The Vimos VLT Deep Survey:
Key Words.:Galaxies: fundamental parameters - Cosmology: observations - Cosmology: large-scale structure of the Universe - Astronomical data bases: Catalogs
Context:The VVDS-Wide survey has been designed with the general aim of tracing the large-scale distribution of galaxies at on comoving scales reaching , while providing a good control of cosmic variance over areas as large as a few square degrees. This is achieved by measuring redshifts with VIMOS at the ESO VLT to a limiting magnitude , targeting four independent fields with size up to 4 each.
Aims:We discuss here the survey strategy which covers 8.6 and present the general properties of the current redshift sample. This includes 32734 spectra in the four regions, covering a total area of 6.1 with a sampling rate of 22 to 24%. This paper accompanies the public release of the first 18143 redshifts of the VVDS-Wide survey from the 4 deg contiguous area of the F22 field at RA=22.
Methods:We have devised and tested an objective method to assess the quality of each spectrum, providing a compact figure-of-merit, particularly effective in the case of long-lasting spectroscopic surveys with varying observing conditions. Our figure of merit is a measure of the robustness of the redshift measurement and, most importantly, can be used to select galaxies with uniform high-quality spectra to carry out reliable measurements of spectral features. We use the data available over the four independent regions to directly measure the variance in galaxy counts. We compare it with general predictions from the observed galaxy two-point correlation function at different redshifts and with that measured in mock galaxy surveys built from the Millennium simulation.
Results:The purely magnitude-limited VVDS Wide sample includes 19977 galaxies, 304 type I AGNs, and 9913 stars. The redshift success rate is above 90% independently of magnitude. A cone diagram of the galaxy spatial distribution provides us with the current largest overview of large-scale structure up to z, showing a rich texture of over- and under-dense regions. We give the mean N(z) distribution averaged over 6.1 for a sample limited in magnitude to . Comparing galaxy densities from the four fields shows that in a redshift bin at one still has factor-of-two variations over areas as large as deg. This level of cosmic variance agrees with that obtained by integrating the galaxy two-point correlation function estimated from the F22 field alone. It is also in fairly good statistical agreement with that predicted by the Millennium mocks.
Conclusions:The VVDS WIDE survey provides the currently largest area coverage among redshift surveys reaching z. The variance estimated over the survey fields shows explicitly how clustering results from deep surveys of even 1 deg size should be interpreted with caution. The survey data represent a rich data base to select complete sub-samples of high-quality spectra and to study galaxy ensemble properties and galaxy clustering over unprecedented scales at these redshifts. The redshift catalog of the 4 deg F22 field is publicly available at http://cencosw.oamp.fr.
distribution of galaxies contains unique information on the structure of
our Universe and the fundamental parameters of the cosmological model.
The relation of galaxy properties to large-scale structure in turn
provides important clues on the physics of galaxy formation within the
standard paradigm in which baryons are assembled inside dark-matter
halos (e.g. White & Rees 1978). Redshift surveys of the “local”
() Universe as the 2dFGRS (Colless et al. 2001) and SDSS (Abazajian et al. 2003)
contain several hundred thousand
galaxies spanning a few thousands square degrees.
These large samples and explored volumes
have allowed large-scale structure studies to be pushed well into the linear
regime h Mpc
having at the same time a detailed characterization of small-scale
clustering and its dependence on galaxy properties like luminosity, colour and
morphology (e.g. Madgwick et al. 2003; Norberg et al. 2001, 2002; Zehavi et al. 2005; Li et al. 2006).
All these features and properties are expected to depend on redshift,
and different evolutionary paths can lead to similar observational
properties in the local universe.
Ideally, one would
like to be able to gather similarly large samples over comparably large
volumes, at cosmologically relevant distances ().
First pioneering deep redshift surveys capable of
measuring the evolution of clustering go back to the 1990’s and were
limited to a few hundred square arcminutes
(e.g. Le Fevre et al. 1996; Shepherd et al. 2001).
Even deeper measurements of clustering evolution were provided by
specific color-selected surveys, using the Lyman-break technique,
although these give a very biased view of large-scale structure
limited to a specific class of objects (e.g. Steidel et al. 1998).
More recent surveys like GOODS (e.g. Giavalisco et al. 2004) and DEEP
(e.g. Koo 1995)
multi-wavelength coverage, but are still limited to small fields.
Only recently, thanks to the increased
multi-plexing ability of spectrographs mounted on 10-m class telescopes,
robust clustering studies of the general galaxy population at have become feasible. This opportunity has been exploited by the
VVDS (Le Fèvre et al. 2005c) and the DEEP2 (Davis et al. 2003) surveys.
The VVDS Deep sample (Le Fèvre et al. 2005c), in particular, covered a
reasonably large area ( ) up to redshift 4 and
to a very deep
magnitude limit (). Major clustering results using these
data have included studies of the evolution of galaxy
clustering since (Le Fèvre et al. 2005a),
its dependence on luminosity,
spectral type and stellar mass (Pollo et al. 2006; Meneux et al. 2006, 2008)
and the evolution and non-linearity of
galaxy bias (Marinoni et al. 2005), together with a direct assessment
of the evolution of enviromental effects, as the dependence of colour
(Cucciati et al. 2006) or luminosity function
(Ilbert et al. 2005; Zucca et al. 2006; Ilbert et al. 2006) on local density.
Still, the area surveyed
by the VVDS Deep is not yet large enough to fully
characterize large-scale structure at high redshift: results from 2dF
show that structures of size Mpc do exist in the local
Universe, while in the VVDS-Deep itself a structure at
is found to fill the full survey field (14
h Mpc) (Le Fèvre et al. 2005b) The Wide part of the VVDS survey has been conceived specifically to
improve upon this, covering structures with size h
Mpc at , while having the ability to measure the variance in
galaxy density on scales of a few square degrees. This will be achieved by
measuring redshifts to over four separated fields on
the sky, with size up to 4 deg each.
