J. P. U. Fynbo 1,2,3 - C. Ledoux 4 - P. Møller 3 - B. Thomsen 1 - I. Burud 5
1 -
Department of Physics and Astronomy,
University of Århus,
Ny Munkegade, 8000 Århus C, Denmark
2 -
Astronomical Observatory, University of Copenhagen,
Juliane Maries Vej 30, 2100 Copenhagen Ø, Denmark
3 -
European Southern Observatory,
Karl-Schwarzschild-Straße 2,
85748, Garching by München, Germany
4 -
European Southern Observatory, Casilla 19001, Santiago 19,
Chile
5 -
Space Telescope Science Institute, 3700 San Martin Drive,
Baltimore, MD21218, USA
Received 17 March 2003 / Accepted 22 May 2003
Abstract
We present the first results of an observational programme at
the ESO Very Large Telescope aimed at detecting a large sample
of high-redshift galaxies fainter than the current spectroscopic limit of
R=25.5 for Lyman-Break galaxies. In this paper, we describe the results of
deep narrow and broad-band imaging and subsequent follow-up multi-object
spectroscopy of faint high-redshift galaxies in the fields of the QSOs
BRI 1346-0322 and Q 2138-4427. These QSOs have intervening high neutral
hydrogen column density absorbers, at redshifts z=2.85 and z=3.15respectively, for which redshifted Ly
emission falls within less than
a few Å from the central wavelengths of existing VLT (
60 Å-wide)
narrow-band filters. We selected 37 and 27 candidate emission-line galaxies in
the two fields respectively. Most (
85%) of the candidates have R-band
magnitudes fainter than R=25.5. The first spectroscopic follow-up of a
sub-sample of the candidates resulted in 41 confirmed candidates and 4
foreground emission line galaxies (three [O II] emitters and one
C IV emitter). The confirmation rate for Ly
emitters is 82% and 68% in the field of BRI 1346-0322 and Q 2138-4427 respectively.
In addition, we serendipitously detect a number of other
emission-line sources on some of the slitlets not used for candidates. Of
these, 9 are also most likely Ly
emitters with redshifts ranging from 1.98 to 3.47. The redshift distribution of confirmed candidates in the field
of BRI 1346-0322 is consistent with being drawn from a uniform distribution
weighted by the filter response curve, whereas the galaxies in the field of
Q 2138-4427 have redshifts clustering very close to the redshift of the
damped Ly
absorber. This latter fact indicates the existence of a
large "pancake''-like structure confirming the earlier suggestions of
Francis & Hewitt (1993).
Key words: cosmology: observations - quasars: individual BRI 1346-0322, Q 2138-4427 - galaxies: high redshift
Our knowledge of the properties of high-redshift galaxies (here
)
currently mainly comes from two very different kinds of studies,
namely, i) the study of Lyman-Break Galaxies (LBGs) selected by the
Lyman-limit break in their spectrum (e.g. Steidel et al. 1996, 2000;
Fontana et al. 2000; Papovich et al. 2001; Vanzella et al. 2002),
and ii) the study of the chemical and kinematical properties of
(proto)galaxies, so called Damped Ly
Absorbers (DLAs), intervening
the lines of sight to QSOs (e.g. Wolfe et al. 1986; Pettini et al. 1997;
Ledoux et al. 1998; Prochaska & Wolfe 2002; Ledoux et al. 2003). However,
there is strong
evidence that there is only a small overlap between the galaxies in
the current LBG samples and the DLAs (Fynbo et al. 1999;
Haehnelt et al. 2000; Schaye 2001; Møller et al. 2002;
Adelberger et al. 2003). The reason for this is that Lyman-Break galaxy
samples are continuum-flux limited and that the current flux limit
corresponding to
is not deep enough to reach the level of
typical damped Ly
absorption selected galaxies. However, due to
the very steep faint-end slope of the z=3 galaxy luminosity function
(Adelberger & Steidel 2000; Poli et al. 2001a, 2001b)
70% of the
observer rest-frame R-band flux from z=3 galaxies is emitted by galaxies
fainter than R=25.5. Therefore, besides accounting for the bulk of the damped Ly
absorption in QSO spectra, the R>25.5 galaxies could dominate
the integrated star-formation rate and the metal enrichment and heating of
the intergalactic medium at
.
