1 Introduction

Deep fields have become a favourite tool of observational cosmology, particularly in conjunction with the construction of multiwavelength datasets. Besides the ubiquitous Hubble Deep Fields, another illustrative example is the ROSAT Deep Survey in the Lockman Hole (Hasinger et al. 1998) and its optical follow-up by imaging and spectroscopy (e.g., Schmidt et al. 1998). However, full spectroscopic coverage is usually impossible to obtain, and the reconstruction of distances often has to rely on photometric redshift estimation. Despite major advances in this area, there is always a certain degree of degeneracy between object classes, such as different galaxy types, and their redshifts. Breaking this degeneracy is possible only by incorporating independent spectroscopic information, such as adding infrared colors or moving towards finer spectrophotometric resolution.

Wolf et al. (2001a,b) demonstrated that the use of medium-band filters leads to a substantial gain in classification accuracy and discriminative power, as compared to simple broad-band photometry. In 1999 we initiated a new multicolor survey using the Wide Field Imager (WFI, Baade et al. 1998, 1999) at the MPG/ESO 2.2 m telescope on La Silla, Chile. The survey was designed to make full use of the capabilities offered by the WFI and at the same time exploit the experience collected in the course of earlier multicolor projects. By incorporating 17 different optical filters into our new survey the COMBO-17 (Classifying Objects by Medium-Band Observations in 17 filters) project goes a major step beyond the traditional multicolor approach. It permits confident spectral classification of objects with into stars, galaxies, quasars, and the recognition of exotic objects, and it facilitates accurate subclassification, redshifts estimation, and SED reconstruction for galaxies and quasars. The survey area comprised four independent WFI fields amounting to a total area of 1 deg². Overall objectives of COMBO-17 are:

• The galaxy catalogue will be complete to R < 24 and contain $\sim$50000 galaxies with accurate redshifts and spectral classification up to $z \simeq 1.5$, to be used to study the evolution of the galaxy luminosity function and clustering properties;
• The quasar sample will contain $\sim$500 QSOs, with nearly uniform completeness over the redshift range , well suited to trace the turnover in QSO evolution;
• Deep high-resolution R-band images with $0\hbox{$.\!\!^{\prime\prime}$ }75$ FWHM permit morphological classification and gravitational lensing studies.

In this paper we focus on the "Chandra Deep Field South'' (Giacconi et al. 2000) subject to a deep X-ray exposure by the Chandra satellite observatory and target of a wide range of multiwavelength studies from several groups. It is particularly worth noting that the COMBO-17 field of view contains the Chandra field entirely facilitating the efficient combination of optical and X-ray data. In this paper, we present a first catalogue of optically selected sources down to R-band magnitudes of . This catalogue is publically available over the world-wide web.