2 Observations and data reduction

2.1 Region surveyed by ISOCAM

Figure 1: Sky map of the $\mbox{$\rho$ ~Ophiuchi}$ clouds L1688, L1689N, and L1689S showing the spatial distribution of the ISOCAM sources. The filled symbols correspond to "red'' YSOs (i.e. Class I and Class II YSOs, see Sect. 3), while the open circles mark the "blue'' sources. The surface areas of these symbols are proportional to the 6.7 $\mu$m flux densities. The open stars mark four young early-type stars (SR3, S1, WL22, WL16 from right to left; see Sect. 3.1). The contours show the L1688 molecular cloud as mapped in CS(2-1) by Liseau et al. (1995). The triangles refer to the peaks of the DCO+ dense cores of Loren et al. (1990). The figure emphasizes the presence of four sub-clusters Oph A, B, EF (between Oph E and Oph F), and L1689S (named after the associated dense cores).

The survey encompasses the ${\rho}$ Ophiuchi central region associated with the prominent dark cloud L1688, as well as the two subsidiary sites L1689N and L1689S. The L1688 field is a $\sim$ $45\hbox{$^\prime$ }\times 45\hbox{$^\prime$ }$ square, while the L1689N and L1689S fields each cover an area of $16.5\hbox{$^\prime$ }\times 16\hbox{$^\prime$ }$ (see Fig. 1). Most previously known members of the ${\rho}$ Ophiuchi cluster lie within these fields. In particular, this is the case for 94 of a total of 113 recognized members from WLY89, AM94, Greene et al. (1994), and MAN98. The known young stars which lie outside the boundaries of our survey are mostly optically visible, weak-line or post T Tauri stars (belonging to Class III) spread over a large area on the outskirts of the molecular complex (e.g. Martín et al. 1998).

2.2 Observational details

The mapping was performed in the raster mode of ISOCAM in which the mid-IR $32\times32$ pixel array imaged the sky at consecutive positions along a series of scans parallel to the right-ascension axis. The offset between consecutive array positions along each scan ( $\Delta \alpha$) was 15 pixels, while the offset between two scans ( $\Delta \delta$) was 26 pixels. Each set of scans was then co-added and combined into a single raster image. The final image of the L1688 field (Fig. 1 and Abergel et al. 1996) actually results from the combination of six separate rasters. A pixel field of view of 6 $\hbox{$^{\prime\prime}$ }$ was used for four of these rasters, but smaller 3 $\hbox{$^{\prime\prime}$ }$ pixels were employed for the other two rasters in order to avoid saturating the array on the brightest sources of the cluster. The L1689N and L1689S fields were imaged with one raster each using 6 $\hbox{$^{\prime\prime}$ }$ pixels.

In order to avoid saturation, the individual readout time for the L1688 rasters was set to $\mbox{$t_\mathrm{int}$ }= 0.28\,$s. About 55 of these readouts (i.e. an integration time of $\sim$$15\,$s) were performed per sky position. Thanks to the half-frame overlap between subsequent individual images, each sky position was observed twice, yielding an effective total integration time of $\sim$$30\,$s. For L1689N and L1689S it was possible to use $\mbox{$t_\mathrm{int}$ }= 2.1\,$s, and about 15 readouts were performed per sky position with the same half-frame overlap, giving an integration time per sky position of $\sim$$60\,$s. A total of 1104 individual images were necessary to mosaic the L1688 field, and an additional 60 images each were used to map L1689N and L1689S. All three fields were mapped in two broad-band filters of ISOCAM: LW2 (5-8.5$\mu$m) and LW3 (12-18$\mu$m). These filters are approximately centered on two minima of the interstellar extinction curve and are situated apart from the silicate absorption bands (at roughly 10 and 18 $\mu$m). However, they include most of the Unidentified Infrared Bands (UIBs, likely due to PAH-like molecules) which constitute a major source of background emission toward star-forming clouds (e.g. Bernard et al. 1993; Boulanger et al. 1996). The ISOCAM central wavelengths adopted here for LW2 and LW3 are 6.7 $\mu$m and 14.3 $\mu$m respectively.