In 2000, we started the programme "Building the Bridge
between Damped Ly
Absorbers and Lyman-break Galaxies: Ly
Selection of Galaxies'' at the European Southern Observatory's Very Large
Telescope (VLT). This project aims at bridging the gap between
absorption- and emission-line selected galaxy populations by creating a
database of
3 galaxies that are fainter than R=25.5, and to study
the properties, such as morphology, star-formation rates, clustering and
continuum colours, of these faint, very numerous, and so far little studied
high-redshift galaxies.
Our method is to obtain deep narrow-band Ly
observations of the fields
of
QSO absorbers whose redshifts match existing VLT narrow-band
filters. We chose to observe the fields of QSO absorbers to anchor our fields
to already known structures at the target redshift and, hence, minimise the
risk of observing a void. Ly
narrow-band imaging can, with
comparatively short integration times, lead to the detection of a significant
sample of
galaxies having much fainter continuum fluxes than LBGs,
as demonstrated by, e.g., Cowie et al. (1998), Kudritzki et al. (2000), and
Fynbo et al. (2000, 2001, 2002). In the following, we shall use the acronym LEGO for Ly
-Emitting Galaxy-building Objects (Møller & Fynbo 2001)
to refer to the Ly
emitters. In this first paper, we describe the
results of deep narrow and broad-band imaging and subsequent follow-up
multi-object spectroscopy of the fields of the QSOs BRI 1346-0322 and
Q 2138-4427. These QSOs have intervening high neutral hydrogen column
density Ly
absorbers at redshifts z=3.15
(
)
and z=2.85 (
)
respectively
(Francis & Hewett 1993; Storrie-Lombardi et al. 1996). The paper is organized
in the following way. First, in Sect. 2 we describe the imaging and the
selection of LEGO candidates. Then, in Sect. 3 we describe the observations and
results from the first spectroscopic run. Finally, in Sect. 4. we discuss our
findings. Throughout this paper we assume a cosmology with
H0=65 km s-1 Mpc-1,
and
.
In this model, a redshift of 3.15(2.85) corresponds to a luminosity
distance
(25.76) Gpc and a distance modulus of
47.31(47.05).
One arcsecond on the sky corresponds to 8.18(8.42) proper kpc and the
look-back time is 12.3(12.1) Gyr (roughly 85% of the time since the event
commonly referred to as the Big Bang).
![]() |
Figure 1: Transmission curves of the two narrow-band and the broad-band Bessel B and R filters used in this study. |
Open with DEXTER |
Table 1:
Log of imaging observations with FORS1. The 5
detection limit
was measured in a circular aperture of
diameter and is given in the
AB system.
The images were reduced (de-biased and corrected for CCD pixel-to-pixel
variations) using the FORS1 pipeline (Grosbøl et al. 1999). The
individual reduced images in each filter were combined using a code that
optimizes the Signal-to-Noise (S/N) ratio for faint, sky-dominated sources
(see Møller & Warren 1993 for details on this code).
![]() |
Figure 2:
The 400 ![]() ![]() |
Open with DEXTER |
The broad-band images were calibrated as part of the FORS1 calibration plan
via observations of Landolt standard stars (Landolt 1992). As the data were
collected over a time span of months, we have data from several photometric
nights which we used to determine independent zeropoints for the combined
images. These zero-points are consistent with each other within 0.02 mag. We
transformed the zero-points to the AB system using the relations given by
Fukugita et al. (1995):
and
.
The Q 2138-4427
images are deeper than the BRI 1346-0322 ones by nearly half a magnitude as
the BRI 1346-0322 data were obtained at higher airmass and with slightly
larger lunar illumination than the Q 2138-4427 data. For the calibration of
the narrow-band images, we used observations of the spectrophotometric
standard stars LTT6248 and LDS749-B for, respectively, the BRI 1346-0322
and Q 2138-4427 fields.
