For many years, studies of the galaxy formation and evolution were largely based on theoretical expectations because of the lack of an observationally identified population of galaxies at . This situation has recently changed owing to high quality data obtained from HST and ground based telescopes (Keck, NTT, CFHT), that are providing new exciting information about abundance and morphology of galaxies in a wide redshift range up to . In particular, the ongoing development of spectroscopic surveys of relatively bright complete samples limited at B<24.5, I<22 or K<19.5 (Lilly et al. 1996, ApJ, 460, L1; Cowie et al 1996 AJ 112, 839) is now providing a fair galaxy sample up to .
The combined informations from deep galaxy counts and spectroscopic surveys have recently shown that the cosmological star-formation rate increases steadily up to , to decline again at , where is comparable to the present-day value. Thus, most of the stars in the Universe appear to have formed at intermediate redshifts. However, quite different cosmological scenarios (such as a high merging rate or a fading away of a population of dwarf galaxies, or open cosmologies) may still explain the excess of blue faint star-forming galaxies at intermediate redshifts: solving this controversy is one of the major challenges to present-day cosmology.
Redder galaxies at intermediate redshift are other valuable cosmological probes, since the comparison between their IR spectral properties with the predictions of population synthesis models may lead to constraints on the models themselves and on the epoch of galaxy formation.
It is worth emphasizing that the published surveys have required large integration times at the largest telescopes, and cannot be conceivably extended to fainter magnitudes with the planned spectrograph. The LRIS instrument at Keck can reach a limiting magnitude but has still a limited field. FORS1 at the VLT will have similar capabilities. Large field spectroscopic instruments are under construction both for Keck (DEIMOS) and the VLT (VIMOS,NIRMOS). They both aim at studying galaxy population and large scale structure down to the limit of spectroscopy.
Crucials tests to discriminate among the different theories of galaxy formation and evolution could be provided by the extension to fainter magnitudes of the redshift surveys, to have direct informations on the faint end of the luminosity function at intermediate and high redshifts and to compare the predicted N(z) counts in magnitude bins with the observed ones.
A new approach recently developed uses deep multicolor surveys to study the faint magnitude galaxies. Deep multiband images are taken with a complete set of standard broad-band filters, in order to cover the overall spectrum of the galaxy and to discriminate the populations at different redshifts. A well-known example is the successful detection of galaxies using the dropout in their UV flux due to the intergalactic medium and the intrinsic Lyman Limit (Steidel et al 1995, AJ, 110, 2519; Steidel et al , 1996, ApJL, 462, 17; Madau et al 1996 MNRAS, in press; Connolly et al. 1997, ApJ, 486, L11; Giallongo et al. 1998 AJ in press). To perform these detection, an extreme sensitivity in the UV part of the spectrum is required. Indeed, magnitudes as faint as mU = 27 or mB = 28 must be reached to identify the high redshift galaxies that are typically found at .
A different approach, that exploits all the information contained in multiband catalogs, is to compare the observed colors with those predicted by spectral population synthesis codes with a best- fitting procedure, like minimization or Maximuum Likelihood Analysis. This `` photometric redshift'' approach is known to work successfully on low redshift bright galaxy samples (Connolly et al 1995 AJ, 110, 2655) and has been extensively adopted on the HDF deep counts (Lanzetta et al 1996, Nature, 381, 759; Sawicki et al 1997 AJ, 113, 1; Giallongo et al. 1998, AJ, 115 in press)). It allows to take the photometric errors into account and to obtain an indication of the statistical reliability of the redshift identification.
The photometric redshift is measured essentially by identifying the wavelength position of (at least one) spectral breaks, and by using the other colors for a consistency check and to derive an estimate of age and reddening. The wavelength sampling allowed by broad- band filters, usually about 1000 Å, is sufficiently adequate to obtain reliable redshift estimates. Some contamination may arise from the misinterpretation of the observed breaks: broad-band filters don't allow in some cases to disentangle effectively between spectral shapes dominated by redshifted Lyman break, 2800 Å or 4000 Å breaks. However, preliminary results from the spectroscopic follow-up of the HDF deep counts show that a typical redshift error of about is to be expected, and some misidentifications have been reported.
