2 Observations and data reduction

The I and Z observations were made using the CFH12K mosaic camera (Cuillandre et al. 2001) at the prime focus of the Canada-France-Hawaii telescope, from December 17 to 19, 2000. The camera is equipped with an array of twelve $2048\times4096$ pixel CCDs with $15~{\mu}{\rm m}$ size pixels, yielding a scale of 0.206''/pixel and a field of view of 42'$\times$28'. The I Mould and Z filter profiles are shown in Fig. 1. We obtained images of 17 fields in the Pleiades whose coordinates are given in Table 2 and the location on the sky relative to the brightest Pleiades members shown in Fig. 2. Fields were selected in order to avoid the brightest stars which would produce reflection halos and scattered light on the detectors, i.e., a higher and non uniform background which is difficult to remove. We also avoided the southwestern region of the cluster where a small CO cloud causes relatively large extinction. The observing conditions were photometric with a I-band seeing of typically 0.5'' FWHM, but occasionally up to 1.1'' FWHM.

Figure 1: Filter profiles. The solid line corresponds to the I Mould filter and the dashed line to the Z filter. The cut-off in the Z-band is imposed by the detector pass band.

   

Table 2: Coordinates of the 17 different pointings observed with the CFH12K camera.

Field	RA (2000)	Dec (2000)
	(h m s)	($^{\circ}$ ' '')
Pl-B	03:51:55	24:47:00
Pl-C	03:51:55	24:14:00
Pl-D	03:48:00	25:20:00
Pl-E	03:48:00	26:00:00
Pl-G	03:54:00	26:00:00
Pl-H	03:50:30	26:40:00
Pl-I	03:55:00	26:30:00
Pl-J	03:46:00	26:30:00
Pl-K	03:45:00	25:38 30
Pl-L	03:45:25	25:08:00
Pl-O	03:40:50	26:08:00
Pl-P	03:42:20	24:40:00
Pl-Q	03:55:00	24:47:00
Pl-R	03:55:00	24:20:00
Pl-T	03:52:40	23:25:00
Pl-U	03:55:45	23:52:00
Pl-V	03:55:45	23:28:00

Figure 2: Location of the 17 fields relative to the bright Pleiades members shown as stars. The size of the symbols depends on the brightness of the sources. The units are degrees on both axes. Overplotted are the circles of radii from 0.75 to 3.5 degrees centered on the cluster center. The dots represent the brown dwarf candidates detected in our survey. The total area covered by the survey amounts to 6.4 square degrees and reaches up to 3 degrees from the cluster center.

Short and long exposures were taken for each field in order to cover a large magnitude range. We obtained a set of four images in both Iand Z filters: one 10 s exposure plus three 300 s exposures in the I-band and one of 10 s exposure plus three of 360 s exposures in the Z-band. The short exposures encompass the magnitude range $13.5 \le I \le 21.5$ whereas the long ones cover $17 \le I \le 24$. A comparison between the number of stellar like objects detected in long exposures of overlapping fields indicates that the survey is about 90% complete down to $I\simeq 22$ (see Fig. 3). Note that the completeness limit of the short exposures ( $I \simeq 19.5$) is much larger than the saturation limit of the long exposures ( $I \simeq 17$), so that the survey completely samples the continuous mass range from $0.03~M_{\odot}$ to  $0.45~M_{\odot}$ in the Pleiades cluster.

Figure 3: The plain histograms represent the number of sources per magnitude bin I (top) or Z (bottom) detected in the field Pl-U and located in the region which overlaps with field Pl-V. Hatched histograms correspond to the number of objects detected in both fields. The percentage of matched sources is given and indicates that the survey is about 90% complete down to $I\simeq Z\simeq 22$.

2.1 Photometry

Each mosaic consists of 12 CCD images which were reduced and analysed separately. The images were overscan corrected, debiased and flat-fielded. The flats were normalized to a reference CCD to retain the appropriate relative scaling between chips. The same photometric zero-point can thus be used for all the CCDs of a mosaic. The images were also fringe-corrected using patterns derived from a smoothed combination of more than 20 images in each band. Then, images of similar exposure time for a given field were stacked together using an optimal CCD clipping to remove the artefacts, and then taking the average of the remaining final stack for each pixel to preserve the photometry.

To detect all the sources in the frames, we averaged all long exposure I and Z images of a field previously shifted to the same location and used the automatic object-finding algorithm from the Sextractor package (Bertin & Arnouts 1996). Then, a PSF-fitting photometry on the I and Z images was performed for all the detected objects using the PSFex package. We discriminated between point-like and extended or corrupted sources using the FWHM distribution of all the detected objects on each field. In practice, we defined conservative lower and upper limits around the well defined stellar peak of this distribution and rejected all the sources located outside these boundaries.

Photometric standard Landolt fields SA98 and SA113 were observed and reduced in the same way as the science images. Three successive exposures of SA98 were obtained with an incremental offset of several arcminutes in RA. Thus, common sets of photometric standard stars were observed on every CCD of the mosaic and we checked that the photometric zero-point in I-band was the same for each CCD: we did find a scatter of only 0.03 mag. The photometry in the Z-band is sensitive to detector pass band and Landolt does not give Zmagnitudes for his standards. Thus we selected unreddened A0 standards and set their Z magnitude equal to their I magnitude assuming that their I-Z color was zero (by definition of an A0 star). Thanks to the several exposures of SA98 shifted of a few arcminutes, those A0 stars were observed on different chips so that we could derive Zzero-points for some of the CCDs. We found a scatter of only 0.04 mag and we then used the mean value as a global zero-point, assuming that it could be used for all the CCDs as in the I-band. We verified this assumption a posteriori by measuring the difference between the magnitudes of the same objects detected in the common region of 2 overlapping pointings. We thus verified that there was no systematic error and estimated the photometric rms error up to the completeness limit ( $I\simeq 22$, see Fig. 4, top panel) which amounts to 0.07 mag or less at $I\le 22$.

2.2 Astrometry

In order to obtain accurate coordinates for cluster brown dwarf candidates, we had to derive an astrometric solution for each CCD of the mosaic. We used the CFHT's Elixir package (Magnier & Cuillandre 2002) to compute the astrometric solution for each image. The algorithm calculates the celestial coordinates for all the detected objects using the approximate solution given by the header, compares them with the USNO2 catalog and refines the solution. Most of the USNO2 stars were saturated in our long exposure images so that we had to use lists of refined coordinates derived from the short exposures as a reference catalog.

We estimated the astrometric error by comparing coordinates of stars present in overlapping regions and found an accuracy better than 0.5 arcsec. The astrometric rms error is shown in Fig. 4 (bottom panel) as a function of I magnitude.

Figure 4: Top: Photometric error rms plotted as a function of I magnitude. The dots correspond to the error on the magnitude I, the open triangles to the error on Z and the crosses to the error on I-Z; Bottom: Astrometric error rms shown as a function of I magnitude. The dots correspond to the error on the right ascension and the triangles denote the declination error.