is larger for the more slowly de­
creasing de Vaucouleurs profiles. Being typically r e
! ¸ 0:6 00
in high­z early­type galaxies (e.g. Fasano et al. 1998), this
means that the underestimate in our sample is at most
!
¸ 0.2 magnitudes at K s = 20 for galaxies with a de
Vaucouleurs profile. The lack of HST imaging for the K20
sample prevented us from applying magnitude corrections
for the galaxies with de Vaucouleurs profiles.
Figure 4 shows the same results as a function of the
luminosity (adopting the luminosity­surface brigthness re­
lations discussed in section 4.2.1). For typical L \Lambda ellipti­
cals we could miss systematically up to 20% of the flux,
and only !10­15% for exponential disks. Such effects be­
come more relevant only for very large and high luminosity
galaxies (e.g. L ¸ 10L \Lambda ), but fortunately such objects are
rare enough not to introduce severe selection effects in our
survey.
Such selection effects will be taken into account in the
next papers including the comparison with the predictions
of galaxy formation models.

6 A. Cimatti et al.: The K20 survey sample
Fig. 4. A comparison of BEST magnitude from SExtractor
and input magnitudes in our simulations for ellipticals (top
panel) and spirals (bottom panel) in the CDFS. Different lines
represent different luminosities, as indicated in the labels, and
the error bars represent the dispersion in the simulations, using
about 100 artificial galaxies at each magnitudes.
4.2.3. Dimming real galaxies
As an additional test on the completeness, we applied the
method described by Saracco et al. (1999, 2001), consist­
ing in dimming observed bright galaxies by various factors,
while keeping the background noise constant and equal to
that of the original image. In particular, all the objects
were extracted from the original frame, kept with the same
original sizes, scaled to a lower flux and then re­inserted in
the real images. A new catalog has then been obtained on
these ``dimmed'' images with the same parameters used for
the original one. This method tests the whole photometric
procedure using a set of objects with clustering and with
a wide variety of morphological properties representative
of the real galaxy population.
These simulations have shown that: a) the flux frac­
tion lost by the BEST magnitudes is negligible to about
K s = 19:5, whereas a slight underestimate of 0.05­0.1
magnitudes is present in the fainter side of the sample
19:5 ! K s ! 20:0; b) 100% of the objects at K s = 20 are
robustly detected in both fields, irrespective of their mor­
phology, although c) the scatter between the input and
the recovered magnitudes is typically 0.1 magnitudes at
K s ? 19:5. As a result, d) the number counts observed
in both fields are not affected by significant instrumental
incompleteness.
Table 2. The differential galaxy counts
Ks n0055\Gamma269 nCDFS n tot N tot
13.75 0 +1:84
\Gamma0:00 1 +2:30
\Gamma0:83 1 +2:30
\Gamma0:83 69.2 +159:2
\Gamma57:5
14.25 0 +1:84
\Gamma0:00 1 +2:30
\Gamma0:83 1 +2:30
\Gamma0:83 69.2 +159:2
\Gamma57:5
14.75 1 +2:30
\Gamma0:83 0 +1:84
\Gamma0:00 1 +2:30
\Gamma0:83 69.2 +159:2
\Gamma57:5
15.25 0 +1:84
\Gamma0:00 1 +2:30
\Gamma0:83 1 +2:30
\Gamma0:83 69.2 +159:2
\Gamma57:5
15.75 2 +2:64
\Gamma1:29 3 +2:92
\Gamma2:37 5 +3:38
\Gamma2:16 346.1 +234:0
\Gamma149:5
16.25 1 +2:30
\Gamma0:83 5 +3:38
\Gamma2:16 6 +3:58
\Gamma2:38 415.4 +247:8
\Gamma164:8
16.75 3 +2:92
\Gamma2:37 5 +3:38
\Gamma2:16 8 +3:95
\Gamma2:77 553.8 +273:5
\Gamma191:8
17.25 7 +3:77
\Gamma2:58 17 +5:20
\Gamma4:08 24 +5:97
\Gamma4:86 1661.5 +413:3
\Gamma336:5
17.75 15 +4:96
\Gamma3:83 22 +5:76
\Gamma4:65 37 +7:14
\Gamma6:06 2561.5 +494:3
\Gamma419:5
18.25 28 +6:35
\Gamma5:26 35 +6:97
\Gamma5:89 63 +8:00
\Gamma7:91 4361.5 +553:8
\Gamma547:6
18.75 29 +6:45
\Gamma5:35 59 +7:75
\Gamma7:66 88 +9:43
\Gamma9:36 6092.3 +652:8
\Gamma648:0
19.25 56 +7:54
\Gamma7:46 69 +8:36
\Gamma8:28 125 +11:18
\Gamma11:18 8653.8 +774:0
\Gamma774:0
19.75 43 +7:61
\Gamma6:53 86 +9:33
\Gamma9:25 129 +11:36
\Gamma11:36 8930.7 +786:5
\Gamma786:5
Ks : bin central magnitude;
n0055 : number of galaxies in the 0055­269 field with 1oe poisso­
nian uncertainties;
nCDFS : number of galaxies in the CDFS with 1oe poissonian
uncertainties;
n tot : total number of galaxies with 1oe poissonian uncertainties;
N tot : gal/deg 2 /0.5 mag (both fields) with 1oe poissonian uncer­
tainties;
4.3. Galaxy counts and colors
Tab. 2 lists the galaxy differential counts in 0.5 magnitude
bins in our survey, and Fig. 5 shows a comparison with a
compilation of literature counts. No corrections for incom­
pleteness were applied to our data, and we excluded the
AGNs and the stars with spectroscopic identification in
our sample. The 1oe poissonian uncertainties of the counts
were estimated according to the prescriptions of Gehrels
(1986). Our counts are in very good agreement with those
of previous surveys.
