ESO Imaging Survey. The stellar catalogue in the Chandra deep field south

Stellar catalogues in five passbands ( UBVRI ) over an area of approximately 0.3 deg 2 , comprising about 1200 objects, and in seven passbands ( UBVRIJK ) over approximately 0.1 deg 2 , comprising about 400 objects, in the direction of the Chandra Deep Field South are presented. The 90% completeness level of the number counts is reached at approximately U = 23.8, B = 24.0, V = 23.5, R = 23.0, I = 21.0, J = 20.5, K = 19.0. These multi-band catalogues have been produced from publicly available, single passband catalogues. A scheme is presented to select point sources from these catalogues, by combining the SExtractor parameter class_star from all available passbands. Probable QSOs and unresolved galaxies are identified by using the previously developed $\chi^2$-technique (Hatziminaoglou et al. 2002), that fits the overall spectral energy distributions to template spectra and determines the best fitting template. Approximately 15% of true galaxies are misclassified as stars by the $\chi^2$ method. The number of unresolved galaxies and QSOs identified by the $\chi^2$-technique, allows us to estimate that the remaining level of contamination by such objects is at the level of 2.4% of the number of stars. The fraction of missing stars being incorrectly removed as QSOs or unresolved galaxies is estimated to be similar. The observed number counts, colour-magnitude diagrams, colour-colour diagrams and colour distributions are presented and, to judge the quality of the data, compared to simulations based on the predictions of a Galactic Model convolved with the estimated completeness functions and the error model used to describe the photometric errors of the data. By identifying outliers in specific colour-colour diagrams between data and model the level of contamination by QSOs is alternatively estimated to be $\la$6.3% in the seven passband and $\la$2.3% in the five passband catalogue. This, however, depends on the exact definition of an outlier and the accuracy of the representation of the colours by the simulations. The comparison of the colour-magnitude diagrams, colour-colour diagrams and colour distributions show, in general, a good agreement between observations and models at the level of less than 0.1 mag; the largest discrepancies being a colour shift in $V-R$ and $R-I$ of order of 0.2 mag possibly due to uncertainties in the bolometric corrections. Although no attempt is made to fit the model to the data, a comparison shows that the lognormal law for the initial mass function proposed by Chabrier (2001) describes the data better than the power law form in that paper. The resulting stellar catalogues and the objects identified as likely QSOs and unresolved galaxies with coordinates, observed magnitudes with errors and assigned spectral types by the $\chi^2$-technique are presented and are publicly available.