In this paper we present a first analysis of the currently available redshifts from the VVDS-Wide spectroscopic survey, including in particular the data collected over the full deg area of the F22 field, which are publicly released to the scientific community. The paper is organized as follows. In section 2 we describe the VVDS Wide survey strategy and report on the status of the observations conducted so far; in section 3 we assess redshift reliability depending on data quality, in section 4 we present the main characteristics of the resulting redshift catalog, while in section 5 we present the widest cone diagram currently available up to , quantify the field to field variance of the redshift ditribution and how it can affect smaller size surveys and compare the observed cosmic variance with model predictions.
Throughout this paper, we have used a Concordance Cosmology with , and . The Hubble constant is normally parameterized via , while a value has been used when computing absolute magnitudes.
2 The VVDS Wide survey
The VVDS Wide survey uses VIMOS at the ESO VLT to target
4 separate fields,
one of which includes the VVDS Deep survey area,
evenly distributed on the sky and covering a total of
With a deg size, each field
can span along the diagonal a transverse comoving size of 116 h
Mpc at .
The names and coordinates of each field are given in
In each of the areas we have excellent photometric coverage, extending from U to K. In addition to the U, , surveys conducted by the VVDS team (Radovich et al. (2004); McCracken et al. (2003); Iovino et al. (2005)), the sky regions at 02 and 22 hours are also covered by the CFHTLS survey
The original plan of the VVDS Wide survey, involved a “two-pass” observing strategy: each of the four areas is covered by two slightly displaced (2 arcmin) grids of adjacent VIMOS pointings (see below for a description of the current implementation of this plan). This strategy allows one to reach a spectroscopic sampling rate of of all galaxies with , which is important for density reconstruction studies (Marinoni et al. 2005), while keeping the required observing time within reasonable limits. At the same time, the 2-arcmin shifts is chosen as to fill (at least partially) the gaps left by the VIMOS footprint. For the VVDS Wide survey, the exposure time of each pointing was 45 minutes in MOS mode, using the Low Resolution Red grism. As in the case of the VVDS Deep survey (see Le Fèvre et al. 2005c), we have used a jitter observing sequence, with 5 steps along the slit, each separated by 0.7 arcsec. This strategy allows us to considerably reduce the fringing produced by the CCDs above (LeFevre et al. 2003), although fringing residuals still appear for the brighter and more extended sources, as well as in those observations where seeing was higher than 1.0 arcsec. Observation preparation, mask layout, and observing strategy is the same as for the VVDS Deep sample: using the VMMPS software (Bottini et al. 2005), we have been able to place slits on average per VIMOS mask-set down to the limiting magnitude of the VVDS Wide survey. Data have been reduced using the VIMOS Interactive Pipeline and Graphical Interface package (VIPGI Scodeggio et al. 2005).
The observations presented here were collected mostly during Guaranteed Time observations (5 extended visitor observing runs from Oct. 2002 to Sep. 2004), with a small fraction acquired during two further runs in Guest Observer standard time (service runs in February 2006 and 2007). As visually summarized in Fig. 1, we completed the first pass on the 4.0 F22 area, plus a second pass on the central deg of the same field. We further include here the redshift measurements from the first pass over 0.8 and 1.2 in F10 and F14 respectively, while further 2.1 deg in these area are under analysis and are not included in this paper (grey dots in Fig. 1). Finally, we also include redshifts for all galaxies with in the of the F02 field covered by the VVDS-Deep survey to . Overall, this data set corresponds to 36% of the original VVDS Wide survey goal.
Given the instrument geometry, with one pass only there are empty crosses not covered by the instrument (see Fig. 1, F10, F14 and outer part of F22 areas). In Table 1 we give both the total area covered so far by the VVDS Wide survey, (i.e. the global area covered by either grey or black points in fig. 1), and the effective area, i.e. the area including only fully reduced pointings (black points only) and net of the empty crosses. In Table 1 we also give the total number of pointings reduced so far for each field, and the average sampling rate of measured redshifts at the given magnitude limit.
|Field||R.A||Dec||Surveyed area||Effective area||N.of pointings||sampling rate|
|For 1 pointing, only 1 quadrant has been reduced so far|
|Reduction of 4 pointings is still partial|
3 Redshift measurement, data quality, and reliability
Redshifts have been measured using the same “double-check” procedure described in Le Fèvre et al. (2005c), adopting the same grading scheme to characterize the reliability of the measured redshift:
flag 4: a 100% secure redshift, with high SNR spectrum and obvious spectral features supporting the redshift measurement;
flag 3: a very secure redshift, strong spectral features;
flag 2: a secure redshift measurement, several features in support of the measurement;
flag 1: a tentative redshift measurement, based on weak spectral features and continuum shape;
flag 0: no redshift measurement possible, no apparent features;
flag 9: only one secure single spectral feature in emission, tipically interpreted as [OII]3727 Å, or .
A similar classification is used for broad line AGN, which we identify as spectra showing at least one ”broad line” (i.e. resolved at the spectral resolution of the VVDS). Flags for broad line AGN have the following meaning
flag 14: secure AGN with 100% secure redshift, at least 2 broad lines;
flag 13: secure AGN with good confidence redshift, based on one broad line and some faint additional feature;
flag 19: secure AGN with one single secure emission line feature, redshift based on one line only;
flag 12: a 100% secure redshift measurement, but lines are not significantly broad, might not be an AGN;
flag 11: a tentative redshift measurement, spectral features not significantly broad.