2.3 Image processing, source extraction, and photometry

Each raster consists of a temporal series of individual integration frames (i.e. of $32\times32$ pixel images) which was reduced using the CAM Interactive Analysis software (CIA). We have subtracted the best dark current from the ISOCAM calibration library, and as a second step we improved it with a second order correction using a FFT thresholding method (Starck et al. 1999). Cosmic-ray hits were detected and masked using the multi-resolution median transform algorithm (Starck et al. 1996). The transients in the time history of each pixel due to detector memory effects were corrected with the inversion method described in Abergel et al. (1996). The images were then flat-fielded with a flat image obtained from the observations themselves. Since these various corrections applied to the images are not perfect, the extraction of faint sources from the images is a difficult task. We have developed an interactive IDL point-source detection and photometry program for raster observations which works in the CIA environment. This program helps to discriminate between astronomical sources and remaining low-level glitches or ghosts due to strong transients (see also Nordh et al. 1996; Kaas et al. 2001). The fluxes of the detected sources were estimated from the series of flux measurements made in the individual images (usually 2 to 4 individual images cover each source) which were obtained from classical aperture photometry. The emission was integrated in a sky aperture, the background emission subtracted, and finally an appropriate aperture correction was applied based on observed point-spread functions available in the ISOCAM calibration library. In practice, the radius of the aperture used was 9 $\hbox{$^{\prime\prime}$ }$(i.e., 3 and 1.5 pixels for a pixel size of $3\hbox{$^{\prime\prime}$ }$ and 6 $\hbox{$^{\prime\prime}$ }$, respectively). For the weakest sources, however, we reduced the aperture radius to 4.5 $\hbox{$^{\prime\prime}$ }$ (i.e. 1.5 pixels for a pixel size of $3\hbox{$^{\prime\prime}$ }$), in order to improve the signal-to-noise ratio. Finally, we applied the following conversion factors: 2.33 and 1.97 ADU/gain/s/mJy for LW2 and LW3 respectively (from in-orbit latest calibration-Blommaert 1998). These calibration factors are strictly valid only for sources with a flat SED ( $F_\nu \sim \nu^{-1}$). Here, a small but significant ( $\lower.5ex\hbox{$\buildrel > \over \sim$ }$1%) color correction needs to be applied to the bluest sources, recognized as Class III YSOs in Sect. 3 below. For these sources, the conversion factors quoted above were divided by 1.05 for LW2 and 1.02 for LW3 to account for the color effect. The 212 ISOCAM sources recognized as cluster members (see Sect. 3 below) are listed in Table 1 (available only in electronic form at http://cdsweb.u-strasbg.fr/cgi-bin/qcat?/A+A/372/173) with their J2000 coordinates, their flux densities and associated rms uncertainties (see Sect. 2.4), as well as the corresponding near-IR identifications.

2.4 Photometric uncertainties and point-source sensitivity

The uncertainties on the final photometric measurements result from systematic errors due to uncertainties in the absolute calibration and the aperture correction factors, and from random errors associated with the flat-fielding noise, the statistical noise in the raw data, the noise due to remaining low-level glitches, and the imperfect correction for the transient behavior of the detectors. The in-orbit absolute calibration has been verified to be correct to within 5% (Blommaert 1998), and we estimate that the maximum systematic error on the aperture correction is $\sim$10% (by comparing theoretical and observed point-spread functions). The maximum systematic error on our photometry is thus $\sim$15%. The magnitudes of the random errors were directly estimated from the data by measuring both a "temporal'' noise (noise in the temporal sequence of individual integrations) and a "spatial'' noise (due to imperfect flat-fielding and/or spatial structures in the local mid-IR background emission) for each source in the automatic detection procedure. The temporal noise was computed as the standard deviation of the individual aperture measurements divided by the square root of the number of measurements. The spatial noise was estimated as the standard deviation around the mean background (linear combination of the median and the mean of the pixels optimized for the source flux estimates) in the immediate vicinity of each source.