On the basis of what discussed in previous sections, the
slight flattening shown by our data for K s ? 19:5 should
not be due to incompleteness, but is likely to be due to
field­to­field fluctuations which, as Figure 5 shows, are the
dominant source of scattering in the counts from survey to
survey. This is supported by the fact that the small drop
in our total counts to K s ? 19:5 is seen in only one of our
two fields (the 0055­269 field) (see Fig. 5).
The detailed color properties of the galaxies in the K20
sample will be discussed elsewhere. Here, as an additional
check of the photometry, we show a comparison between
the R \Gamma K s galaxy colors of the galaxies in the K20 sample
(the same color used to select the K20 ERO sample dis­
cussed in Cimatti et al. 2002 and Daddi et al. 2002) and
the sample taken from the Caltech Faint Galaxy Redshift
Survey (CFGRS; Cohen et al. 1999a, 1999b), a spectro­
scopic survey of a complete sample of K s ! 20:0 galaxies
over a field of 14.6 arcmin 2 .
Fig. 6 shows a comparison between the R \Gamma K s color
distributions. As suggested by Fig. 6 (top panel), such a

A. Cimatti et al.: The K20 survey sample 7
Fig. 5. Differential Ks­band galaxy counts. The filled circles
represent the K20 survey counts. Literature survey counts are
shown with different symbols. The insert plot shows the counts
separately for the CDFS and the 0055­269 field, and the contin­
uous line represents the average Ks­band galaxy counts taken
from Hall et al. (1998).
comparison may be biased by the presence of three clus­
ters (or rich groups) of galaxies present in both surveys:
one at z = 0:67 and one at z = 0:73 in the 0055­269
and in the CDFS respectively, and one at z = 0:58 in the
CFGRS. To investigate this possibility, we excluded the
members of the richest clusters in both samples (Fig. 6
bottom panel). As Figure 6 (bottom panel) clearly sug­
gests, a Kolmogorov­Smirnov test indicates that the hy­
pothesis that the two distributions are drawn from the
same population cannot be rejected at a confidence level
of 0.05, thus demonstrating that the difference (Fig. 6, top
panel) was due to color biases introduced by the galaxy
clusters (or groups).
5. Optical spectroscopy
Optical multi­object spectroscopy was obtained with the
ESO VLT UT1 and UT2 equipped respectively with
FORS1 (with MOS mode: 19 movable slitlets) and FORS2
(with MXU mode: up to 52 targets observed through a
laser­cut slit mask) during 0.5 00 ­2.0 00 seeing conditions and
with 0.7 00 ­1.2 00 wide slits depending on the seeing. The
Fig. 6. Top panel: the K20 and the CFGRS (Cohen et al.
1999a, 1999b) survey R \Gamma Ks color distributions for Ks ! 20:0
galaxies (continuous and dotted lines respectively). Bottom
panel: the comparison between the K20 and the CFGRS R \GammaK s
color distributions after removing the richest clusters in the two
samples.
grisms 150I, 200I, 300I were used, providing dispersions
of 5.5, 3.9, 2.6 š A/pixel and spectral resolutions of R=260,
380, 660 respectively. The integration times ranged from
0.5 to 3 hours. Dithering of the targets along the slits be­
tween two fixed positions was applied as much as possible
for the faint target observations in order to efficiently re­
move the CCD fringing and the strong OH sky lines at
– obs ? 7000 š A.
