The empirical approach adopted by the astronomers to investigate galaxy formation and evolution is to search for and to study the populations of distant galaxies. Samples selected in the optical bands allow to cull star-forming galaxies where the redshifted ultraviolet (UV) radiation is dominated by hot, massive and short-lived stars. However, optical samples are affected by severe biases due to the strong influence of dust extinction in the UV and to the wide range of shapes that spectral energy distributions (SEDs) have in the UV depending on the level of star formation activity and the age of the galaxy. Instead, at longer wavelengths, the above problems are alleviated, as the rest-frame optical and, even better, the near-infrared radiation is dominated by low mass, long-lived stars. Also, the shapes of the SEDs in the optical/near-IR are very similar for all galaxy types, and the effects of dust extinction become less severe. In addition, the rest-frame optical/near-IR luminosity is known to correlate with the galaxy mass. The above advantages make galaxy samples selected in the near- (e.g. K-band at 2.2mum) or, even better, in the mid-IR (e.g. 4-8mum, now possible with the Spitzer Space Telescope) more suitable than optical samples to investigate galaxy evolution and, particularly, the history of galaxy stellar mass assembly, because they allow to observe the rest-frame optical and near-IR for high redshift galaxies.

Recent results at z<1 and z>3 suggest that most of the galaxy mass was built at z~1-3. Thus, it is important to investigate the nature and the evolutionary status of galaxies in this redshift range in order to constrain their formation and evolution and the theoretical models.

OBSERVATIONAL APPROACH

The z~1-3 is a difficult redshift range because galaxies become rapidly very faint and, in particular, massive and/or dusty galaxies are the most difficult targets for conventional spectroscopy due to their spectral features and the lack or weakness of emission lines. This difficulty is exacerbated in the so called ``redshift desert'' (i.e. z~1.3-2.3) where spectroscopic redshifts are hard to derive, especially for red galaxies. Photometric redshifts provide little clues on the nature and evolutionary status of galaxies, and no constraints on their 3D clustering.

Our approach is to perform ultra-deep ESO VLT+FORS2 multi-object spectroscopy with very long integration times (15-40 hours per mask) of infrared-selected galaxies at z>1.4 selected based on high-quality photometric redshifts. The spectroscopy is done either in the blue or in the red depending on the SEDs of the target galaxies, their zphot and their location in diagnostic color-color diagrams. Due to the selection, depth and redshift coverage, our project is complementary to other surveys such as the ESO/GOODS spectroscopy, the K20 survey and the GDDS.