Contour images of the combined narrow-band images of the
arcsec2 fields surrounding the QSOs BRI 1346-0322 and Q 2138-4427 are
shown in Fig. 2. Both QSOs are identified by an "
'' at the
field centre and the positions of selected LEGO candidates (see
Sect. 2.1.1) are shown with boxes.
![]() |
Figure 3:
The top three panels concern the BRI 1346-0322 field and the
lower three panels the Q 2138-4427 field. Left panels: simulated
colour-colour diagram based on Bruzual & Charlot galaxy
SEDs. The filled squares are 0<z<1.5 galaxies with ages from a
few to 15 Gyr and the open triangles are 1.5<z<3.0 galaxies
with ages from a few Myr to 1 Gyr. The dotted box encloses the
simulated galaxy colours. The dashed line indicates colours of
objects with a particular broad-band colour and with an SED in
the narrow-band filter ranging from an absorption line in the upper
right part to an strong emission line in the lower left part of the
diagram. Middle panels:
colour-colour diagrams for all objects detected at S/N>5 in either
the narrow band or the B band. As expected, most objects have colours
consistent with being in the dotted box. Due to the damped Ly![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
For the final selection of LEGO candidates, we used the "narrow minus
on-band broad'' versus "narrow minus off-band broad'' colour/colour
plot technique (Møller & Warren 1993; Fynbo et al. 1999, 2000, 2002 and
Fig. 3). In order to constrain where objects with no special
spectral features in the narrow filter are in the diagram, we calculated
colours based on synthetic galaxy SEDs taken from the
Bruzual & Charlot (1995) models. We have used models with ages ranging from a
few Myr to 15 Gyr and with redshifts from 0 to 1.5 (open squares in
Fig. 3) and models with ages ranging from a few Myr to 1 Gyr with
redshifts from 1.5 to 3.0 (open triangles). For the colours of high-redshift
galaxies, we included the effect of Ly
blanketing
(Møller & Jakobsen 1990; Madau 1995). Figure 3 shows the
versus
colour diagram for the simulated galaxy
colours (left panels) and for the observed sources in the two target fields
(middle and right panels). The dashed line indicates where objects with a
particular broad-band colour and either absorption (upper right) or emission
(lower left) in the narrow filter will fall.
In the middle panel, we show the colour-colour diagram for all of the objects
detected in the two fields. Due to the damped Ly
line, Q 2138-4427
has a large positive
colour and is hence seen in the upper right
corner. Due to Lyman-forest blanketing, the B-band flux of BRI 1346-0322 is
suppressed hence decreasing its
colour. In the lower left part of
the diagram, a large group of objects are found to lie significantly away from
the locus of continuum objects. In the right panel, we show with the solid
line the median
colour in the range
.
The
two dotted lines show the 98% and 2% percentiles in the
colour. We selected as emission line sources objects with S/N>5 in the
narrow-band image and whose 1
upper limit on
is below
the 98% line. We detect 27 and 37 such objects in the BRI 1346-0322 and
Q 2138-4427 fields respectively which we consider as LEGO candidates in the
following. The colours of the candidates are shown again in the two right
panels.
Table 2: Log of spectroscopic observations with FORS1.
![]() |
Figure 4:
Our selection criterion is illustrated by plotting the line
equivalent width against the Ly![]() ![]() |
Open with DEXTER |
The MOS data were reduced and combined as described in Fynbo et al. (2001).
The accuracy in the wavelength calibration is about 0.1 pixel for
a spectral resolution R=900, which translates to
.
Average object extraction was performed within a variable window size
matching the spatial extension of the emission line(s). Therefore, the
flux should be conserved. When two or more individual exposures on the
same target have been obtained through different masks, the spectra
were appropriately combined with rescaling and weights, and using a mask for
rejecting cosmic ray impacts.
We first analysed the slitlets containing the spectra of the LEGO candidates.