Serendipitous objects appearing by chance within the slit of the main target are identified by adding a “2” in front of the flag. We have classified with flag = -10 objects in slits with a clear observational problem, like e.g. objects for which the automated spectra extraction algorithm in VIPGI (Scodeggio et al. 2005) failed, or objects too close to the edge of a slit to allow for a proper sky subtraction. In the following, redshifts with a flag between 2 and 9 (or 12 and 19 in the case of AGN) are referred to as secure redshifts.
3.1 Data quality
When conducting a large spectroscopic survey, carried out over
years, under different weather conditions, and with
different people involved at different times in the data reduction and
redshift measurement process, it is important to identify an objective way
to assess the quality of the data and of the reduction process,
independent on the redshift measurement success or failure.
For the VVDS Deep and Wide surveys, we have devised a method which
takes into account the most important
Slit obscuration due to field vignetting, effective exposure time, seeing and sky transparency directly impact on the number of photons collected for each spectrum; sky brightness at constant exposure time determines the S/N ratio, and the quality of the wavelength calibration has an impact on the accuracy of the redshift measurement. The goal we set was to devise an objective quality parameter which could be used to make an a priori selection of the best data at hand. The final figure of merit we assigned to each spectrum is the combination of all these factors in such a way that the highest is the figure of merit, the highest the spectrum quality. In the following, we shall discuss in turn each separate contribution to this quality parameter, show the overall results for both surveys and relate it to the redshift confidence level.
VIMOS takes advantage of the full Nasmyth field of view, but, due to the design of the guiding probe, a fraction of the field of view can be partially vignetted for some positions of the guiding star. This has happened especially during the first observations, before enough experience had been gained on the choice of the guiding star. Obscured slits can be easily identified by looking at the average level of sky counts in each slit, and comparing it with the average sky level for all the slits in the quadrant. When the single slit has a sky level which is lower than 70% the average sky level, the slit has been flagged as “bad” assigning to it an . This happens for a total of 75 objects in the VVDS Deep data, and 181 in the VVDS Wide data. An a posteriori check shows that these slits account for 68% and 50% of the spectra where no object is detected in the VVDS Deep and VVDS Wide sample respectively. As all the contributions to the quality parameter are eventually combined in a multiplicative way, all obscured slits will end up in a global quality parameter equal to zero.
As described in Scodeggio et al. (2005), wavelength calibration is performed using both a global fitting, and a slit per slit refinement. For some particular slits towards the edge of the field of view, and in particularly unfavourable positions of the instrument during the observation, flexures can be important, and it is not possible to get a wavelength calibration of the same quality as usual. Using the wavelength calibration rms for each slit (see Scodeggio et al. 2005) as a measure of the wavelength calibration quality, we can identify such deviant cases. The distribution of the wavelength calibration rms for the VVDS Deep (solid line) and the VVDS Wide (dotted line) survey spectra is shown in Fig. 2. There is a small number of slits (1% both in the VVDS Deep survey, and in the VVDS Wide survey) showing a wavelength calibration rms above , which essentially means that wavelength calibration has totally failed. Such slits get a ”wavelength calibration quality flag” .
Furthermore, for some pointings the arc calibration exposure was not usable, and we were forced to calibrate the corresponding spectra using the sky lines. In such cases, the wavelength calibration is never as good as in the standard case, since the sky lines are often broad and/or unresolved at our resolution. This is reflected by the distribution of the wavelength calibration rms for that specific quadrant/pointing, which peaks at rather than the usual . We have considered these slits as an intermediate category, and assigned to them a wavelength calibration quality flag of -1. As all the contributions to the quality parameter are eventually combined in a multiplicative way, a shaky ( rms ) or bad ( rms ) wavelength calibration will end up in a global quality parameter being negative or equal to zero respectively.
Sky brightness depends (at zero order) on moon phase and moon distance. In principle, knowing these two parameters and using some calibration table, the expected sky brightness could be computed. In practice, as we are interested in the global background level, a simpler approach has been adopted:
for each slit, the mean one dimensional sky spectrum is obtained by taking the median over the sky two-dimensional image along the spatial direction, to reject border effects;
after having discarded obscured slits (see 3.1.1), all one dimensional sky spectra are combined and the median sky spectrum for the whole quadrant derived;
such median sky spectrum is then normalized for the exposure time and integrated over the full wavelength range to get the median sky value for that quadrant in that pointing ();
by comparing the median sky values obtained for the same quadrant in the different pointings, we define a “reference sky value” () as the mean of the three lowest ”median sky values”;
finally, the median sky value per quadrant per pointing is compared to for that quadrant, and a sky quality factor is defined as
In Fig. 3, top left, the distribution of the sky quality parameter for the VVDS Deep (solid line) and VVDS Wide (dotted line) surveys is shown. Overall, the VVDS Wide Survey shows a sky quality parameter distribution broader than the VVDS Deep survey. This is expected, as the VVDS Deep survey observations have been carried out during dark time, while the VVDS Wide survey ones have been partially performed during grey time.
The exposure time for each pointing of the VVDS Deep survey had been
planned to be 16200 seconds, while for the Wide it should have been
2700 seconds per pointing. As a matter of fact, in some cases the effective
exposure time has been less than what foreseen, mainly because
metereological conditions had badly deteriorated during the observation,
and of course exposure time has a direct impact on
the signal to noise ratio as a multiplicative factor.