The data reduction was carried out using IRAF pack­
ages. The bias subtraction, overscan correction and flat­
fielding were performed in the standard mode. For the
non­dithered spectra, the sky background was removed
from the two­dimensional frames obtained by fitting a
third­order polynomial along the spatial direction and
avoiding the regions with object spectra. The resulting
sky­subtracted images were kept for following analysis in
the course of the redshift determination. The 1D spectra
were then optimally extracted and wavelength calibrated
using exposures of He­Hg­Ar lamps, taken with the same
instrumental configuration. Finally, the spectra were flux­
calibrated using standard spectrophotometric stars, de­
reddened for atmospheric extinction, corrected for telluric
absorptions and scaled to the total R­band magnitudes.
The spectroscopic analysis of the optical spectra and
the redshift estimates were done using IRAF tasks, both
automatic (xcsao) and interactive ones (rvidlines and
splot), and always through visual inspection of the 1D
and 2D spectra.

8 A. Cimatti et al.: The K20 survey sample
We followed different approaches in the redshift mea­
surements, according to the quality and type of the ana­
lyzed spectrum. In case of a spectrum with emission lines,
the redshift was obtained with rvidlines by means of
multiple Gaussian fitting of the spectral features. For the
objects with fairly good signal--to--noise ratio and with an
absorption line dominated spectrum, we performed an au­
tomatic cross­correlation with xcsao. As templates for the
cross­correlation we used the set of Kinney et al. (1996)
template spectra and also high S=N composite spectra
obtained from our own K20 survey database. For objects
with both emission and absorption lines, we used both
rvidlines and xcsao. For spectra with poor S=N ratio,
with a single emission line and/or uncertain spectral fea­
tures, the redshift was estimated both by checking the re­
ality of the structures on the 2D frames and interactively
measuring the wavelengths of the features using splot
and cross­checking the results with rvidlines. Several
and independent cross­checks on ambiguous cases were
done independently by 2­3 of us in order to ensure in­
dependent redshift estimates and to eliminate discrepant
cases. Objects with multiple observations made in differ­
ent mask configurations or different runs provided always
consistent results in terms of detected features and red­
shift measurements, and their spectra were then coadded
to increase the final signal­to­noise ratio.
The spectroscopic redshifts were divided into two cat­
egories: ``high­quality and secure'' redshifts (when several
features were detected at high significance), and ``lower­
quality'' redshifts (when only one weak emission line was
detected and ascribed to [OII]–3727 emission, or when the
detected spectral features were more marginal). Objects
with ``lower­quality'' redshifts are 5% of the total sam­
ple to K s Ÿ 20:0, whereas their fractions decrease to 4%
and 2% for K s Ÿ 19:5 and K s Ÿ 19:0 respectively. One
of the following spectral classes was preliminarly assigned
to each object: star, Type 1 (i.e. with broad lines) AGN,
emission line galaxy, early­type galaxy with emission lines,
early­type galaxy with no emission lines. Fig. 7 shows ex­
amples of the different spectral types. A more detailed
spectral classification is under way and will be presented
together with the K20 spectroscopic catalog.
6. Near­infrared spectroscopy
A small fraction (22 objects, 3.9%) of the K20 sample
was observed with near­IR spectroscopy using the VLT
UT1+ISAAC (Moorwood et al. 1999) in order to derive
the redshifts of the galaxies which were too faint for optical
spectroscopy and/or expected to be in a redshift range
difficult for optical spectroscopy.
During the first ISAAC run (Table 3), targets in the
CDFS were selected according to the photometric red­
shifts available at that time in order to search mostly for
Hff emission redshifted in the H­band for 1:3 ! z ! 1:8.
The observations were carried out under photometric and
0.5 00 ­1.5 00 seeing conditions. The low spectral resolution
mode (R=500 with 1 00 slit) was adopted in order to max­
3000 3500 4000
0
2
4
6
3000 3500 4000
0
1
2
3
2500 3000 3500
0
2
4
1500 2000 2500 3000
0
0.5
1
1.5
2
FeII FeII MgII
Fig. 7. Examples of high­z galaxies with the four adopted
spectral classifications. From top to bottom: an early­type at
z = 1:096 (R = 23:0), an early­type + [OII]–3727 emission
at z = 0:735 (R = 23), an emission line galaxy at z = 1:367
(R = 23:0), an absorption line galaxy at z = 1:725 (R = 23:5).
