New spectroscopic redshifts from the CDFS and a test of the cosmological relevance of the GOODS-South field

Context. This paper prepares a series of papers analysing the Intermediate MAss Galaxy Evolution Sequence ( IMAGES) up to a redshift of one. Intermediate mass galaxies ( M-J <= - 20.3) are selected from the Chandra Deep Field South ( CDFS) for which we identify a serious lack of spectroscopically determined redshifts. Aims. Our primary aim in this study is therefore to obtain a sample of intermediate-mass galaxies with known spectroscopic redshift to be used for further analysis of their 3D-kinematics. We also intend to test whether this important cosmological field may be significantly affected by cosmic variance. Methods. The spectroscopic observations were carried out using VIMOS on the ESO VLT. The data reduction was done using a set of semi-automatic IRAF procedures developed by our team. Results. We have spectroscopically identified 691 objects including 580 galaxies, 7 QSOs, and 104 stars. The overall completeness achieved is about 76% for objects with I-AB <= 23.5 in the CDFS after excluding instrumental failures. This study provides 531 new redshifts in the CDFS. It confirms the presence of several large-scale structures in the CDFS, which are found to be more prominent than in other redshift surveys. To test the impact of these structures in the GOODS-South field, we constructed a representative redshift catalog of 640 galaxies with I-AB <= 23.5 and z <= 1. We then compared the evolution of rest-frame U, B, V, and K galaxy luminosity densities to the one derived from the Canada France Redshift Survey ( CFRS). The GOODS South field shows a significant excess of luminosity densities in the z = 0.5 - 0.75 range, which increases with the wavelength, reaching up to 0.5 dex at 2.1 mu m. Stellar mass and specific star formation evolutions might be significantly affected by the presence of the peculiar large-scale structures at z = 0.668 and at z = 0.735, which contain a significant excess of evolved, massive galaxies when compared to other fields. Conclusions. This leads to a clear warning about the results based on the CDFS/GOODS South fields, especially those related to the evolution of red luminosity densities, i.e. stellar mass density and specific star formation rate. Photometric redshift techniques, when applied to that field, are producing quantities that are apparently less affected by cosmic variance ( 0.25 dex at 2.1 mu m), however, at the cost of any ease in disentangling the evolutionary and cosmic variance effects.