There are also a
few observations, performed during visitor runs, which have been
lengthened in the attempt to compensate for high airmass or unstable
By comparing the actual total exposure time of
each quadrant in each pointing (), to the nominal exposure time “a
priori” established for the survey (), we can
In Fig. 3, top right, the distribution of the Exposure Time quality parameter for the VVDS Deep (solid line) and VVDS Wide (dotted line) surveys is shown. For the vast majority of the pointings, the exposure time effectively used is what had been foreseen for that depth.
Sky Transparency, seeing and slit losses
Atmospheric conditions have a direct influence on sky
transparency and seeing,
which in turn, and coupled with slit losses,
contribute to flux losses in a way which is
not possible to disentangle. In order to estimate their global
an empirical approach has been adopted:
for each object, we have integrated its spectrum under the I filter
response curve, and compared the thus obtained
to the equivalent quantity as obtained
from photometry ().
Then one could in
principle compute in one shot the effect of seeing, slit losses and transparency on
Such ratio should always be below one, by definition, but, as shown in
Le Fèvre et al. (2005c),
a small fraction (around few percent) of objects have a value of
above 1.0. This is
due to a number of second-order effects affecting the measure, such as: 1)
is affected by how well
zero orders or fringing residuals have been removed; 2)
has its own errors, larger for fainter magnitudes ( 0.2 mag for
objects fainter than , for the VVDS Deep survey,
McCracken et al. (2003));
3) the brighter and more
extended the object, the more inaccurate is the sky subtraction:
the sky region that can be used to compute the sky level is
small and dominated by pixels affected by slit edge effects. This can
lead to an underestimate of the sky level.
This more often occurs in the VVDS Wide survey pointings, where
the fraction of brighter ( larger) objects is higher, and/or in bad
To quantify the overall contribution of such second
order effects, we can define
actually due to sky transparency and seeing/slit width ratio,
and, for pointlike sources,
should be constant within one observation.
contribution of all the second order effects like those listed above,
and as such can vary from object to object.
In optimal atmospheric conditions, we should have
and . Indeed,
in the magnitude range between and
, where the error on photometric magnitude is negligible and
object sizes are small enough not to be affected by slit losses, or to hamper a good background
estimate, the mean flux ratio is
always below one, being affected by sky transparency only.
Thus, on a per quadrant and per pointing basis,
using only the range
we can compute the mean transmission as
Subsequently, and for each object for which
is above one, we can estimate
the residuals as
Finally, the contribution of all these factors to the observation quality
can be computed as
In Fig. 3, panel e, the distribution of the parameter for the VVDS Deep (solid line) and VVDS Wide (dashed line) surveys is shown. Also objects for which no redshift has been measured are included in this panel. Panel c and d show the distribution of the same parameter for stars and galaxies separately. Comparing the two distributions obtained for stars, we see that in the VVDS Deep survey most stars have a figure of merit close to 1, while for the VVDS Wide survey the peak is around 0.8. As stars are less affected by slit losses than galaxies, panel c tells us that the overall better figure of merit for the VVDS Deep survey is mainly due to the better average atmospheric conditions during observations.
Global quality parameter
The above quality parameters have been computed for each object and
combined in a multiplicative way as
The resulting distribution is shown in Fig. 3 bottom panels, for the VVDS Deep (solid
line) and VVDS Wide (dotted line) data.
Negative values of quality pertain to
objects with poor wavelength calibration, while a quality parameter of
0 is due to either bad wavelength calibration or to obscured slits.
From Fig. 3,
panel h, it is apparent that the global distribution of
the quality parameter is (slightly) better for the VVDS Deep survey data than
for the VVDS Wide survey data. This is essentially
due to the parameter, which is better for the deep survey,
while the VVDS Wide survey, in which large and extended galaxies
are more abundant, is globally more affected by slit
The percentage of galaxies with quality above 0.5 is 79% in the VVDS Deep sample and 58% in the VVDS Wide sample. We note, however, that the quality parameter is not an absolute measure of data quality, but just a relative one: it allows to select the best quality spectra we have in our samples (i.e. those for which slit losses are small, observed for the nominal exposure time in excellent atmospheric conditions), or conversely, to discard those data for which something during observations or reduction went wrong. We will see in the following section that this does not necessarily prevent, nor assures, 100% reliable redshift measurements.
Data quality parameter and redshift flag
Once the global quality parameter is obtained, it is interesting to
see how it relates to the redshift flag.
If both estimates are reliable, we expect that
objects with a bad value of the
quality parameter (i.e. below 0.5) should have a higher probability of
an unsuccessful redshift measurement (flag =0),
while objects with a good value of the quality parameter (above 0.5)
should have a higher probability of a
very secure redshift flag (i.e. 3 or 4). Still, we do not expect
the opposite to be
totally true, i.e. there may exist spectra with a not so good quality
but to which the redshift can be securely assigned: in fact, the quality
parameter is related to the continuum intensity, its signal to
the absolute flux calibration of the data, while the
flag is a measure of the reliability of the redshift, and is
strongly affected by the presence, or absence, of prominent
In Fig 4, for each flag, the distribution of the
quality parameter is shown for the VVDS Deep (left) and VVDS Wide
(right) surveys. The dotted line indicates the 0.5 value of the
quality parameter. As expected, more secure flags are assigned to
objects showing, on average, a higher quality parameter,
as shown by the peak of the histogram
moving towards higher values of quality
going to more secure flags. An exception are the flag 9 objects, which
show a distribution of the
quality parameter comparable to that of the flag 1
objects. This is not a surprise: we recall that a flag 9 is assigned when
one secure single spectral feature in emission is visible, and, as
anticipated before, an emission line, if strong enough, can be
detected even in presence of low S/N continuum, or residual
More quantitatively, in the VVDS Deep survey only 42% of the failed spectra have a quality parameter larger than 0.5, a percentage which goes down to 20% for the Wide survey. On the other hand, 82% of the objects with a very secure redshift flag ( flag 3 or 4) in the Deep survey are derived from spectra of good quality (quality ), a percentage which decreases to 73% in the Wide case.