Some typical absorption and emission lines are also indicated
for each spectrum.
imize the covered spectral range per integration, the slit
was 1 00 wide, providing a spectral resolution FWHM of
32 š A in H­band. The observations were done by nod­
ding the target(s) along the slit between two positions A
and B with a typical nod throw of 10­20 00 . The integra­
tion time was 10 minutes per position. For instance, a
total integration time of 1 hour was obtained following a
pattern ABBAAB. Whenever possible, two targets were
observed simultaneously in the slit in order to maximize
the multiplex. The spectral frames were flat­fielded, recti­
fied, sky­subtracted, coadded and divided by the response
curve obtained using the spectra of O stars. Absolute flux
calibration was achieved by normalizing the spectra to
the JK s broad­band photometry or to the interpolated
H­band magnitude in case of H­band spectra.
During the second ISAAC run (Tab. 3), the targets
were selected from the CDFS and the 0055­269 fields
whenever their spectra did not show features in the optical
and if the photometric redshifts suggested z ? 1:3. The
observations were carried out as for the first run. Despite
a large fraction of the time lost because of bad weather
and poor seeing, such a run was more successful in iden­
tifying a few high­z star­forming galaxies thanks to their
Hff emission.
The bottom line for the ISAAC runs is that the ab­
sence of a multi­object spectroscopic mode, being each
integration limited to a single band (J or H or K), and
the lack of very accurate photometric redshifts at the time

A. Cimatti et al.: The K20 survey sample 9
Table 3. The VLT spectroscopic runs
Dates Telescope Instrument Mode Multiplex Seeing range Run completion
1999 01­05.10 UT1 FORS1 MOS Ÿ19 1.0 00 ­2.5 00 50%
1999 06­09.11 UT1 FORS1 MOS Ÿ19 0.4 00 ­1.5 00 100%
1999 28­30.11 UT1 ISAAC LR Ÿ2 0.5 00 ­1.5 00 100%
2000 24­28.11 UT2 FORS2 MXU Ÿ52 0.4 00 ­1.5 00 100%
2000 04­08.12 UT1 ISAAC LR Ÿ2 0.5 00 ­2.0 00 50%
of the observations made it difficult to derive a substantial
number of spectroscopic redshifts based on near­IR spec­
troscopy alone. All in all, redshifts were derived for four
galaxies at 1:3 ! z ! 1:9 out of the 22 observed (see Fig.
8 for two examples of successful ISAAC observations).
Fig. 8. ISAAC H­band low resolution spectra of two emission
line galaxies with Hff redshifted at z = 1:729 (top panel) and
z = 1:715 (bottom panel).
7. Spectroscopic completeness
Table 4 summarizes the status of the sample and of the
spectroscopic redshift identifications.
Only about 6% of the sample targets to K s = 20 were
not observed either because it was not possible to fit them
in the observing masks or because too faint in the optical
bands or because of the nights lost due to bad weather or
bad seeing conditions.
The efficiencies in deriving the spectroscopic redshifts
(N identified =N observed ) was 96%, 91% and 85% for K s Ÿ
19:0, 19:0 ! K s Ÿ 19:5 and 19:5 ! K s Ÿ 20:0 respec­
tively, whereas the overall spectroscopic redshift complete­
ness (N identified =N total ) in the same magnitude ranges is
94%, 85% and 71%.
The spectroscopic completeness of the total K s Ÿ 20:0
sample is 87% and is one of the highest reached for samples
of faint K­selected galaxies.
8. Optical imaging and photometric redshifts
The availability of high­quality and deep imaging (besides
the K s images) and of spectroscopic redshifts covering
also the critical region of 1:3 ! z ! 1:9 allowed us to
optimize the estimate of the photometric redshifts in a
self­consistent way, and to obtain the most reliable infor­
mation possible on the redshifts of the targets with no
spectroscopic z or with no observations available.
Although the final multicolor data set is not homoge­
neous (in terms of bandwidths and depth) over the two
fields, the strategy followed in both fields is identical,
aimed at assuring that the UV and optical coverage and
depth are adequate to follow the spectral shape of the red­
dest galaxies in the sample. In practice, we have imaged
the 0055­269 field over 10 bands (UBGV RRw IzJK s ), ob­
tained with the ESO NTT + SUSI2 (UBGV Rw I) and
SOFI (J), and VLT + FORS1 (R and z), for a total
of about 45 hours of integration. In the CDFS field, we
used a combination of public EIS NTT data (U) and deep
FORS1 images (BV RIz) obtained in the framework of a
different program (courtesy of P. Rosati & M. Nonino).
