Next Contents Previous

3.3. Metallicity

We derive the Metallicity Distributions (MD) of the studied fields from the Color and Magnitude distribution of RGB stars (transformed to the absolute (V - I)0 vs. MI plane adopting the reddening and distance modulus described above) by interpolation on a grid of RGB ridgelines of template globular clusters, adopting the same scheme as Bellazzini et al. (2003). Essentially the same approach is adopted also by T04. The MDs are obtained in different metallicity scales, e.g. the ZW scale, the scale by Carretta & Gratton (1997, hereafter CG) and the global metallicity scale described in Ferraro et al. (1999), to make the comparison with other studies easier. While the individual photometric metallicities provided by the adopted procedure may be quite uncertain, the overall metallicity distribution and its average properties are sufficiently well characterized to provide interesting insights into the stellar population under analysis and it has been widely used, in recent years, in the study of resolved galaxies (see Bellazzini et al., 2003, for details, discussion and references). The color distribution of RGB stars should depend - to a lesser extent - also on the age distribution of the underlying population. T04 studied this problem in the case of M 33 by mean of synthetic CMDs drawn from theoretical evolutionary tracks and concluded that "...the ages of the RGB stars are not likely to significantly affect the derived MDs...". They estimate that the maximum expected shift of the peak of the MDs is of the order of 0.1-0.2 dex. Finally, the young MS population observed in F1 - and not in F2 - cannot affect the comparison between the MDs of the two fields since their evolved counterparts should be negligible in number and do not fall in the selection box we adopt to obtain MDs (see Fig. 5 and 7, below). When dealing with photometric metallicities it should be kept in mind that the underlying age distribution may affect the derived MDs.

In Fig. 5 a direct comparison between the observed RGBs of F1 and F2 and the adopted template ridge lines is presented. It is immediately clear that the large majority of M 33 RGB stars (in both fields) are enclosed within the ridge lines of M 5 ([Fe / H]CG = - 1.11; [M / H] = - 0.90) and of 47 Tuc ([Fe / H]CG = - 0.70; [M / H] = - 0.60). In the CMD of F2 an anomalous clustering of stars can be noted around the ridge line of NGC 6553 ([Fe / H]CG = - 0.16; [M / H] = - 0.06) that has no counterpart in the CMD of F1. However, as can be appreciated from Fig. 3, such red stars are just above the limiting magnitude at their color. For this reason we will not discuss in detail this feature in the following. Deeper photometry is needed to firmly assess the possible presence of an excess of very red RGB stars in this region.

Figure 5

Figure 5. The CMDs of the RGB stars in F1 (left panel) and F2 (right panel) are compared to the grid of RGB ridge lines of template globular clusters we adopted to derive the metallicity distributions. From blue to red, the template clusters are: M 92, M 13, M 5, 47 Tuc, NGC 6553 and NGC 6528. The thick box encloses the RGB stars actually selected for the derivation of the metallicity distribution (see Bellazzini et al., 2003, for details).

In Fig. 6 we present the MDs (in the form of generalized histograms) as a function of the global metallicity for F1 (upper left panel) and F2 (upper right panel). To study in finer detail the radial behavior of he MD we split F1 in two subregions (A and B) of similar area. In particular F1 A contains all the F1 stars less distant than 15' from the center of M 33, while the F1 stars with r geq 15' are assigned to F1 B. The MDs of F1 A and F1 B are plotted in the lower left and lower right panels of Fig. 6, respectively. Note that if we exclude from the interpolation the stars with MI geq - 2.5, i.e. in the range where the sensitivity of color to metallicity is lower, the obtained MDs are unchanged. This experiment also demonstrates that the derived MD are not sensitive to the effect of incompleteness, as expected. The average properties of all the considered MDs are summarized in Table 2, for two different assumptions of E(B-V). There are a number of considerations emerging from the inspection of Fig. 6 and Table 2:

Figure 6

Figure 6. Metallicity Distributions (continuous lines) of the F1 field as a whole (upper left panel), of F2 (upper right panel), of the inner part of F1 (10' ltapprox R ltapprox 15'; F1 A, lower left panel) and of the outer part of the same field (15' ltapprox R ltapprox 20'; F1 B, lower right panel). The dotted line in the upper left panel shows the instrumental response of the method, the MD that would be obtained if the width of the RGB was entirely due to the photometric errors.