The combined spectra are displayed in Figs. 5
and 6. Out of the 27 candidates in the BRI 1346-0322 field, we confirm 20 as being emission-line objects. We consider a candidate
confirmed if there is an emission line detected with at least 3significance at the correct position in the slitlet within the wavelength
range corresponding to the filter transmission. Five candidates were
not observed and for the remaining two an emission line is not detected. Two
of the confirmed emission-line sources are foreground galaxies with
the [O II] line located in the narrow-band filter. The overall
efficiency for detection and confirmation of LEGOs is therefore
#(confirmed LEGOS)/#(observed LEGOs) = 18/22 = 82% so
far. For the Q 2138-4427 field, the overall efficiency is smaller.
Three candidates were not observed and 9 observed candidates were
not confirmed. For the remaining 25, we confirm the presence of an
emission line. Two of the confirmed emission-line sources are
foreground objects. One is an [O II] emitter and the other is
a z=2.0364 AGN with C IV located in the narrow-band filter. Hence,
the fraction of confirmed LEGOs is 23/34=68%.
![]() |
Figure 5:
![]() |
Open with DEXTER |
For the confirmed LEGOs, we derived Ly
fluxes, equivalent
widths (EWs) and star-formation rates (SFRs) as described in detail
in Fynbo et al. (2002).
![]() |
Figure 6:
![]() |
Open with DEXTER |
![]() |
Figure 6: continued. |
Open with DEXTER |
Based on observed Ly
fluxes, the range of SFRs for the confirmed LEGOs
is 0.20-15
yr-1 if the extinction is negligible. The observed EWs range from less than 100 Å to more than 1000 Å. This corresponds to
about 20 Å to 250 Å in the rest-frame and is consistent with
the theoretically expected Ly
EWs for metal-poor starburst galaxies
(Charlot & Fall 1993; Valls-Gabaud 1993; Schaerer 2003).
The surface density of confirmed sources is of the order of 10 arcmin-2per unit redshift down to a Ly
flux limit of
erg s-1 cm-1 and the EW limit shown in
Fig. 4. This is about a factor of five higher than the
surface density of LBGs down to R=25.5,
(Steidel et al. 1999) even if most LBGs are not themselves Ly
emitters. In other
words, the LBGs are the tip of an iceberg consistent with the conclusion of
Fynbo et al. (1999). This reflects the steepness of the luminosity
function for z=3 galaxies.
In future papers, we will address the morphology, clustering properties and luminosity function of the LEGOs from this and complementary VLT surveys.
In each mask about half of the available 19 slitlets could be used
on candidates. The remaining slitlets were placed either on stars to check
the slit alignment, on faint galaxies, or on regions of blank sky.
We have carefully analyzed all the 2D-spectra for other emission line
sources that could be serendipitously detected LEGOs at
other redshifts (see e.g. Manning et al. 2000 for other such
cases). In addition to a number of [O II] emitters we detect
a significant number of emission line sources that are most likely
Ly
emitters based on their extremely faint continua and
the lack of other lines in the spectra. The spectra and images of these
are shown in Fig. 10. One of the sources, called
LEGO1346_2b, a is a neighbour to the z=3.1301 LEGO1346_2 and has a
redshift within the narrow filter used for the BRI 1346-0322 field,
z=3.1251.
However, in the narrow-band image it is fainter than our 5
cut.
The serendipitously detected LEGOs have redshifts ranging from 1.98 to 3.47.
![]() |
Figure 7:
Histogram of R(AB) magnitudes of 45 confirmed emission-line sources.
Objects detected at less than 5![]() |
Open with DEXTER |
It is interesting to ask how the LEGOs compare with the LBGs at the
bright end of the luminosity function. Given the importance of dust for the
escape of Ly
emission from galaxies the naive expectation
would be that Ly
emitters are drawn from the younger, less
chemically evolved and in general faint end of the population of
high-redshift galaxies (Fynbo et al. 2001; Malhotra & Rhoads 2002).