Unfortunately, the FORS1 images do not overlap com­
pletely with the area surveyed here, so that multicolor
imaging and photometric redshifts are available for a sub­
set of 272 over 348 objects in the CDFS. In addition, and
at variance with the 0055­269 field, the multicolor cata­
log of the CDFS is not in its final form, since we plan to
eventually include the forthcoming J EIS public images,
but the current (UBV RIzK) data set already provides a
highly satisfactory solution for photometric redshifts.
For the purpose of estimating photometric redshifts
and deriving the Spectral Energy Distributions (SEDs)
of the target galaxies, we developed a technique to ob­
tain a seeing--corrected ``optimal'' aperture photometry to
cope with the inhomogeneous image quality over the dif­
ferent bands. Photometric redshifts have been obtained
through template fitting (see Fontana et al 2000 et al. for
the details), using in this case the PEGASE 2.0 template
spectral package (Fioc & Rocca--Volmerange 1999).

10 A. Cimatti et al.: The K20 survey sample
Table 4. Summary of the spectroscopic results
Magnitude Field N total N observed N identified
N identified
N total
N identified
N observed
N galaxies NAGN Nstars
Ks Ÿ19.0 0055­269 97 95 92 0.95 0.97 81 4 7
CDFS 180 176 168 0.93 0.95 137 5 26
TOTAL 277 271 260 0.94 0.96 218 9 33
19:0 ! Ks Ÿ19.5 0055­269 58 56 49 0.84 0.88 47 0 2
CDFS 75 73 69 0.92 0.95 63 2 4
TOTAL 133 129 118 0.85 0.91 118 2 6
19:5 ! Ks Ÿ20.0 0055­269 43 31 26 0.60 0.84 26 0 0
CDFS 93 82 70 0.75 0.85 63 1 6
TOTAL 136 113 96 0.71 0.85 89 1 6
Ks Ÿ20.0 0055­269 198 182 167 0.84 0.92 154 4 9
CDFS 348 331 307 0.88 0.93 263 8 36
TOTAL 546 513 474 0.87 0.92 417 12 45
The whole procedure has been tested against the sam­
ple with secure spectroscopic redshift, and the results are
shown in Fig. 9­10 The comparison between spectroscopic
and photometric redshifts shows that it is possible to
achieve a high reliability and accuracy even with ground--
based data sets, provided that the imaging data are of
adequate quality.
The results in the two fields are slightly different, be­
cause of the different quality of the imaging data set.
To compute a suitable statistics on z spe \Gamma z phot , one
has to define a recipe to account for the non­gaussian
shape of its distribution function. The one we adopt here
is to compute mean values and standard deviations of
z spe \Gamma z phot on all the objects with a fractional error
\Deltaz = (z spe \Gamma z phot )=(1 + z spe ) smaller (in modulus) than
0:15. This choice is based on the fact that the fractional er­
ror \Deltaz is roughly constant with redshift with a oe--clipped
dispersion of about 0.05 in both these samples and in the
HDFN (Cohen et al. 2000, Fontana et al 2002). The cor­
responding selection is delimited by the two dashed lines
in Fig. 9. Objects outside this range can be described as
''outliers''. As it emerges from Fig. 9, most of the objects
lie within the boundaries that we use to define the ``sat­
isfactory'' fits, and most of the outliers are close to the
same boundaries.
In the 0055­269 field, the number of outliers is fairly
small (7 over 140), and the r.m.s. dispersion of the remain­
ing fraction in the range 0 ! z ! 2 is oe \Deltaz = 0:078, with
a median value ! \Deltaz ?= \Gamma0:004, very close to the val­
ues obtained in the HDF--N over the same redshift range
and with the same template set (Fontana et al 2000). In
the case of the CDFS, we reached a final accuracy of
oe \Deltaz = 0:095 with a median value ! \Deltaz ?= 0:02, with
a fraction of discrepant objects of 13 over the 198 galaxies
with secure redshifts used in the comparison.
It is also important to notice that objects with ``lower
quality'' spectroscopic redshifts (see section 5) generally
follow the relation between photometric and spectroscopic
redshifts (Fig. 9).
In both cases, the high accuracy and the low level
of misidentified objects allow us to use the photomet­
ric redshift estimate both to support the redshift iden­
tification of objects with uncertain spectroscopic red­
shift, and to adopt the photometric redshift for the un­
observed/unidentified objects.