Thanks to the complete spectral coverage, several additional information can be simultaneously deduced for all the objects, such as ongoing star-formation rate, total mass in stars, and hints on the details of the IMF and on the age of the stellar population. For example, Giallongo et al. 1998 have obtained photometric redshifts from the galaxies selected in an NTT field centered on the high redshift (z=4.7) quasar BR 1202-0725. The wide spectral coverage obtained from deep BVRIK multicolor photometry has allowed a reliable redshift estimate for each galaxy. This has been obtained comparing the observed colors with those predicted by spectral synthesis models including UV absorption by the IGM and dust reddening. The main results can be summarized in the following: 1) The redshift distribution of the galaxies is peaked at z=0.6 with 16% of the sample at z>1.5. The derived surface density of the galaxies having in our field is 1.6 arcmin-2 in agreement with that derived in the HDF (1.5 arcmin-2) at about the same magnitude. This corresponds to a comoving volume density of 10-3Mpc-3 similar to the local density of galaxies with the same luminosities. The derived surface density at in our field is lower, 1 arcmin-2. 2) The estimated luminosity function at shows a strong steepening for MBAB>-19with respect to the extrapolation derived from brighter redshift surveys. A comoving volume density of Mpc-3 at MBAB=-17.5 is obtained. Comparing with the local luminosity function, a luminosity evolution by about 2 magnitudes is suggested for galaxies with MBAB>-19. 3) The bulk of the intermediate redshift population mostly consists of very young star-forming galaxies with a median age yr and a small stellar mass M. In particular, the blue fraction with B'-I'<1.4 shows a median age of yr and stellar mass M. 3) The observed 2800 Å luminosity density and the associated star formation rate in our sample show an increase to W Hz-1 Mpc-3 (or M yr-1 Mpc-3) at , i.e. only by a factor 2.5 larger than the local value. At z>1 the UV luminosity density and the corresponding SFR decrease to values comparable to the local one. Thus evidence of a marked maximum in the luminosity density and SFR at appears blurred especially if we consider that an significant corrections for fainter undetected galaxies are expected at z>1. A comparison between the average cosmological luminosity density and the corresponding star formation rate at z=0.4-1 implies an average , adopting the Calzetti (1997) attenuation law and a Miller-Scalo IMF.
The results of this example show that galaxies (which are at the limit for spectroscopic investigation on 10m class telescopes) have a surface density <1 per arcmin2 at z<0.4 and/or at z>1.5 in a binsize. Thus a large and deep multicolor imaging is needed to adequately sample the luminosity function in these redshift intervals. A large area is needed to collect a significant bright (R=24-25) subsample for spectroscopic follow-up, while deep photometry (R=25-27) is needed to sample at any z the faint end of the luminosity function with the photometric redshift technique.
To overcome the uncertainties inherent to the use of broad-band filters, and to achieve a higher accuracy in the photometric redshifts, a dedicated set of intermediate-band filters, of 30-40 nm of bandpass, could be used to follow in a more accurate way the overall spectral shape of the galaxy from the near IR to the UV band. Preliminary simulation show that this sampling allows a much higher accuracy in the redshift estimate ( depending on bandwidth) and a significant reduction of possible misidentifications.
It is important to realize that the typical field of view of present images at large telescopes sample few comoving Mpc of transverse size at cosmological redshifts (e.g. 5' corresponds to about 2.4 Mpc at z=0.7 for H0=50 km s-1 Mpc-1 and q0=0.5). At this scale, large scale structures may dominate the statistical errors in faint counts (Glazebrook et al 1994, MNRAS, 266,65). At the redshift accuracy obtainable with intermediate band filters, wider angular fields must be covered to smooth out the large scale structure noise and to obtain reliable average quantities in each redshift bin.
Moreover, the possibility of covering with the intermediate-band filters sufficiently wide areas, with extensions of the order of or larger than the degree scale, would allow to study the three dimensional large scale structure (LSS) at magnitudes significantly fainter than those reachable with planned, future spectroscopic surveys. For example, the VIRMOS deep survey will cover with spectroscopy (about 50,000 redshifts) galaxies down to I = 24 in about one square degree. Most of these galaxies are expected to be at z<1. The wide field LBT photometric survey will allow both to measure the amplitude of the LSS in the linear regime (r 50 Mpc) at higher redshift and to compare the clustering properties of low luminosity galaxies with those of higher luminosity galaxies at z< 1. Both these results are crucial for discriminating about different cosmological models, since the growth term of the structure is directly related to . With the same data, thanks to the accurate photometric redshift, we will also be able to collect a sizable sample of rich clusters up to z> 1. A statistically controlled sample of clusters at the highest possible redshift is important for cosmological studies, because the number of clusters at high z is a strong function of .
Summarizing, the data collected in a multicolor survey could be used to obtain samples of galaxies at the highest redshifts with an unprecedented statistics, and to study the physics and the spatial distributions of galaxies at intermediate redshifts well beyond the spectroscopic capabilities.
A wide-field double channel imager at LBT, which is highly efficient
from the UV to the NIR J band, is the detector of choice to perform
deep galaxy surveys for the following reasons:
- The use of intermediate-band filters is very time demanding,
and requires the large collecting area of an 8m telescope to reach the
faintest magnitudes (
that are beyond the spectroscopic
- Many interesting targets (e.g. galaxies at or
) have a very low surface density (
per arcmin-2): a large field of view is needed to collect a
significant statistics of these rare objects minimizing the effects of
field-to-field variations and to select the few among them that are
within reach of spectroscopy for more detailed studies.
- Large transverse sizes must be sampled to study the large
scale fluctuations of the faint galaxies;
- A good image quality and sampling are needed to perform an
accurate photometry and avoid confusion at the faintest limits.
- Extreme sensitivity in the U and B bands is essential for both the search of high redshift galaxies, that can be selected through their Lyman Limit absorption, and to derive the star-formation rate of galaxies at low and intermediate redshifts.
- At the same time the NIR wide field facility of the instrument can allow deep and wide surveys at 1 m and beyond to select old red galaxies or dust-reddened galaxies. The full spectral coverage of the instrument from the UV to the J band allow to discriminate between these two populations.