2. The color-magnitude diagram
The material used in this work are the J and F DPOSS plates of the field 278. For each band, we extracted from the whole digitized plate a sub-image (size: pixels), corresponding to an area of , with a pixel size of , centered on M 92 at coordinates (Harris 1996):
The two images were linearized by using a density-to-intensity (DtoI) calibration curve, provided by the sensitometric spots available on the DPOSS plates. The F plate is contaminated by two very similar satellite tracks (as an alternative, the two tracks come from a high altitude civil airplane) lying and from the cluster center and crossing the field in a South-East/North-West direction. The effect of these tracks can be seen as empty strips on the lower panel of Fig. 1. Other thin, fainter tracks and some galaxies are present on the same plate, but at larger distances from the cluster core region. We applied the CMD technique to datasets obtained with different astronomical packages, in order to test the reliability of object detection and photometry in crowded stellar fields. On the DPOSS plates containing M 92, we used both the SKICAT and DAOPHOT packages. SKICAT, written at Caltech (see Weir et al. 1995a, and refs. therein), is the standard software used by the CRoNaRio collaboration for the DPOSS plate processing and catalog construction. DAOPHOT is a well-tested program for stellar photometry, developed by Stetson (1987), and widely used by stellar astronomers. In this work we have used DAOPHOT only to obtain aperture photometry, with APPHOT, of objects detected with the DAOFIND algorithm on the DPOSS plates.
2.1. The data set
The SKICAT output catalog only contains objects classified as stars in both filters. For each object, we used (the magnitude computed from the central nine pixels), because the other aperture magnitude is measured on an area far too large for crowded regions. The final SKICAT catalog consists of 108779 objects. Since SKICAT is optimized for the detection of faint galaxies, in the present case we needed to test its performances in crowded stellar fields to ensure that it properly detected the stellar population around the cluster.
Thus, SKICAT has been compared to DAOPHOT, which is specifically designed for crowded fields stellar photometry and has been repeatedly tested in a variety of environments, including globular clusters. The DAOPHOT dataset was built using aperture photometry on the objects detected with the DAOFIND. The threshold was set at , similar to the one used by SKICAT. Aperture photometry was preferred to PSF fitting photometry, due to the large variability of the DPOSS point-spread function which makes the PSF photometry less accurate than the aperture photometry. We used an aperture of 1.69 pixels of radius, corresponding to an area of approximately 9 pixels, i.e. equivalent to the area used by FOCAS/SKICAT to compute . Indeed, the advantages of using PSF fitting are more evident in the central and more crowded regions of the cluster, while we are mainly interested in the outskirts, where crowding is less dramatic. Thus, we adopted the results from the aperture photometry, and we refer to this dataset as the DAOFIND+PHOT dataset.
The total number of objects detected in the J and F plates is, respectively, 240138 and 253977. The larger number of objects detected by DAOFIND, compared to those from SKICAT, is mainly due to the better capacity of DAOFIND in detecting objects in the crowded regions of the core. In the case of DAOFIND, since the convolution kernel, which is set essentially by the pixel size and seeing value, is much smaller than in SKICAT, we also have objects measured near the satellite tracks.
The FOCAS/SKICAT and DAOFIND+PHOT aperture photometry were then compared, and the results are shown in Fig. 2 where the SKICAT aperture magnitude is plotted versus the difference between itself and . The average difference is zero, with an error distribution typical of this kind of tests, i.e. a fan-like shape with growing dispersion at fainter magnitudes. The distribution of F magnitudes in Fig. 2 clearly shows the effects of saturation at the bright end, and in both plots there are several outliers, owing to the field crowding. In fact, these outliers are much more concentrated in the inner , where their density is 0.439 arcmin-2, than at larger distances from the center, where the density drops to 0.022 arcmin-2. These objects are mostly classified as non stellar by SKICAT and DAOPHOT, since they are either foreground galaxies or, more often, unresolved multiple objects, and were rejected in the final catalogs. Their area is taken into account later on, when we compute the effective area of the annuli in the construction of the radial profile. The outliers show an asymmetric distribution with SKICAT magnitudes being brighter at bright magnitudes and viceversa at fainter magnitudes. This is due to two reasons: in the case of large objects, SKICAT splits them into multiple entries, but keeps the value of the originally detected (big) object; at fainter magnitudes, where objects are small, is computed on a number of pixels less than 9, while the aperture photometry of the objects in the DAOFIND catalog are always computed on a circle of 1.69 pixel radius.However, the contribution of these outliers to the counts is far below 1 percent of the total, as can be seen from the histogram plotted on the right hand side of Fig. 2.
The above analysis shows that SKICAT catalogs are, after a suitable cleaning, usable "as they are" also for studies of moderately crowded stellar fields. We shall use the DPOSS-DAOPHOT dataset because it can better detect objects in highly crowded fields, which allows us to probe into the inner () regions of the cluster and merge our star counts profile with the published one of Trager et al. (1995, hereafter T95).
2.2. The color-magnitude diagram
In order to build the CMD of the cluster, individual catalogs were matched by adopting a searching radius of 5 ", and keeping the matched object with the smallest distance. The derived CMD is shown in Fig. 3 for two annular regions: the inner one, between 5´ and 12´ from the center (left-hand panel) and the outer one, referring to the background, between 60´ and 67´ (right-hand panel). The M 92 turn-off region, as well as part of the horizontal branch, are clearly visible. At bright magnitudes, the giant branch turns to the blue, due to plate saturation. At large angular distances from the center of M 92, most objects are galactic stars with only a small contribution from the cluster.
For reducing the background/foreground field contribution, we used an approach similar to that of Grillmair et al. 1995. First of all, We selected an annular region () around the cluster center to find the best fiducial CMD sequence of the cluster stars. Then, the CMD of this region was compared with the CMD of the field at a distance greater than from the cluster center. The two CMD's were normalized by their area and then we binned the CMD and computed the ratio of each element, just like in G95 (their Eq. 2). Finally, we then obtained the final contour of the best CMD region by cutting at a . Contours are shown in Fig. 3, on which a solid line marks the CMD region used to select the "bona fide" cluster stars as described above. We must say that this CMD selection is not aimed at finding all the stars in the cluster but only at the best possible enhancement of the cluster stars as compared to the field stars. This is why the region of the sub-giant/giant branch is not included, since here the cluster stars are fewer than in the field. By extracting objects at any distance from the center of the cluster, in the selected CMD region, the field contamination is reduced by a factor of . In absence of strong color gradients, the fraction of lost stars does not depend on the distance from the center.
© European Southern Observatory (ESO) 2000
Online publication: March 28, 2000