A. Cimatti et al.: The K20 survey sample 11
The dispersion on the global sample (CDFS + 0055­
269 fields) is oe = 0:089 and the average is ! z spe \Gamma
z phot ?=0.012.
In the whole K20 database, there are only 9 objects (in
the CDFS) for which neither z spe nor z phot are available.
Thus, the overall redshift completeness reached for our
sample by using both z spec and z phot is 537/546=98.3%.
0 1 2 3
0
1
2
3
Fig. 9. The comparison between spectroscopic (x--axis) and
photometric (y--axis) redshifts of the galaxies in the K20 sur­
vey: 140 galaxies with secure redshifts in the 0055­269 field in­
dicated by filled circles and 198 galaxies in the CDFS indicated
by filled squares. The solid line shows the zspe = zphot relation.
The two dashed line delimit the region where jzspe \Gamma zphot j Ÿ
0:15(1+zspe),that has been used to compute the overall statis­
tics in the sample (see text for details). The open symbols
indicate objects with ``lower quality'' spectroscopic redshifts.
9. Release of the K s
­band photometric catalog
We make the K s ­band photomet­
ric sample available at the web site
http://www.arcetri.astro.it/¸k20/releases/.
This version of the sample includes for the moment
only the names of the targets, the coordinates, the total
(BEST) K s ­band magnitudes and their photometric
uncertainties. An extract of such catalog is shown in
Table 5. In the next papers, we will also release the full
spectroscopic and multicolor catalogs.
10. Summary
We discussed the general scientific background and aims of
the K20 survey and we described the sample photometric
and spectroscopic properties.
­1 ­0.5 0 0.5 1
0
10
20
30
40
Fig. 10. The histogram of the observed scatter zspe \Gamma zphot for
all the galaxies of the sample with spectroscopic redshift.
The relevant advantages of the K20 database are the
small K­correction effects due to the K s ­band selection, its
size (about 500 galaxies), the distribution of the targets in
two independent fields, the use of near­IR spectroscopy for
a subsample of the targets, and the availability of a large
deep imaging database from the optical to the near­IR.
Extensive simulations show that the sample is highly
complete to K s = 20:0 and not affected by strong selection
effects on the galaxy population, with the possible excep­
tion of a slight underestimate of the total flux for large and
luminous early­type galaxies with de Vaucouleurs surface
brightness profiles. Such a selection effect can be taken
into account when comparing the observed K20 sample
with model predictions. The observed galaxy counts and
the R \Gamma K s color distribution are in agreement with liter­
ature results.
Optical spectroscopy, aided by near­IR spectroscopy,
allowed us to achieve a high redshift completeness for a
sample of galaxies selected in the near­IR (94% to K s !
19:0 and 87% K s ! 20:0), and to obtain spectroscopic
redshifts in the ``desert'' of 1:4 ! z ! 2:2.
The high­quality deep imaging database allowed us
also to obtain tested and reliable photometric redshifts for
the unobserved or spectroscopically unidentified galaxies.
Using the photometric redshifts, the global completeness
(z spec + z phot ) increases to 98.3% of the total sample.
Such a rich photometric and spectroscopic database
makes the K20 sample a key tool to investigate the for­
mation and evolution of galaxies.
Acknowledgements. We thank the VLT support astronomers
for their kind assistance and competent support during the
observations. AC warmly thanks ESO (Garching) for the hos­
pitality during his visits. We thank J. Cohen for providing

12 A. Cimatti et al.: The K20 survey sample
Table 5. Extract of the Ks­band photometric catalog of the 0055­269 field
ID Right Ascension (J2000) Declination (J2000) Ks oe(Ks)
00013 00:57:55.27 ­26:40:58.3 18.83 0.03
00017 00:57:58.24 ­26:40:59.6 19.68 0.07
00018 00:58:02.90 ­26:41:01.5 16.16 0.01
00019 00:58:06.68 ­26:40:59.7 18.98 0.04
00021 00:57:51.54 ­26:41:03.6 18.54 0.03
00022 00:57:59.53 ­26:41:04.6 19.19 0.05
00023 00:57:56.82 ­26:41:05.2 19.87 0.08
00024 00:57:54.74 ­26:41:07.9 19.94 0.06
00025 00:57:54.52 ­26:41:09.7 18.46 0.02
..... ........... ........... ..... ....
the CFGRS sample in digital form. We warmly thank Piero
Rosati and Mario Nonino for providing the reduced and cali­
brated BV RI FORS1 images of the CDFS. We also thank the
anonymous referee for the useful comments. The imaging ob­
servations of the 0055­269 field were performed during SUSI2
guaranteed time of the Observatory of Rome in the framework
of the ESO­Rome Observatory agreement for this instrument.