Thus, the quality parameter statistically strengthens the redshift flag, and justifies it on the basis of the quality of the data. Furthermore, the coupling of the two pieces of information allows one to easily select subsamples of objects for which not only the redshift has the highest degree of reliability, but the data are above a given quality and thus particularly suitable for detailed studies of the continuum emission.
4 General properties of the spectroscopic sample
|primary targets||secondary targets|
In Table 2 we summarize the statistics of redshift measurement for the VVDS Wide sample. For reference, we also report the redshift statistics for the VVDS Deep sample, once it is cut at a limiting magnitude of . So far, in the 3 Wide survey fields we have accumulated 28166 spectra for primary targets, including 16670 galaxies, 258 QSOs and 9164 stars. There are only 2074 spectra for which the redshift measurement failed; this corresponds to a success rate larger than 92%. Secure redshift objects (flag between 2 and 9), are 21894, almost 80% of the sample. Although the magnitude limit is only , thanks to the large surveyed area ( of effective area), we have a fairly large sample of rare, luminous galaxies at high redshift: 979 with and 225 with . The highest secure redshift measured for a galaxy is 4.0573, while the highest secure redshift object is a QSO at z=5.0163. On top of the targeted sample, we also have 772 additional redshifts of objects accidentally falling within the slit. Adding the data collected in the F02 field limited to , the VVDS Wide sample comprises almost 20000 galaxies and 304 QSOs with measured redshift over 6.1 .
4.1 Magnitudes, sampling rate and redshift distribution
In Fig. 5, the magnitude distribution of the photometric parent
catalog (top panels, empty histogram) and of the final
spectroscopic sample (top panels, shaded histogram)
is shown for the three VVDS Wide areas. For comparison, we show the
same plot for the VVDS-Deep F02 field, limited to .
The bottom panels
show the fraction of observed over total objects vs. magnitude. In
all the three VVDS Wide areas, the fraction of observed objects is
and 25%, over the full magnitude range,
very similar to the
sampling rate of the VVDS Deep area, once limited at .
The slight trend
favouring a better sampling at the faintest magnitude in the F02
field is due to the intrinsic
deeper limiting magnitude of the spectroscopic selection in this area,
, which increases the probability of brighter objects to be
discarded in favour of fainter ones (see Bottini et al. 2005).
In Fig. 6, the redshift distributions using all available redshifts (irrespective of flags) for the four different areas are shown. Table 3 shows that there are no statistically significant differences between the N(z) obtained using all redshifts, and the ones obtained using only secure redshifts (i.e. flag 2 to 9, redshift confidence ).
|Field||1 quartile||median||3 quartile||1 quartile||median||3 quartile|
|all flags||secure flags|
4.2 Galaxy luminosities and stellar masses
The large areas explored, coupled with the relative bright magnitude limit, make the VVDS Wide the ideal survey to explore the bright/massive ends of the luminosity/mass function up to redshift . As an example of the potential of this sample for these studies, in Fig. 7 and 8 we show the absolute B magnitude and stellar mass vs. redshift distribution for the galaxies with secure redshifts in the four areas. Absolute B magnitudes and stellar masses have been derived by fitting the photometric and spectroscopic data with a grid of stellar population synthesis models generated with the PEGASE2 population synthesis code (Fioc & Rocca-Volmerange 1997), and using the GOSSIP Spectral Energy Distribution tool (Franzetti et al. 2007), where we have adopted a Salpeter IMF and a delayed exponential SFH (see Pozzetti et al. (2007) for a thorough discussion on the dependence of mass values on the different IMF adopted). We can define a unique complete sample of 3542 bright galaxies with up to : 2136 galaxies in the F22 field, 412 in F10, 520 in F14, 474 in F02 (we remind that we have cut the F02 area to a limiting apparent magnitude ). At the same limit in redshift, we have more than 11000 galaxies more massive than log(M)=10. (6547 galaxies in the F22 field, 1367 in F10, 2009 in F14, 1271 in F02), a sample which will allow a detailed study of the properties of medium to high mass galaxies.
4.3 A direct test of star-galaxy separation techniques
As mentioned earlier, the VVDS was deliberately carried out
without any star-galaxy separation prior to spectroscopy.
When the survey was planned, only ground based photometry was
available (and not over all fields), thus preventing us from using the
most efficient color based methods to discriminate between stars and
Furthermore, the image quality of such ground based photometry was not
good enough to apply geometrical arguments discriminate between point-like and
extended sources down to the magnitude limits of the Deep and Wide
Thus, we decided to follow the conservative approach of
not attempting any a priori removal of starlike objects based on
colors or compactness.
This has lead to the high stellar contamination of the spectroscopic
sample (up to 1/3 for the lower galactic latitude fields).
Using UKIDSS K photometry, and CFHTLS z photometry available in the
F22 and F02 field, we
can test with excellent statistics the
performances of these
star identification methods.
thus applied to the spectroscopic sample the BzK criterium
described in Daddi et al. (2004), coupled with a compactness criterium
based on the stellarity index provided by Sextractor: any
object with a stellarity index above or equal 0.9 is catalogued as
compact. An object is considered as a star if both
criteria are satisfied.
To optimize the test, we used only objects
with secure redshift (redshift flag ) and small photometric
errors (err) in the B, K and z Bands.