Contrary to this expectation, Shapley et al. (2001) suggest, based on
rest-frame optical properties of LBGs, that the Ly
emitting
sub-sample of the LBGs are the oldest objects among the Lyman-break
selected galaxies. In this scenario LBGs start out in a dusty burst
phase for 50-100 Myr followed by a more quiescent phase with less
extinction for several hundred Myr up to a Gyr. However,
most of the Ly
emitters in our samples are too faint to be included
in the LBG samples and it is therefore likely that the properties of the Ly
emitting LBGs are different from LEGOs in general.
![]() |
Figure 8: Redshift distributions of LEGOs in the two fields relative to the filter response curves. The redshifts of LEGOs in the field of BRI 1346-0322 fill out the volume probed by the filter, whereas LEGOs in the field of Q 2138-4427 have a very narrow redshift distribution. |
Open with DEXTER |
During most of the previous decade Ly
emission was considered
an inefficient survey method for high-redshift galaxies due to a number
of unsuccessful surveys (e.g. Prichet 1994 and references therein). It is
now clear that the first surveys for Ly
emitters were unsuccessful
mainly because they reached significantly too shallow detection limits.
The theoretical expectation was that todays large ellipticals formed
in a fast, monolithic collapse (e.g. Patridge & Peebles 1967). In the
hierarchical picture of galaxy formation the high-redshift galaxies are
smaller and hence fainter than expected when the first surveys were
planned.
![]() |
Figure 9:
Shown are
![]() ![]() |
Open with DEXTER |
![]() |
Figure 10:
Shown are
![]() |
Open with DEXTER |
The main advantage of LBG surveys is that they probe a very large volume and hence provide a large number of galaxies per field. However, there is a number of studies that are most efficiently done with LEGOs as probes: i) LEGOs can be used to probe the faint end of the luminosity function (Fynbo et al. 2001); ii) LEGOs can be detected and spectroscopically confirmed at both lower (Fynbo et al. 1999, 2002) and higher redshifts (Dey et al. 1998; Ellis et al. 2001; Venemans et al. 2002; Hu et al. 2002; Taniguchi et al. 2003) than is currently possible with techniques based on the continuum; iii) the large space density reachable with surveys for LEGOs allows a detailed study of the underlying large scale structure and to probe the environments of other high-redshift objects such as radio galaxies (Kurk et al. 2000; Venemans et al. 2002), Gamma Ray Burst host galaxies (Fynbo et al. 2002) or QSO absorbers (e.g. Møller & Warren 1993; Francis et al. 1995; and this paper).
The optimal way to proceed with Ly
surveys seems to be the use
of large area cameras on
8-m class telescopes. First results regarding the luminosity function and
clustering properties of LEGOs using the Suprime Camera (Miyazaki et al.
2002) on SUBARU has been reported by
(Ouchi et al. 2003), however that survey is not as deep as the survey
presented here and it targets a higher redshift (z=4.86). A narrow-band image
targeting z=3 LEGOs to the same depth as in the present survey, but obtained
with the Suprime Camera will provide of order 500 candidates per field.
Furthermore, it is mandatory that candidates based on narrow-band imaging are
subsequently confirmed (or rejected) based on spectroscopy to make sure that
conclusions based on surveys of LEGOs can be trusted. A major reason for the
success of the Lyman-Break surveys is the spectroscopic confirmation of most
of their candidates. Studies of LEGOs should follow this good example.
Furthermore, the measurement of several hundred redshifts in one field can be
used to map out the underlying filamentary structure
(Møller & Fynbo 2001) and even provide an independent measurement of the
cosmological constant (Weidinger et al. 2002).
Acknowledgements
We thank our referee C. Steidel for comments that helped us improve the discussion and the Paranal staff for excellent support during the visitor run in July 2002. JPUF and CL gratefully acknowledge the receipt of an ESO research fellowship. We acknowledge helpful discussions with Bruno Leibundgut, Nicolas Bouche, Masami Ouchi, Vincenzo Mainieri and Max Pettini. JPUF gratefully acknowledge support from the Carlsberg Foundation.