References
Angeretti L., Pozzetti L., Zamorani G., 2002, in ''A new era
in cosmology'', Durham, UK, ASP Conference Series, ed.
N.Metcalfe & T.Shanks, in press
Baugh C.M., Cole S., Frenk C.S., 1996, MNRAS 283, 1361
Baugh C.M., Benson A.J., Cole S., Frenk C.S., Lacey C.,
2002, in `The Mass of Galaxies at Low and High Redshift',
Venice 2001, eds. R. Bender, A. Renzini, in press, astro­
ph/0203051
Bender R., Burstein D., Faber S.M., 1992, ApJ, 399, 462
Bertin E., Arnouts S. 1996, A&A, 117, 393
Broadhurst T.J., Ellis R.S., Glazebrook K. 1992, Nature, 355,
55
Bruzual G., Charlot S. 1993, 405, 538
Cimatti A. et al. 2002, A&A, 381, L68
Cohen J.G., Blandford R., Hogg D.W., Pahre M.A., Shopbell
P. 1999a, ApJ, 512, 30
Cohen J.G., Hogg D.W., Pahre M.A., Blandford R., Shopbell
P., Richberg K. 1999b, ApJS, 120, 171
Cohen J.G., Hogg D.W., Blandford R., Cowie L.L., Hu E.M.,
Songaila A., Shopbell P., Richberg K. 2000, ApJ, 538, 29
Cole S., Lacey C.G., Baugh C.M., Frenk C.S. 2000, MNRAS,
319, 168
Cole S. et al. 2001, MNRAS, 326, 255
Cowie, L.L., Gardner J.P., Hu E.M., Songaila A., Hoddapp
K.­W., Wainscoat R.J. 1994, ApJ, 434, 114
Cowie, L.L., Songaila A., Hu E.M., Cohen J.G. 1996,
AJ,112,839
Daddi E., Cimatti A., Pozzetti L., et al., 2000, A&A 361, 535
Daddi E., Broadhurst T., Zamorani G., Cimatti A., R¨ottgering
H.J.A., Renzini A., 2001, A&A, 376,825
Daddi E. et al. 2002, A&A, 384, L1
Dalcanton J.J., 1998, ApJ, 495, 251
Djorgovski, S. et al., 1995, ApJ, 438, L13
Drory N., Bender R., Snigula J., Feulner G., Hopp U.,
Maraston C., Hill G.J., de Oliveira C.M. 2001, ApJ, 562,
L111
Fasano G., Cristiani S., Arnouts S., Filippi M. 1998, AJ, 115,
1400
Firth A.E., Somerville R.S., McMahon R.G. et al. 2001,
MNRAS, submitted (astro­ph/0108182)
Fontana A., Menci N., D'Odorico S., Giallongo E., Poli F.,
Crisitiani S., Moorwood A., Saracco P. 1999, MNRAS, 310,
L27
Fontana A. et al. 2002, A&A, submitted
Fontana A., D'Odorico S., Poli F., Giallongo E., Arnouts S.,
Cristiani S., Moorwood A., Saracco P. 2000, AJ, 120, 2206
Gardner, J. P.; Cowie, L. L.; Wainscoat, R. J., 1993, ApJ, 415,
L9
Gardner, J. P.; Sharples, R. M.; Carrasco, B. E.; Frenk, C. S.,
1996, MNRAS, 282, L1
Gavazzi G., Pierini D., Boselli A. 1996, A&A, 312, 397
Gehrels N. 1986, ApJ, 303, 336
Giacconi R., Rosati P., Tozzi P. et al. 2001, ApJ,551,624
Glazebrook, K.; Peacock, J. A.; Collins, C. A.; Miller, L., 1994,
MNRAS, 266, 65
Hall P.B., Green R.F., Cohen M. 1998, ApJS, 119, 1
Huang, J.­S.; Cowie, L. L.; Gardner, J. P.; Hu, E. M.; Songaila,
A.; Wainscoat, R. J., 1997, ApJ, 476, 12
Impey C. D., Sprayberry D., Irwin M. J., Bothun G. D, 1996,
ApJS, 105, 209
Impey C. D., Bothun G. D, 1997, ARA&A, 35, 267
Kauffmann G., White S.D.M., Guiderdoni B. 1993, MNRAS,
264, 201
Kauffmann G., 1996, MNRAS, 281, 487
Kauffmann G., Charlot S. 1998, MNRAS, 297, L23
Kinney A.L., Calzetti D., Bohlin R.C. et al. 1996, ApJ,467,38
Kochanek C.S., Pahre M.A.,Falco E.E. et al. 2001,ApJ,560,566
Kron R.G., 1980, ApJS, 43, 305
K¨ummel, M. W.; Wagner, S. J., 2000, A&A, 353, 867
Lilly S.J., Le Fevre O., Crampton D., Hammer F., Tresse L.,
1995, ApJ, 455, 50
Madau P., Pozzetti L., Dickinson M., 1998, ApJ, 498, 106
Maihara T. et al., 2001, PASJ, 53, 25
Marleau F.R., Simard L. 1998, ApJ, 507, 585
Martini P., 2001, AJ, 121, 598
McCarthy et al. 2001, ApJL, 560, L131
McLeod, B. A.; Bernstein, G. M.; Rieke, M. J.; Tollestrup, E.