Figure 2: a) Distribution of the $\mbox{$F_\nu^{6.7}$ }$ flux densities for the 425 sources detected by ISOCAM. The dark histogram shows the flux distribution of the cluster members (i.e., YSOs) discussed in Sect. 3, while the light histogram corresponds to the remaining sources which are likely dominated by background Galactic objects. b) Same as a) for the $\mbox{$F_\nu^{14.3}$ }$ flux densities. The dark histogram shows the flux distribution for the "red'' ISOCAM sources (i.e., Class I and II YSOs), while the light histogram comprises only the "blue'' ISOCAM sources not associated with recognized YSOs (see Sect. 3 and Fig. 3). The adopted completeness levels equal to 10 and $15\,$mJy for $\mbox{$F_\nu^{6.7}$ }$ and $\mbox{$F_\nu^{14.3}$ }$ respectively are shown as vertical dotted lines. The solid and dashed curves show the expected number of Galactic sources in each bin according to the model of Wainscoat et al. (1992) for AV = 0 and AV = 10 respectively.

The sensitivity limit of the survey was estimated by calculating the average value of the quadratic sum of the temporal and spatial noises measured on the weakest detected sources. The total rms flux uncertainty found in this way, $\sigma_{\rm tot} = (\sigma_{\rm temp}^2 +\sigma_{\rm spat}^2)^{1/2}$, is $\mbox{$\sigma_{6.7}$ }= 2.2$ mJy at 6.7 $\mu$m and $\mbox{$\sigma_{14.3}$ }= 4.1$ mJy at 14.3 $\mu$m, $\sim$75% of which is due to the spatial noise component. The large contribution of the spatial noise originates in the highly structured diffuse mid-IR emission from the ambient molecular cloud itself (see Abergel et al. 1996). Figure 2 displays the distributions of fluxes at 6.7 $\mu$m and 14.3 $\mu$m for all the detected ISOCAM sources. We used the Wainscoat et al. (1992) Galactic model of the mid-IR point source sky to estimate the expected number of foreground and background sources up to a distance of 20$\,$kpc. The model predictions are shown by solid and dashed curves in Fig. 2 for cloud extinctions of A_V = 0 and A_V = 10, respectively (see Kaas et al. 2001 for more details). It can be seen that the flux histograms of the ISOCAM sources not associated with YSOs (light shading in Fig. 2) are remarkably similar in shape to the model distributions down to $\sim$6 mJy at 6.7 $\mu$m and $\sim$10 mJy at 14.3 $\mu$m. These flux densities can be used to estimate the completeness level of our observations which is not uniform over the spatial extent of the survey. The histograms with light (grey) shading in Fig. 2 are dominated by background sources preferentially located in low-noise regions (i.e., outside the crowded central part of L1688), where the total rms flux uncertainty is $\sim$2.0 mJy at 6.7 $\mu$m and $\sim$3.5 mJy at 14.3 $\mu$m. The effective completeness level in these regions is thus $\sim$ $3\,\sigma_{\rm tot}$, where $\sigma_{\rm tot}$is the total flux uncertainty (see above). However, most of the YSOs are located in regions where the noise is somewhat larger. The largest rms noise is reached in the Oph A core area (see Fig. 1), where $\sigma_{\rm tot} \approx 3.4\,$mJy at 6.7 $\mu$m and $\sigma_{\rm tot} \approx 5.0\,$mJy at 14.3 $\mu$m. Therefore, we conservatively estimate the completeness levels of the global ISOCAM survey to be $\sim$10 mJy at 6.7 $\mu$m and $\sim$15 mJy at 14.3 $\mu$m.

Finally, we note that the A_V = 10 model curve in Fig. 2b accounts for essentially all the "blue'' sources detected at 14.3 $\mu$m and not associated with known YSOs. At 6.7 $\mu$m, the predictions of the Wainscoat et al. model suggest that there might still be a slight excess of $\sim$30 unidentified sources belonging to the cloud (Fig. 2a).