Applying this technique to the F02 data (which are deeper and at high galactic latitude), we end up with a residual stellar contamination of only . In the F22 field, which has a brighter magnitude limit and is located at lower galactic latitude, the residual contamination decreases from 35% to 14%. The price to be paid in terms of galaxies which would have been a priori discarded is about 5% in the F22 field, and about 2% in the F02 case. We have checked which kind of galaxies were typically discarded and found out that they have the spectrum of a normal elliptical galaxy. Overall, we can state on the basis of observed data that the performance of these two coupled methods in discarding stars is highly efficient, at the low price of a small loss of normal early type galaxies In addition, we note that we have applied the colour method in the standard form. Exploring in detail alternative color-color selections using the other available bands is beyond the scope of this paper.
5 Large-scale structure and density fluctuations in the VVDS Wide fields
5.1 Galaxy spatial distribution in the F22 field
In Fig. 9 we show the redshift space cone diagram of all galaxies observed in the F22 area, in co-moving coordinates and projected onto the right ascension plane. The figure shows two declination slices, of 1 degree each, to better show the extension of the different structures. Note that the aspect ratio is stretched along the vertical direction. We can identify galaxy overdensities at , 0.33, 0.41, 0.53, 0.75, 0.82 and 0.9, some of which extend over the full surveyed area, both in right ascension and declination: at z=0.33 a very thin wall covers the whole field of view of ; the structure at is the most prominent and massive, extending for almost 80 along the line of sight, and 40 across. Its presence strongly influences the redshift distribution in this field, lowering its median value and steepening its rise at low redshifts. Such “thick wall” has several subconcentrations, better visible in the slices in declination of Fig. 9. The other visible structures look rather more compact, with a comoving transverse size of the order of 20 , and confined within the 2 square degrees.
5.2 Mean redshift distribution up to
In terms of their broad shape and peak position, the galaxy
redshift distributions in the four areas
are relatively similar.
At the same time, however, significant field-to-field
variations are evident (e.g. the thick wall at 0.53 in the F22 field,
as outlined in the previous section). In this and the following
this variance and compare it to theoretical expectations, as obtained both from
the observed two-point correlation function and from mock surveys
built using numerical/semi-analytic models.
Combining the four fields, appropriately taking into account the effective area and the sampling rate of each field, we can derive our current best estimate of the redshift distribution of a magnitude selected sample to . The result is shown in Figure 10 and the corresponding values are reported in Table 4 for convenience. In this figure and table, we use a binning of z=0.1 up to z=1, and 0.2 at higher redshift, in order to smooth out the smaller structures present in the different fields. This represents the most accurate redshift distribution mesured to date at these faint magnitudes, based on galaxies over a total area of 6.1 deg, and it can provide an important reference for galaxy formation models.
5.3 Field to field variations
With this unprecedented area surveyed, it becomes
possible to quantify the variations in each of the four
fields with respect to this average distribution. This is shown
in Fig. 11. The top panel reports the redshift
distribution of the four fields, using a z=0.1 binning.
For reference, around the peak of the distribution z=[0.5,0.6]
such a redshift bin corresponds to a comoving radial size of 222
. Error bars correspond to Poissonian
errors. In the bottom panel of figure 11, we show
the fractional difference between the observed N(z) in each field, and the
average distribution. This comparison of the fluctuations in the
different fields for fixed redshift bins is inevitably
qualitative. In fact, given the very different areas covered
in the four fields, the same redshift range corresponds to rather
For example, the smallest field, F02, covers 0.5 deg, i.e. 8
times smaller than the largest one, F22 (4 deg). This is
certainly one reason for the higher variance in the F02 field
(Fig. 11, filled circles).
To properly estimate the intrinsic variance as a function of
scale, we have therefore defined a set of square sub-fields over the
four survey areas, with increasing angular size. The variance is
then computed among the set of N homogeneous volumes having
identical size on the sky and along the redshift direction. The
result is summarized in Fig. 12. In practice,
each sky region (represented by a different color and symbol) has
been divided – for a given size – into the largest possible number
of subareas that could be accomodated.
Galaxy densities have been computed in each sub-field and for
different redshift bins, properly correcting for the average
sampling of the area. Table 5 shows quantitavely the
results illustrated in Fig. 12. For each area
size, and each redshift bin, we give the number of sub-fields of that size available,
the median value,
upper/lower quartiles and maximum and minimum of the observed galaxy
For the largest area, where only 2 measurements are available, we
computed the arithmetic mean instead of the median. For the smaller
scales represented (190 arcmin) the observed large
fluctuations (up to a factor of four between adjacent areas) are not
surprising, as we are essentially looking at scales of the order of
a few Mpc. At redshifts approaching unity, and for large angular
sizes (30 arcmin correspond to 10 at this
redshift) the spread can still be a factor of 2, an indication that
important structures exist and are not uncommon at such redshift.
It is interesting to see, using the larger areas, how much variance
we expect in a field of (1800 square arcminutes) like
F02, i.e. the field of the VVDS-Deep survey. For example, at
redshift 0.75 we still see peak to peak fluctuations of 30 % in
the counts. We also notice a significant excess fluctuation in the data
from the field F22 (red circles) in the redshift bin 0.5-0.6. In
particular, all counts are shifted towards higher values, reflecting
the presence of the global large-scale fluctuation covering the full
field already mentioned in section 5.1.
5.4 Measuring cosmic variance
The results of Fig. 12 can be translated from the
observational space into a framework which is theoretically easier to
interpret in terms of cosmic variance and expectations from galaxy
clustering. Given a set of N identical volumes with volume , we can
define the observed variance among them as
(e.g. Somerville et al. (2004)), where the last term is the correction for Poissonian shot noise. In the following we will compute following eq. 1 only when the Poisson shot noise is smaller than 10%. The observed variance in the counts at a given redshift can be compared to that expected from the two-point correlation function of the galaxy sample. Following (Peebles 1980)
If the galaxy correlation function can be described as a power law, , then this expression becomes
where and and are measured from the observations.