V.; Fazio, G. G., 1995, ApJS, 96, 117
Minezaki, T., Kobayashi Y., Yoshii Y., Peterson B.A., 1998,
ApJ, 494, 111
Mobasher, B.; Ellis, R. S.; Sharples, R. M., 1986, MNRAS, 223,
11

A. Cimatti et al.: The K20 survey sample 13
Moorwood A.F.M., Cuby J.­G., Lidman C. 1998, The
Messenger, 91,9
Moorwood A.F.M. et al. 1999, The Messenger, 95, 1
Moustakas, L.A.; Davis, M., Graham, J.R., Silk, J., Peterson,
B.A.; Yoshii, Y., 1997, 475, 445
Pahre M.A., 1999, ApJS, 124, 127
Peebles P.J.E. 2002, in ``A new era in cosmology'', Durham,
U.K., ASP Conference Series, ed. N. Metcalfe & T. Shanks,
in press,astro­ph/0201015
Persson S.E., Murphy D.C., Krzeminski W., Roth M., Rieke
M.J. 1998, AJ, 116, 2475
Rengelink R. et al. 1998, submitted to A&A, astro­ph/9812190
Renzini A. 1999, in ``The formation of galactic bulges'', edited
by C.M. Carollo, H.C. Ferguson, R.F.G. Wyse. Cambridge,
U.K., Cambridge University Press, p.9
Renzini A., Cimatti A. 1999, in ``The Hy­Redshift Universe:
Galaxy Formation and Evolution at High Redshift'',
Berkeley, USA, ASP Conference Proceedings, Vol. 193,
Edited by Andrew J. Bunker and Wil J. M. van Breugel,
p. 312
Rudnick G., Franx M., Rix H.­W., Moorwood A., Kuijken K.,
van Starkenburg L., van der Werf P., R¨ottgering H.J.A.,
van Dokkum P., Labbe I. 2001, AJ, 122, 2205
Saracco, P.; Iovino, A.; Garilli, B.; Maccagni, D.; Chincarini,
G., 1997, AJ, 114, 887
Saracco P., D'Odorico S., Moorwood A., Buzzoni A., Cuby J.­
G., Lidman C. 1999, A&A, 349, 751
Saracco, P.; Giallongo, E.; Cristiani, S.; D'Odorico, S.;
Fontana, A.; Iovino, A.; Poli, F.; Vanzella, E., 2001, A&A,
375, 1
Scalo J.M. 1986, Fundam. Cosm. Phys., 11, 1
Schlegel D.J., Finkbeiner D.P., Davis M. 1998, ApJ, 500, 525
Soifer, B. T. et al., 1994, ApJ, 420, L1
Somerville R.S, Primack J.R.,Faber S.M. 2001, MNRAS, 320,
504
Songaila A., Cowie, L.L., Hu E.M., Gardner J.P. 1994, ApJS,
94, 461
Stern D., Connolly A., Eisenhardt P., Elston R., Holden B.,
Rosati P., Stanford A., Spinrad H., Tozzi P., Wu K.
2001, in ``Deep Fields'', Proceedings of the ESO Workshop,
Garching, Germany, Springer­Verlag, p. 76
Szokoly, G.P.; Subbarao, M.U.; Connolly, A.J.; Mobasher, B.,
1998, ApJ, 492, 452
van Dokkum P.G. & Stanford S.A. 2001, ApJL,562,L35