2 Observations and data reduction

2.1 The observations and the sample

All data used in this paper were obtained in imaging mode with DFOSC at the ESO 1.54 m Danish telescope (La Silla - Chile) during two observing runs (March 1999 and March 2000) blessed by dark time and photometric conditions. The CCD (LORAL/LESSER C1 W7) has $2052 \times 2052$ pixels, each pixel covering $0.39^{\prime\prime}$, corresponding to a field of $\sim$ $13.3^{\prime} \times 13.3^{\prime}$. Data were taken in the g, r and i filters of the Thuan & Gunn system (Thuan & Gunn 1976; Wade et al. 1979). Seeing averaged $\sim$ $1.2^{\prime\prime}$ and was always better than $1.5^{\prime\prime}$. The slight difference between our setup and the original Thuan & Gunn filters resulted in a significant color correction which had to be taken into account in the calibration procedure. Exposure times ranged from 40 to 50 min in the g band and 20 to 30 min in the r and i bands, depending on the target. In order to obtain higher S/N and more accurate photometric measurements, exposures for those clusters at intermediate redshift were usually repeated in two or three slightly offset frames. The observed fields (all located in a region with $0^{\circ} < \delta < +12^{\circ}$ and therefore observable from both hemispheres) included 12 candidate clusters plus 8 clusters with known redshifts to be used as comparison sample. In order to use the same material to both calibrate the corresponding DPOSS fields and to validate our algorithm, we selected galaxy overdensities in such a way that we had at least two (up to four) candidates and/or clusters in each DPOSS field. One candidate, observed on two different nights, also provided an independent check of the photometric accuracy of the second run. For the comparison sample we selected 8 clusters from the X-ray selected sample of Gioia & Luppino (1994) and Ebeling et al. (1996). The equatorial coordinates of the observed fields are given in Tables 1 and 2. We wish to stress that one of the main problems encountered in our work was the well known lack of a suitable set of photometric standards for the Gunn-Thuan system which, along with the lack of faint stars suitable for CCD observations, very often prevents good coverage of the airmass-color plane. The problem is even worse for observers in the Southern hemisphere where the number of available standards is uncomfortably small. We succeeded, however, in observing an average of 4-5 standard stars per observing night.

2.2 Data reduction and photometric calibration

The raw images (both scientific and calibration) were prereduced using the standard procedures available in the IRAF package. First, the frames were corrected for instrumental effects (overscan and bias) and flat fielded. Individual dome and sky flats in each filter were median stacked to increase the S/N ratio. For the first run, flatfielding was performed using sky flats only, but the experience gained in this run suggested a slightly different procedure for the second run, using dome flats to achieve better correction of the small scale pixel-to-pixel variations. Dome flats were first used to correct the sky flat frames for the higher frequency fluctuations, and the resulting frames were then smoothed and stacked to map the lower frequency fluctuations and combined with the average dome flat to produce the final master flats. In the first run, we divided the exposures into two or three frames for the same field; in these cases, the images were combined in each filter by medianing (three exposures) or averaging (two exposures) the aligned frames. Standard star photometry was performed using the apphot package in IRAF. Due to the need to defocus most of the stars to avoid saturation, stars were measured through 10 apertures with diameters up to 90 pixels ( $35.1^{\prime\prime}$), and the local sky was determined using a 10 pixel wide annulus outside of the largest aperture. In order to determine the zero-point offset and the airmass and color terms we used the 40 pixels ($\simeq$ $15^{\prime\prime}$) aperture, for the focused and unsaturated stars, and the asymptotic magnitudes for the defocused ones. The IRAF fitparams task was used to fit the data with the relation

$${m_{\rm true} = m_{\rm inst} + Z_{\rm p} + K_{\rm e} \cdot X + CI \cdot {\rm color}_{\rm inst}}$$ (1)

where $Z_{\rm p}$ is the zero point of the magnitude scale; $K_{\rm e}$ is the extinction coefficient and X the airmass; CI is the instrumental color term coefficient. For the first run (March 1999), due to the paucity of standard stars, we could determine single night coefficients only for the r and i filters. The coefficients were consistent from one night to the other and we used a mean fit for the g-r color. For the second run (March 2000) we instead derived the coefficients for each night and in each band (using the g-r and r-i colors). The resulting calibration coefficients of the various nights are consistent within the errors and, therefore, in order to improve the quality of the fit, we adopted a unique pair of extinction coefficient and color term for the whole run. These constants were then used to derive the zero points for each night. In Fig. 2 we show for the g and r filters, the fit residuals as a function of the estimated magnitude, using different symbols for different nights.

Figure 2: Residuals for the g (upper panel) and r (lower panel) fit. Different symbols refer to different nights. The outlier point is Ross 683, which turns out to be consistently brighter than expected in all four observed nights.

Figure 3: Comparison between the photometry of the candidate cluster 64_781 between nights 2 and 4: the plot shows the g or r magnitude offset vs. the g or r magnitude.

Figure 4: Color-color diagram for the Gunn & Thuan standards (open circles) and for a sample of stars (filled symbols) extracted from the field 5_778.

2.3 Object detection and photometry

The object catalogs were produced individually for each band using S-Extractor: all objects larger than 4 pixels and $2\sigma$ above the background counts were included and their photometric and morphological features measured. We used a photometric reference aperture with a diameter $\sim$3 times larger than the average seeing. For each CCD field, the three single band catalogs where matched taking into account the shifts between pointings (measured using the geomap and geotrans IRAF tasks). To obtain an estimate of the external photometric errors, the candidate cluster 64_781 was observed on two different nights. In this way we could evaluate possible night-to-night magnitude offsets in both the g and r filters. The typical weighted mean values for these offsets are 0.05 for the g filter and 0.007 for the r filter, i.e. they are of the same order as the rms errors from the three parameter calibration fit (Fig. 3). Since our goals require high accuracy for the color determination, we further checked the photometric calibration, using the following test: in the color-color diagram (Fig. 4) we plotted the linear sequence of the Gunn-Thuan standards (open circles) together with all the unsaturated stars (S-Extractor stellarity index <0.8) within the limiting magnitude, selected from some CCD cluster candidate fields (in Fig. 4 we show the 5_778 field). For all of these fields, the sequence of the selected stars is linear (excluding the very red sources, dominated by stars of spectral type M, which have a constant g-r color while r-i depends on the spectral subtype; see Fukugita et al. 1996; Finlator et al. 2000) and overlap quite well with the standard sequence. This means that the colors of the main sequence stars are well determined, since the relation between g-r and r-i is the same for the cluster field stars and for the standards.