Following the same approach as in Le Fèvre et al. (2005a),
and using the VVDS-Deep F02 data limited at
, we have estimated
the best fit correlation function parameters in different redshift
bins, and used eq. 3 to check whether the observed
variance measured from the field-to-field scatter (as from eq.
1) can be recovered
consistently by extrapolating the correlation function measured from a
much smaller field. The results are shown in Fig.
13, where we plot the observed square-root
of the variance (i.e. the value of the rms
fluctuation) against the volume. The red asterisks correspond to the
direct measurement, obtained from the scatter among N samples within
the given volume. The dashed area shows the same quantity as obtained
using the 3 confidence intervals of the VVDS-Deep F02
correlation function (limited at
One notices immediately that the variance directly estimated from the galaxy counts in the different fields is in excellent agreement with the cosmic variance as estimated from the correlation function. Only at redshift 0.35 and 0.65 and for volumes of h , field to field variance appears smaller than the one predicted from the correlation function parameters. Looking back at the distribution of the number counts (fig. 11), at these two redshifts we note a remarkable similarity among the different fields, as well as in the galaxy surface density distribution in fig. 12 and table 5. This similarity automatically converts in a lower field to field variance, which in any case remains compatible with the one computed from correlation function at a level.
Using 100 quasi-independent mock samples of degrees (for details, see Guzzo et al. 2008) built applying to the Millennium simulation (Springel et al. 2005; Blaizot et al. 2005) the semi-analytic prescription of De Lucia & Blaizot (2007), we have computed model predictions for cosmic variance again using eq. 1. The model predictions (green filled squares) are quite well consistent with the observed field-to-field variance, with a difference which is at most at z=0.65 for volumes of h .
6 Public Data Release and Database Access
We are publicly releasing all redshift measurements in the F22 area through the CENCOS (CENtre de COSmologie) database environment on our web site http://cencosw.oamp.fr with access to the database built under the Oracle environment, and through VO services (VVDS_WIDE ConeSearch service). The catalog can be searched by coordinates, redshift interval, identification number, in combination with the spectra quality flags. Spectra in FITS format are already available on the same site (or through VO SSA service) for the F02 area, both from the CENCOS site and the VO SSA service VVDS_F02_DEEP. The remaining redshifts, together with all spectra in FITS format, will be available as soon as the whole set of available data will be measured.
The VVDS Wide survey is still ongoing but it has already measured
redshifts for 26864 objects (including serendipitous objects)
in 3 areas covering a total of 5.6
to a limiting magnitude of , to which we
can add 3130 redshifts to the same limiting magnitude obtained by
the VVDS Deep survey in the F02 field. The
success rate in redshift measurement is more than 92% and more than three
quarters of the redshifts
have a confidence higher than 80%. Overall (i.e. including the F02 field
and serendipitous objects) the current sample includes 19777 galaxies,
line QSOs, and 9913 stars, while the total area covered amounts to
6.1 . When completed, the total area coverage will be of 8.6
, and the total number of redshifts of the order of 50000.
The large number of redshifts available in the F22 field, coupled with a sampling rate of 23%, allows to identify and describe several prominent structures present along the line of sight up to 2500 . Typical sizes are of the order of 20 , but one large clumpy structure extends for almost 80 along the line of sight, and 40 across.
We give the mean N(z) distribution averaged over 6.1 (fig. 10) for a sample limited in magnitude to . We have estimated field to field variations in terms of number counts (see fig. 11 and table 4) and galaxy surface density both as a function of field size and redshift (see fig. 12 and table 5), showing that differences as high as a factor of two can exist at z=1 for still relatively large scales of the order 30 arcminutes, like those considered in many deep surveys today. For fields limited to smaller scales (of the order 10 arcminutes), the spread in galaxy densities can be up to a factor 2.5. Still, the observed cosmic variance is consistent both with what derived from the correlation function parameters, and from theoretical simulations (see fig.13).
In addition to the evolution of clustering and large-scale structure, this data set is best suited to study in detail the bright end of the luminosity function, as well as the massive end of the mass function, up to in four different fields observed with identical purely magnitude limited selection.
The redshift catalog for the F22 area is available at the web site http://cencosw.oamp.fr, or via ConeSearch VO service VVDS_WIDE.
Acknowledgements.This research has been developed within the framework of the VVDS consortium.
This work has been partially supported by the CNRS-INSU and its Programme National de Cosmologie (France), and by Italian Ministry (MIUR) grants COFIN2000 (MM02037133) and COFIN2003 (num.2003020150) and by INAF grants (PRIN-INAF 2005).
The VLT-VIMOS observations have been carried out on guaranteed time (GTO) allocated by the European Southern Observatory (ESO) to the VIRMOS consortium, under a contractual agreement between the Centre National de la Recherche Scientifique of France, heading a consortium of French and Italian institutes, and ESO, to design, manufacture and test the VIMOS instrument.
- offprints: B.Garilli, email@example.com
- Based on observations obtained with MegaPrime/MegaCam, a joint project of CFHT and CEA/DAPNIA, at the Canada-France-Hawaii Telescope (CFHT) which is operated by the National Research Council (NRC) of Canada, the Institut National des Science de l’Univers of the Centre National de la Recherche Scientifique (CNRS) of France, and the University of Hawaii. This work is based in part on data products produced at TERAPIX and the Canadian Astronomy Data Centre as part of the Canada-France-Hawaii Telescope Legacy Survey, a collaborative project of NRC and CNRS.