  1. The MD of all the considered fields shows a strong peak at [M / H] appeq - 0.7. This justifies our assumption of the median metallicity as the characteristic value of the dominant population in our determination of the TRGB distance. A sparsely populated tail of metal-poor stars extending to [M / H] < - 2.0 is also present in all the presented MDs. This general similarity over large areas of the galaxy (F1 covers a range of galactocentric distances from ~ 2.4 Kpc to ~ 5 Kpc, F2 from ~ 5.5 Kpc to ~ 8.2 Kpc) is in agreement with the results by Kim et al. (2002) and is reminiscent of what is observed in M 31 (Bellazzini et al., 2003). Our MDs are very similar to those obtained by T04.

  2. The bell-shaped curve plotted as a dotted line in the upper left panel of Fig. 6 displays the response of the adopted interpolation scheme to a Simple Stellar Population (SSP, i.e. a population of stars having the same age and chemical composition, Renzini & Fusi Pecci, 1988) observed under the same conditions as our real data. It has been obtained adopting, as input for the interpolation, a synthetic RGB population whose color width is entirely due to the photometric errors. The I magnitude of the "synthetic" stars is extracted from the observed RGB luminosity function, the V-I color is obtained from the average ridge line of the observed RGB plus a photometric error drawn at random from a Gaussian distribution having sigma equal to the average observed photometric uncertainty at the considered magnitude (as done in Bellazzini et al., 2002).

    To obtain the true width of the underlying MD the described "instrumental response" should be deconvolved from the observed MD (continuous line). The main peak of the observed MD is well fitted by a Gaussian distribution with sigma = 0.34 dex, while the instrumental response curve is well approximated by a Gaussian distribution with sigma = 0.23 dex. It may be concluded that the true intrinsic dispersion of the main peak of the MD is sigma appeq 0.25 dex.

  3. By comparison with ridge lines of template globular clusters, Mould & Kristian (1986) estimated < [Fe / H]ZW > appeq -2.2 ± 0.8 for a field at a galactocentric distance similar to F2. On the other hand, we find, for both F1 and F2, < [Fe / H]ZW > appeq -1.03 ± 0.40, in excellent agreement with the results by Davidge (2003), Cuillandre, Lequeux & Loinard (1999) and T04. The difference is partly justified by the different assumptions about distance ((m - M0) = 24.8 instead of our (m - M0) = 24.64). However, even adopting their distance modulus we find < [Fe / H]ZW > appeq -1.22 ± 0.40, much more metal-rich than what was found by Mould & Kristian (1986). We tentatively ascribe this difference to a possible problem in the absolute calibration of Mould & Kristian's photometry. This hypothesis is confirmed by the results of T04.

  4. Assuming the same distance modulus as Mould & Kristian (1986), Kim et al. (2002) obtain -0.61 leq < [Fe / H]ZW > leq - 0.86 for 10 fields covering (approximately) the same radial range than our F1. The agreement with our results is much better than with Mould & Kristian's one, still the difference is not negligible and it is not justified by the different distance modulus assumed (a larger distance modulus should imply a brighter and hence more metal-poor RGB). However Kim et al. (2002) derived their mean metallicity from the mean (V - I)0 color at MI = - 3.5. We consider our median/mean metallicities, derived from all the RGB stars brighter than MI = - 2.0 and based on an accurately checked photometric calibration, as more robust and safer than those by Kim et al. (2002).

  5. Kim et al. (2002) found a weak radial gradient in the mean metallicity (0.2 dex) in the range 1 leq R leq 5 Kpc (see also T04). We find no sign of variation of the mean metallicity in the range 2.4 leq R leq 8.2 Kpc, but such feeble differences may have gone undetected at the level of accuracy of our relative photometry (e.g. may be hidden in the "instrumental width" of the observed sequences).

  6. The weak shoulder at [M / H] ~ - 0.2 in the MD of F2 is due to the handful of red stars around the ridge line of NGC 6553 discussed above, hence, at the present stage, cannot be trusted as a real feature of the MD (but it deserves further investigation).

Table 2. Distance Modulus, median Metallicity and standard deviations in different metallicity scales, and with different assumptions on the reddening. Note that the median and the mean metallicity are nearly coincident in all the considered cases.
  F1 F2 F1 F2
E(B-V) 0.04 0.08
         
[Fe / H]ZW -1.03 -1.03 -1.17 -1.18
sigmaZW 0.40 0.40 0.40 0.44
[Fe / H]CG -0.89 -0.89 -0.97 -0.98
sigmaCG 0.27 0.28 0.28 0.32
[M / H] -0.75 -0.74 -0.81 -0.81
sigmaM/H 0.23 0.23 0.23 0.27
         
(m - M)0 24.64 ± 0.15 24.59 ± 0.15

Next Contents Previous