4.1. The colour-magnitude diagram
The CMD of M22 is presented in Fig. 4. It contains all stars brighter than and with r . The stars from the inner part have been omitted because of the high star density in the center of M22 and the resulting large photometric errors. The shown sample has been cleaned by error selection as done for M55.
The basic features of the red giant branch (RGB) and the blue horizontal branch (BHB) are well defined. In striking difference to M55 is the large colour spread among the red giant branch, as already mentioned by Arp & Melbourne (1959). A similar spread is also visible on the BHB, which suggests that differential reddening is the dominating effect for the origin of the colour dispersion in the CMD. However, since a contribution to the colour range by a possible metallicity spread in M22 can not be excluded, the CMD alone can not clarify the situation.
In the following, we select red giants by defining the ridges of the red giant branch by two parallel curves which at are separated in colour by 0.08 mag.
4.2. The diagram
Fig. 5 shows the diagram of M22. Selected are stars from the red giant branch with and . In sharp contrast to M55, the distribution of stars does not exhibit a well defined straight line but shows a strongly scattered distribution, so that an accurate determination of a mean cluster metallicity with the method described in Sect. 3.2 is not possible without knowing the reasons for the colour spread.
Previous measurements indicate that the foreground reddening in the direction to M22 is mag (e.g. Alcaino & Liller 1983, Crocker 1988). In addition, recent spectroscopic studies in M22 lead to a mean cluster metallicity of -1.48 dex (Carretta & Gratton 1997). The diagram shown in Fig. 5 is reddening corrected with , equivalent to the value given by Alcaino & Liller (1983). Using Hilker's calibration, some of the giant branch stars with lie in a range higher than -1.2 dex, so that, with respect to the expected mean metallicity, the giants seem to be scattered to higher metallicities. Since any additional reddening would further increase the metallicity, differential reddening is not a likely candidate in explaining the broad distribution. Moreover, we demonstrate directly that differential reddening does not play a role but that the CN band strengths are responsible.
Dividing the red giants in a red and a blue sample, as indicated in the upper panel of Fig. 6 by filled dots and open triangles, respectively, provides a simple means to assess the role of differential reddening. The lower panel shows the distribution projected on the sphere. The redder stars clearly tend to be on the southern side, especially along a narrow region from the center towards the south of the cluster. In contrast, many of the bluer giants are located on the north-eastern side. Subdividing the cluster into quarters we find an excess of red stars in the south-eastern quarter with a significance of 2. In contrast, there is a deficiency of red stars (at a 4 level) in the north-eastern quarter (see Fig. 6, lower panel). With regard to the cumulative distribution of red and blue stars in the azimuthal angle we obtain a probability of percent that the red and the blue giants are taken from a different distribution function (Kolmogorov-Smirnov test). This inhomogeneous distribution of the red and blue cluster giants indicates variations in the reddening, obviously caused by patchily structured dust in the foreground of M22. From the width of the giant branch (Fig. 6, upper panel) we estimate a total range of 0.07 mag for the reddening variations, after taking a photometric error in of 0.019 mag into account.
Fig. 7 provides an assessment of the appearance of "red" and "blue" red giants in the diagram, but no correlation between the scatter in this diagram and the colour spread in the CMD is visible. We would expect that "red" RGB stars show systematically lower metallicities, if reddening would be dominantly responsible for a stellar locus in the , diagram. Variable CN-abundances remain as a plausible explanation for the scatter.
4.3. CN abundances
The strong CN-band at 4216 Å can reduce the flux in the v filter significantly and thus may lead to a higher value, while the overall metallicity remains the same (Bell & Gustafsson 1978). The distribution in the diagram therefore may be an indication that the M22 giants exhibit strongly variing CN-bands. A large range of CN abundances in M22 has indeed been spectroscopically observed by NF.
60 giant stars from the sample of NF with known CN abundances have been identified and compared with our data. Table 5 shows these stars with the identification numbers from Arp & Melbourne (1959), magnitudes and colours from our photometry, and the CN band abundances from NF, given by the index . Values for the index vary between -0.14 (CN weak) and (CN strong) and the CN variations therefore are significantly higher than observed in M55. To investigate the influence of the CN strengths on the index in the case of M22, all CN strong stars with values larger than have been plotted in the diagram (Fig. 8, upper panel) as filled circles, while the CN weak stars () are shown as open circles. The correlation between the positions of the giants in the diagram and their CN abundances is clearly visible. All stars with strong CN abundances are located in the upper part of the diagram and it is reasonable to conclude that also all the other stars in this area, that have no spectroscopic measurements, are CN rich as well.
The influence of the cyanogen strengths on the index can be seen in more detail by using the metallicity calibration for the diagram. The mean Strömgren metallicity (using the calibration of Hilker) of all 60 giant stars with known CN abundances is -1.73 dex, taking a reddening value of into account (the CN strengths definitely influence this value; it is therefore too high). In Fig. 8 (lower panel), (as defined in Sect. 3.3) to the calibration line of -1.73 dex is plotted versus . In contrast to M55, the influence of the cyanogen strenghts on the index in M22 is significant. becomes larger by increasing CN abundances and the CN variations therefore shift the giant stars to higher values. This effect dominates the shape of the diagram, it is even stronger than the reddening variations of about 0.07 mag. Our results are similar to investigations of Anthony-Twarog et al. (1995), who investigated CN and Ca abundances in relation with the Strömgren index for M22. We compared 15 stars from their observations with our data stars and find a good agreement in the measured Strömgren colours. Mean deviations are , and (see also Sect. 3.3).
The fact that only strong cyanogen abundances influence the index offers us the possibility to determine the mean Strömgren metallicity and the mean reddening for M22 using only the 40 spectroscopically measured CN weak giants. With the method described in Sect. 3.2 we derive a cluster metallicity of dex and a reddening of . As for M55, the derived metallicity of M22 is (with regard to the expected errors) in agreement with recent spectroscopic results, such as from Carretta & Gratton (1997, dex) and Lehnert et al. (1991, dex).
© European Southern Observatory (ESO) 1999
Online publication: October 4, 1999