- Abazajian, K., Adelman-McCarthy, J. K., Agüeros, M. A., et al. 2003, AJ, 126, 2081
- Arnouts, S., Schiminovich, D., Ilbert, O., et al. 2005, ApJ, 619, L43
- Blaizot, J., Wadadekar, Y., Guiderdoni, B., et al. 2005, MNRAS, 360, 159
- Bondi, M., Ciliegi, P., Zamorani, G., et al. 2003, A&A, 403, 857
- Bottini, D., Garilli, B., Maccagni, D., et al. 2005, PASP, 117, 996
- Chiappetti, L., Tajer, M., Trinchieri, G., et al. 2005, A&A, 439, 413
- Ciliegi, P., Zamorani, G., Bondi, M., et al. 2005, A&A, 441, 879
- Colless, M., Dalton, G., Maddox, S., et al. 2001, MNRAS, 328, 1039
- Cucciati, O., Iovino, A., Marinoni, C., et al. 2006, A&A, 458, 39
- Daddi, E., Cimatti, A., Renzini, A., et al. 2004, ApJ, 617, 746
- Davis, M., Faber, S. M., Newman, J., et al. 2003, in Discoveries and Research Prospects from 6- to 10-Meter-Class Telescopes II. Edited by Guhathakurta, Puragra. Proceedings of the SPIE, Volume 4834, pp. 161-172 (2003)., 161–172
- De Lucia, G. & Blaizot, J. 2007, MNRAS, 375, 2
- Fioc, M. & Rocca-Volmerange, B. 1997, A&A, 326, 950
- Franzetti, P., Garilli, B., Fumana, M., et al. 2007, in Proceedings of the ESA Workshop ”Astronomical Spectroscopy and the Virtual Observatory, ”
- Giavalisco, M., Ferguson, H. C., Koekemoer, A. M., et al. 2004, ApJ, 600, L93
- Guzzo, L., Pierleoni, M., Meneux, B., et al. 2008, Nature, 451, 541
- Ilbert, O., Cucciati, O., Marinoni, C., et al. 2006, ArXiv Astrophysics e-prints
- Ilbert, O., Tresse, L., Zucca, E., et al. 2005, A&A, 439, 863
- Iovino, A., McCracken, H. J., Garilli, B., et al. 2005, A&A, 442, 423
- Koo, D. 1995, in Wide Field Spectroscopy and the Distant Universe, ed. S. J. Maddox & A. Aragon-Salamanca, 55–+
- Le Fèvre, O., Guzzo, L., Meneux, B., et al. 2005a, A&A, 439, 877
- Le Fevre, O., Hudon, D., Lilly, S. J., et al. 1996, ApJ, 461, 534
- Le Fèvre, O., Paltani, S., Arnouts, S., et al. 2005b, Nature, 437, 519
- Le Fèvre, O., Vettolani, G., Garilli, B., et al. 2005c, A&A, 439, 845
- LeFevre, O., Saisse, M., Mancini, D., et al. 2003, in Proceedings of the SPIE, ed. M. Iye & A. F. M. Moorwood, Vol. 4841, 1670–1681
- Li, C., Kauffmann, G., Jing, Y. P., et al. 2006, MNRAS, 368, 21
- Lonsdale, C. J., Smith, H. E., Rowan-Robinson, M., et al. 2003, PASP, 115, 897
- Madgwick, D. S., Hawkins, E., Lahav, O., et al. 2003, MNRAS, 344, 847
- Marinoni, C., Le Fèvre, O., Meneux, B., et al. 2005, A&A, 442, 801
- McCracken, H. J., Radovich, M., Bertin, E., et al. 2003, A&A, 410, 17
- Meneux, B., Guzzo, L., Garilli, B., et al. 2008, A&A, 478, 299
- Meneux, B., Le Fèvre, O., Guzzo, L., et al. 2006, A&A, 452, 387
- Norberg, P., Baugh, C. M., Hawkins, E., et al. 2002, MNRAS, 332, 827
- Norberg, P., Baugh, C. M., Hawkins, E., et al. 2001, MNRAS, 328, 64
- Peebles, P. J. E. 1980, The large-scale structure of the universe (Princeton University Press)
- Pierre, M., Valtchanov, I., Altieri, B., et al. 2004, Journal of Cosmology and Astro-Particle Physics, 9, 11
- Pollo, A., Guzzo, L., Le Fèvre, O., et al. 2006, A&A, 451, 409
- Pozzetti, L., Bolzonella, M., Lamareille, F., et al. 2007, A&A, 474, 443
- Radovich, M., Arnaboldi, M., Ripepi, V., et al. 2004, A&A, 417, 51
- Schiminovich, D., Ilbert, O., Arnouts, S., et al. 2005, ApJ, 619, L47
- Scodeggio, M., Franzetti, P., Garilli, B., et al. 2005, PASP, 117, 1284
- Shepherd, C. W., Carlberg, R. G., Yee, H. K. C., et al. 2001, ApJ, 560, 72
- Somerville, R. S., Lee, K., Ferguson, H. C., et al. 2004, ApJ, 600, L171
- Springel, V., White, S. D. M., Jenkins, A., et al. 2005, Nature, 435, 629
- Steidel, C. C., Adelberger, K. L., Dickinson, M., et al. 1998, ApJ, 492, 428
- Warren, S. J., Hambly, N. C., Dye, S., et al. 2007, MNRAS, 375, 213
- White, S. D. M. & Rees, M. J. 1978, MNRAS, 183, 341
- Zehavi, I., Zheng, Z., Weinberg, D. H., et al. 2005, ApJ, 630, 1
- Zucca, E., Ilbert, O., Bardelli, S., et al. 2006, A&A, 455, 879